SlideShare a Scribd company logo
1 of 27
Download to read offline
Common	
  Design	
  Elements	
  for	
  
Data	
  Movement	
  
Eli	
  Dart,	
  Network	
  Engineer	
  
ESnet	
  Science	
  Engagement	
  
Lawrence	
  Berkeley	
  Na<onal	
  Laboratory	
  
Cosmology	
  CrossConnects	
  Workshop	
  
Berkeley,	
  CA	
  
February	
  11,	
  2015	
  
Overview	
  
2/10/15	
  2	
  
•  Context	
  
•  Design	
  paJerns	
  
•  What	
  do	
  we	
  need	
  to	
  do?	
  
Context	
  
•  Data-­‐intensive	
  science	
  con<nues	
  to	
  need	
  high-­‐performance	
  data	
  movement	
  
between	
  geographically	
  distant	
  loca<ons	
  
–  Observa<on	
  (or	
  instrument)	
  to	
  analysis	
  
–  Distribu<on	
  of	
  data	
  products	
  to	
  users	
  
–  Aggrega<on	
  of	
  data	
  sets	
  for	
  analysis	
  
–  Replica<on	
  to	
  archival	
  storage	
  
•  Move	
  computa<on	
  to	
  data?	
  	
  Of	
  course!	
  	
  Except	
  when	
  you	
  can’t…	
  
–  A	
  liquid	
  market	
  in	
  fungible	
  compu<ng	
  alloca<ons	
  does	
  not	
  exist	
  
–  Users	
  get	
  an	
  alloca<on	
  of	
  <me	
  on	
  a	
  specific	
  compute	
  resource	
  –	
  if	
  the	
  data	
  
isn’t	
  there	
  already,	
  it	
  needs	
  to	
  be	
  put	
  there	
  
–  If	
  data	
  can’t	
  be	
  stored	
  long-­‐term	
  where	
  it’s	
  generated,	
  it	
  must	
  be	
  moved	
  
–  Other	
  reasons	
  too	
  –	
  the	
  point	
  is	
  we	
  have	
  to	
  be	
  able	
  to	
  move	
  Big	
  Data	
  
•  Given	
  the	
  need	
  for	
  data	
  movement,	
  how	
  can	
  we	
  reliably	
  do	
  it	
  well?	
  
2/10/15	
  3	
  
The	
  Task	
  of	
  Large	
  Scale	
  Data	
  Movement	
  
•  Several	
  different	
  ways	
  to	
  look	
  at	
  a	
  data	
  movement	
  task	
  
•  People	
  perspec<ve:	
  
–  I	
  am	
  a	
  member	
  of	
  a	
  collabora<on	
  
–  Our	
  collabora<on	
  has	
  accounts	
  with	
  compute	
  alloca<ons	
  and	
  data	
  
storage	
  alloca<ons	
  at	
  a	
  set	
  of	
  sites	
  
–  I	
  need	
  to	
  move	
  data	
  between	
  those	
  sites	
  
•  Organiza<on/facility	
  perspec<ve:	
  
–  ANL,	
  NCSA,	
  NERSC,	
  ORNL	
  and	
  SDSC	
  are	
  all	
  used	
  by	
  the	
  collabora<on	
  
–  All	
  these	
  sites	
  must	
  have	
  data	
  transfer	
  tools	
  in	
  common	
  
–  I	
  must	
  learn	
  what	
  tools	
  and	
  capabili<es	
  each	
  site	
  has,	
  and	
  apply	
  those	
  
tools	
  to	
  my	
  task	
  
•  Note	
  that	
  the	
  integra<on	
  burden	
  is	
  on	
  the	
  scien<st!	
  
2/10/15	
  4	
  
Service	
  Primi<ves	
  
•  There	
  is	
  another	
  way	
  to	
  look	
  at	
  data	
  movement	
  
•  All	
  large-­‐scale	
  data	
  movement	
  tasks	
  are	
  composed	
  of	
  a	
  set	
  of	
  primi<ves	
  
–  Those	
  primi<ves	
  are	
  common	
  to	
  most	
  such	
  workflows	
  
–  If	
  major	
  sites	
  can	
  agree	
  on	
  a	
  set	
  of	
  primi<ves,	
  all	
  large-­‐scale	
  data	
  workflows	
  
will	
  benefit	
  
•  What	
  are	
  the	
  common	
  primi<ves?	
  
–  Storage	
  systems	
  (filesystems,	
  tape	
  archives,	
  etc.)	
  
–  Data	
  transfer	
  applica<ons	
  (Globus,	
  others)	
  
–  Workflow	
  tools,	
  if	
  automa<on	
  is	
  used	
  
–  Networks	
  
•  Local	
  networks	
  
•  Wide	
  area	
  networks	
  
•  What	
  if	
  these	
  worked	
  well	
  together	
  in	
  the	
  general	
  case?	
  
•  Compose	
  them	
  into	
  common	
  design	
  paJerns	
  
2/10/15	
  5	
  
The	
  Central	
  Role	
  of	
  the	
  Network	
  
•  The	
  very	
  structure	
  of	
  modern	
  science	
  assumes	
  science	
  networks	
  exist:	
  high	
  
performance,	
  feature	
  rich,	
  global	
  scope	
  
•  What	
  is	
  “The	
  Network”	
  anyway?	
  
–  “The	
  Network”	
  is	
  the	
  set	
  of	
  devices	
  and	
  applica<ons	
  involved	
  in	
  the	
  use	
  of	
  a	
  
remote	
  resource	
  
•  This	
  is	
  not	
  about	
  supercomputer	
  interconnects	
  
•  This	
  is	
  about	
  data	
  flow	
  from	
  experiment	
  to	
  analysis,	
  between	
  facili<es,	
  etc.	
  
–  User	
  interfaces	
  for	
  “The	
  Network”	
  –	
  portal,	
  data	
  transfer	
  tool,	
  workflow	
  engine	
  
–  Therefore,	
  servers	
  and	
  applica<ons	
  must	
  also	
  be	
  considered	
  
•  What	
  is	
  important?	
  	
  Ordered	
  list:	
  
1.  Correctness	
  
2.  Consistency	
  
3.  Performance	
  
©	
  2014,	
  Energy	
  Sciences	
  Network	
  
6 – ESnet Science Engagement (engage@es.net) - 2/10/15
TCP	
  –	
  Ubiquitous	
  and	
  Fragile	
  
•  Networks	
  provide	
  connec<vity	
  between	
  hosts	
  –	
  how	
  do	
  hosts	
  see	
  the	
  
network?	
  
–  From	
  an	
  applica<on’s	
  perspec<ve,	
  the	
  interface	
  to	
  “the	
  other	
  end”	
  is	
  a	
  
socket	
  
–  Communica<on	
  is	
  between	
  applica<ons	
  –	
  mostly	
  over	
  TCP	
  
•  TCP	
  –	
  the	
  fragile	
  workhorse	
  
–  TCP	
  is	
  (for	
  very	
  good	
  reasons)	
  <mid	
  –	
  packet	
  loss	
  is	
  interpreted	
  as	
  
conges<on	
  
–  Packet	
  loss	
  in	
  conjunc<on	
  with	
  latency	
  is	
  a	
  performance	
  killer	
  
–  Like	
  it	
  or	
  not,	
  TCP	
  is	
  used	
  for	
  the	
  vast	
  majority	
  of	
  data	
  transfer	
  
applica<ons	
  (more	
  than	
  95%	
  of	
  ESnet	
  traffic	
  is	
  TCP)	
  
©	
  2014,	
  Energy	
  Sciences	
  Network	
  
7 – ESnet Science Engagement (engage@es.net) - 2/10/15
A small amount of packet loss makes a huge
difference in TCP performance
Metro	
  Area	
  
Local	
  
(LAN)	
  
Regional	
  
Con<nental	
  
Interna<onal	
  
Measured (TCP Reno) Measured (HTCP) Theoretical (TCP Reno) Measured (no loss)
With loss, high performance
beyond metro distances is
essentially impossible
©	
  2014,	
  Energy	
  Sciences	
  Network	
  
8 – ESnet Science Engagement (engage@es.net) - 2/10/15
Design	
  PaGern	
  –	
  The	
  Science	
  DMZ	
  Model	
  
•  Design	
  paJerns	
  are	
  reusable	
  solu<ons	
  to	
  design	
  problems	
  that	
  recur	
  in	
  the	
  real	
  
world	
  
–  High	
  performance	
  data	
  movement	
  is	
  a	
  good	
  fit	
  for	
  this	
  
–  Science	
  DMZ	
  model	
  
•  Science	
  DMZ	
  incorporates	
  several	
  things	
  
–  Network	
  enclave	
  at	
  or	
  near	
  site	
  perimeter	
  
–  Sane	
  security	
  controls	
  
•  Good	
  fit	
  for	
  high-­‐performance	
  applica<ons	
  
•  Specific	
  to	
  Science	
  DMZ	
  services	
  
–  Performance	
  test	
  and	
  measurement	
  
–  Dedicated	
  systems	
  for	
  data	
  transfer	
  (Data	
  Transfer	
  Nodes)	
  
•  High	
  performance	
  hosts	
  
•  Good	
  tools	
  
•  Details	
  at	
  hJp://fasterdata.es.net/science-­‐dmz/	
  	
  
2/10/15	
  9	
  
Context:	
  Science	
  DMZ	
  Adop<on	
  
•  DOE	
  Na<onal	
  Laboratories	
  
–  Both	
  large	
  and	
  small	
  sites	
  
–  HPC	
  centers,	
  LHC	
  sites,	
  experimental	
  facili<es	
  
•  NSF	
  CC-­‐NIE	
  and	
  CC*IIE	
  programs	
  leverage	
  Science	
  DMZ	
  
–  $40M	
  and	
  coun<ng	
  (third	
  round	
  awards	
  coming	
  soon,	
  es<mate	
  addi<onal	
  $18M	
  to	
  $20M)	
  
–  Significant	
  investments	
  across	
  the	
  US	
  university	
  complex,	
  ~130	
  awards	
  
–  Big	
  shoutout	
  to	
  Kevin	
  Thompson	
  and	
  the	
  NSF	
  –	
  these	
  programs	
  are	
  cri<cally	
  important	
  
•  Na<onal	
  Ins<tutes	
  of	
  Health	
  
–  100G	
  network	
  infrastructure	
  refresh	
  
•  US	
  Department	
  of	
  Agriculture	
  
–  Agricultural	
  Research	
  Service	
  is	
  building	
  a	
  new	
  science	
  network	
  based	
  on	
  the	
  Science	
  DMZ	
  model	
  
–  hJps://www.ro.gov/index?s=opportunity&mode=form&tab=core&id=a7f291f4216b5a24c1177a5684e1809b	
  
•  Other	
  US	
  agencies	
  looking	
  at	
  Science	
  DMZ	
  model	
  
–  NASA	
  
–  NOAA	
  
•  Australian	
  Research	
  Data	
  Storage	
  Infrastructure	
  (RDSI)	
  
–  Science	
  DMZs	
  at	
  major	
  sites,	
  connected	
  by	
  a	
  high	
  speed	
  network	
  
–  hJps://www.rdsi.edu.au/dashnet	
  
–  hJps://www.rdsi.edu.au/dashnet-­‐deployment-­‐rdsi-­‐nodes-­‐begins	
  
•  Other	
  countries	
  
–  Brazil	
  
–  New	
  Zealand	
  
–  More	
  
	
  
2/10/15	
  10	
  
Context:	
  Community	
  Capabili<es	
  
•  Many	
  Science	
  DMZs	
  directly	
  support	
  science	
  applica<ons	
  
–  LHC	
  (Run	
  2	
  is	
  coming	
  soon)	
  
–  Experiment	
  opera<on	
  (Fusion,	
  Light	
  Sources,	
  etc.)	
  
–  Data	
  transfer	
  into/out	
  of	
  HPC	
  facili<es	
  
•  Many	
  Science	
  DMZs	
  are	
  SDN-­‐ready	
  
–  Openflow-­‐capable	
  gear	
  
–  SDN	
  research	
  ongoing	
  
•  High-­‐performance	
  components	
  
–  High-­‐speed	
  WAN	
  connec<vity	
  
–  perfSONAR	
  deployments	
  
–  DTN	
  deployments	
  
•  Metcalfe’s	
  Law	
  of	
  Network	
  U<lity	
  
–  Value	
  propor<onal	
  to	
  the	
  square	
  of	
  the	
  number	
  of	
  DMZs?	
  n	
  log(n)?	
  
–  Cyberinfrastructure	
  value	
  increases	
  as	
  we	
  all	
  upgrade	
  
2/10/15	
  11	
  
Strategic	
  Impacts	
  
•  What	
  does	
  this	
  mean?	
  
–  We	
  are	
  in	
  the	
  midst	
  of	
  a	
  significant	
  cyberinfrastructure	
  upgrade	
  
–  Enterprise	
  networks	
  need	
  not	
  be	
  unduly	
  perturbed	
  J	
  
•  Significantly	
  enhanced	
  capabili<es	
  compared	
  to	
  3	
  years	
  ago	
  
–  Terabyte-­‐scale	
  data	
  movement	
  is	
  much	
  easier	
  
–  Petabyte-­‐scale	
  data	
  movement	
  possible	
  outside	
  the	
  LHC	
  experiments	
  
•  3.1Gbps	
  =	
  1PB/month	
  
•  (Try	
  doing	
  that	
  through	
  your	
  enterprise	
  firewall!)	
  
–  Widely-­‐deployed	
  tools	
  are	
  much	
  beJer	
  (e.g.	
  Globus)	
  
•  Raised	
  expecta<ons	
  for	
  network	
  infrastructures	
  
–  Scien<sts	
  should	
  be	
  able	
  to	
  do	
  beJer	
  than	
  residen<al	
  broadband	
  	
  
•  Many	
  more	
  sites	
  can	
  now	
  achieve	
  good	
  performance	
  
•  Incumbent	
  on	
  science	
  networks	
  to	
  meet	
  the	
  challenge	
  
–  Remember	
  the	
  TCP	
  loss	
  characteris<cs	
  
–  Use	
  perfSONAR	
  
–  Science	
  experiments	
  assume	
  this	
  stuff	
  works	
  –	
  we	
  can	
  now	
  meet	
  their	
  needs	
  
	
  
2/10/15	
  12	
  
High	
  Performance	
  Data	
  Transfer	
  -­‐	
  Requirements	
  
•  There	
  is	
  a	
  set	
  of	
  things	
  required	
  for	
  reliable	
  high-­‐performance	
  data	
  transfer	
  
–  Long-­‐haul	
  networks	
  	
  
•  Well-­‐provisioned	
  
•  High-­‐performance	
  
–  Local	
  networks	
  
•  Well-­‐provisioned	
  
•  High-­‐performance	
  
•  Sane	
  security	
  
–  Local	
  data	
  systems	
  
•  Dedicated	
  to	
  data	
  transfer	
  (else	
  too	
  much	
  complexity)	
  
•  High-­‐performance	
  access	
  to	
  storage	
  
–  Good	
  data	
  transfer	
  tools	
  
•  Interoperable	
  
•  High-­‐performance	
  
–  Ease	
  of	
  use	
  
•  Usable	
  by	
  people	
  
•  Usable	
  by	
  workflows	
  
•  Interoperable	
  across	
  sites	
  (remove	
  integra<on	
  burden)	
  
2/10/15	
  13	
  
Long-­‐Haul	
  Network	
  Status	
  
•  100	
  Gigabit	
  per	
  second	
  networks	
  deployed	
  globally	
  
–  USA/DOE	
  Na<onal	
  Laboratories	
  –	
  ESnet	
  
–  USA/.edu	
  –	
  Internet2	
  
–  Europe	
  –	
  GEANT	
  
–  Many	
  state	
  and	
  regional	
  networks	
  have	
  or	
  are	
  deploying	
  100Gbps	
  cores	
  
•  What	
  does	
  this	
  mean	
  in	
  terms	
  of	
  capability?	
  
–  1TB/hour	
  requires	
  less	
  than	
  2.5Gbps	
  (2.5%	
  of	
  100Gbps	
  network)	
  
–  1PB/week	
  requires	
  less	
  than	
  15Gbps	
  (15%	
  of	
  100Gbps	
  network)	
  
–  hJp://fasterdata.es.net/home/requirements-­‐and-­‐expecta<ons	
  
–  The	
  long-­‐haul	
  capacity	
  problem	
  is	
  now	
  solved,	
  to	
  first	
  order	
  
•  Some	
  networks	
  are	
  s<ll	
  in	
  the	
  middle	
  of	
  upgrades	
  
•  However,	
  steady	
  progress	
  is	
  being	
  made	
  
2/10/15	
  14	
  
Local	
  Network	
  Status	
  
•  Many	
  ESnet	
  sites	
  now	
  have	
  100G	
  connec<ons	
  to	
  ESnet	
  
–  2x100G:	
  BNL,	
  CERN,	
  FNAL	
  
–  1x100G:	
  ANL,	
  LANL,	
  LBNL,	
  NERSC,	
  ORNL,	
  SLAC	
  
•  Capacity	
  provisioning	
  is	
  much	
  easier	
  in	
  a	
  LAN	
  environment	
  
•  Security	
  requires	
  aJen<on	
  (see	
  Science	
  DMZ)	
  
•  Major	
  DOE	
  compu<ng	
  facili<es	
  have	
  a	
  lot	
  of	
  capacity	
  deployed	
  to	
  their	
  data	
  
systems	
  
–  ANL:	
  60Gbps	
  
–  NERSC:	
  80Gbps	
  
–  ORNL:	
  20Gbps	
  
•  Big	
  win	
  if	
  sites	
  use	
  Science	
  DMZ	
  model	
  
2/10/15	
  15	
  
Progress	
  So	
  Far	
  
•  There	
  is	
  a	
  set	
  of	
  things	
  required	
  for	
  reliable	
  high-­‐performance	
  data	
  transfer	
  
–  Long-­‐haul	
  networks	
  	
  
•  Well-­‐provisioned	
  
•  High-­‐performance	
  
–  Local	
  networks	
  
•  Well-­‐provisioned	
  
•  High-­‐performance	
  
•  Sane	
  security	
  
–  Local	
  data	
  systems	
  
•  Dedicated	
  to	
  data	
  transfer	
  (else	
  too	
  much	
  complexity)	
  
•  High-­‐performance	
  access	
  to	
  storage	
  
–  Good	
  data	
  transfer	
  tools	
  
•  Interoperable	
  
•  High-­‐performance	
  
–  Ease	
  of	
  use	
  
•  Usable	
  by	
  people	
  
•  Usable	
  by	
  workflows	
  
•  Interoperable	
  across	
  sites	
  (remove	
  integra<on	
  burden)	
  
2/10/15	
  16	
  
Local	
  Data	
  Systems	
  
•  Science	
  DMZ	
  model	
  calls	
  these	
  Data	
  Transfer	
  Nodes	
  
–  Dedicated	
  to	
  high-­‐performance	
  data	
  transfer	
  tasks	
  
–  Short,	
  clean	
  path	
  to	
  outside	
  world	
  
•  At	
  HPC	
  facili<es,	
  they	
  mount	
  the	
  global	
  filesystem	
  
–  Transfer	
  data	
  to	
  the	
  DTN	
  
–  Data	
  available	
  on	
  HPC	
  resource	
  
•  High-­‐performance	
  data	
  transfer	
  tools	
  
–  Globus	
  Transfer	
  
–  Command-­‐line	
  globus-­‐url-­‐copy	
  
–  BBCP	
  
•  These	
  are	
  deployed	
  now	
  at	
  many	
  HPC	
  facili<es	
  
–  ANL,	
  NERSC,	
  ORNL	
  
–  NCSA,	
  SDSC	
  
2/10/15	
  17	
  
Data	
  Transfer	
  Tools	
  
•  Interoperability	
  is	
  really	
  important	
  
–  Remember,	
  scien<sts	
  should	
  not	
  have	
  to	
  do	
  the	
  integra<on	
  
–  HPC	
  facili<es	
  should	
  agree	
  on	
  a	
  common	
  toolset	
  
–  Today,	
  that	
  common	
  toolset	
  has	
  a	
  few	
  members	
  
•  Globus	
  Transfer	
  
•  SSH/SCP/Rsync	
  (yes,	
  I	
  know	
  –	
  ick!)	
  
•  Many	
  niche	
  tools	
  
•  Globus	
  appears	
  to	
  be	
  the	
  most	
  full-­‐featured	
  
–  GUI,	
  data	
  integrity	
  checks,	
  fault	
  recovery	
  
–  Fire	
  and	
  forget	
  
–  API	
  for	
  workflows	
  
•  Globus	
  is	
  also	
  widely	
  deployed	
  
–  ANL,	
  NERSC,	
  ORNL	
  
–  NCSA,	
  SDSC	
  (all	
  of	
  XSEDE)	
  
–  Many	
  other	
  loca<ons	
  
2/10/15	
  18	
  
More	
  Progress	
  
•  There	
  is	
  a	
  set	
  of	
  things	
  required	
  for	
  reliable	
  high-­‐performance	
  data	
  transfer	
  
–  Long-­‐haul	
  networks	
  	
  
•  Well-­‐provisioned	
  
•  High-­‐performance	
  
–  Local	
  networks	
  
•  Well-­‐provisioned	
  
•  High-­‐performance	
  
•  Sane	
  security	
  
–  Local	
  data	
  systems	
  
•  Dedicated	
  to	
  data	
  transfer	
  (else	
  too	
  much	
  complexity)	
  
•  High-­‐performance	
  access	
  to	
  storage	
  
–  Good	
  data	
  transfer	
  tools	
  
•  Interoperable	
  
•  High-­‐performance	
  
–  Ease	
  of	
  use	
  
•  Usable	
  by	
  people	
  
•  Usable	
  by	
  workflows	
  
•  Interoperable	
  across	
  sites	
  (remove	
  integra<on	
  burden)	
  
2/10/15	
  19	
  
Mission	
  Scope	
  and	
  Science	
  Support	
  
•  Resource	
  providers	
  each	
  have	
  their	
  own	
  mission	
  
–  ESnet:	
  high-­‐performance	
  networking	
  for	
  science	
  
–  ANL,	
  NERSC,	
  ORNL:	
  HPC	
  for	
  DOE	
  science	
  users	
  
–  NCSA,	
  SDSC,	
  et.	
  al.:	
  HPC	
  for	
  NSF	
  users	
  
–  Globus:	
  full-­‐featured,	
  high-­‐performance	
  data	
  transfer	
  tools	
  
•  No	
  responsibility	
  for	
  individual	
  science	
  projects	
  
–  Resource	
  provider	
  staff	
  usually	
  not	
  part	
  of	
  science	
  projects	
  
–  Science	
  projects	
  have	
  to	
  do	
  their	
  own	
  integra<on	
  (see	
  beginning	
  of	
  talk)	
  
•  However,	
  resource	
  providers	
  are	
  typically	
  responsive	
  to	
  user	
  requests	
  
–  If	
  you	
  have	
  a	
  problem,	
  it’s	
  their	
  job	
  to	
  fix	
  it	
  
–  I	
  propose	
  we	
  use	
  this	
  to	
  get	
  something	
  done	
  
2/10/15	
  20	
  
Hypothe<cal:	
  HPC	
  Data	
  Transfer	
  Capability	
  
•  This	
  community	
  has	
  significant	
  data	
  transfer	
  needs	
  
–  I	
  have	
  worked	
  with	
  some	
  of	
  you	
  in	
  the	
  past	
  
–  Simula<ons,	
  sky	
  surveys,	
  etc.	
  
–  Expecta<on	
  over	
  <me	
  that	
  needs	
  will	
  increase	
  
•  Improve	
  data	
  movement	
  capability	
  
–  ANL,	
  NERSC,	
  ORNL	
  
–  NCSA,	
  SDSC	
  
–  This	
  is	
  an	
  arbitrary	
  list,	
  based	
  on	
  my	
  incomplete	
  understanding	
  
–  Should	
  there	
  be	
  others?	
  
•  Goal:	
  per-­‐Globus-­‐job	
  performance	
  of	
  1PB/week	
  
–  I	
  don’t	
  mean	
  we	
  have	
  to	
  transfer	
  1PB	
  every	
  week	
  
–  But,	
  if	
  we	
  need	
  to,	
  we	
  should	
  be	
  able	
  to	
  
–  Remember,	
  this	
  only	
  takes	
  15%	
  of	
  a	
  100G	
  network	
  path	
  
2/10/15	
  21	
  
What	
  Would	
  Be	
  Required?	
  
•  We	
  would	
  need	
  several	
  things:	
  
–  Specific	
  workflow	
  (move	
  dataset	
  D	
  of	
  size	
  S	
  from	
  A	
  to	
  Z,	
  frequency	
  F)	
  
–  A	
  commitment	
  by	
  resource	
  providers	
  to	
  see	
  it	
  through	
  
•  ESnet	
  (+	
  other	
  networks	
  if	
  needed)	
  
•  Compu<ng	
  facili<es	
  
•  Globus	
  
•  Is	
  it	
  100%	
  plug-­‐and-­‐play?	
  	
  No.	
  
–  There	
  are	
  almost	
  certainly	
  some	
  wrinkles	
  
–  However,	
  most	
  of	
  the	
  hard	
  part	
  is	
  done	
  
•  Networks	
  
•  Data	
  transfer	
  nodes	
  
•  Tools	
  
•  Let’s	
  work	
  together	
  and	
  make	
  this	
  happen!	
  
2/10/15	
  22	
  
Ques<ons	
  For	
  You	
  
•  Would	
  an	
  effort	
  like	
  this	
  be	
  useful?	
  (I	
  think	
  so)	
  
•  Does	
  this	
  community	
  need	
  this	
  capability?	
  	
  (I	
  think	
  so)	
  
•  Are	
  there	
  obvious	
  gaps?	
  (probably,	
  e.g.	
  performance	
  to	
  tape)	
  
•  Which	
  sites	
  would	
  be	
  involved?	
  
•  Am	
  I	
  crazy?	
  (I	
  think	
  not)	
  
2/10/15	
  23	
  
Thanks!	
  
Eli	
  Dart	
  
Energy	
  Sciences	
  Network	
  (ESnet)	
  
Lawrence	
  Berkeley	
  Na<onal	
  Laboratory	
  
hJp://fasterdata.es.net/	
  
hJp://my.es.net/	
  
hJp://www.es.net/	
  
Extra	
  Slides	
  
2/10/15	
  25	
  
Support	
  For	
  Science	
  Traffic	
  
•  The	
  Science	
  DMZ	
  is	
  typically	
  deployed	
  to	
  support	
  science	
  traffic	
  
–  Typically	
  large	
  data	
  transfers	
  over	
  long	
  distances	
  
–  In	
  most	
  cases,	
  the	
  data	
  transfer	
  applica<ons	
  use	
  TCP	
  
•  The	
  behavior	
  of	
  TCP	
  is	
  a	
  legacy	
  from	
  the	
  conges<on	
  collapse	
  of	
  the	
  Internet	
  in	
  
the	
  1980s	
  
–  Loss	
  is	
  interpreted	
  as	
  conges<on	
  
–  TCP	
  backs	
  off	
  to	
  avoid	
  conges<on	
  à	
  performance	
  degrades	
  
–  Performance	
  hit	
  related	
  to	
  the	
  square	
  of	
  the	
  packet	
  loss	
  rate	
  
•  Addressing	
  this	
  problem	
  is	
  a	
  dominant	
  engineering	
  considera<on	
  for	
  science	
  
networks	
  
–  Lots	
  of	
  design	
  effort	
  
–  Lots	
  of	
  engineering	
  <me	
  
–  Lots	
  of	
  troubleshoo<ng	
  effort	
  
2/10/15	
  26	
  
A small amount of packet loss makes a huge
difference in TCP performance
2/10/15	
  
Metro	
  Area	
  
Local	
  
(LAN)	
  
Regional	
  
Con<nental	
  
Interna<onal	
  
Measured (TCP Reno) Measured (HTCP) Theoretical (TCP Reno) Measured (no loss)
With loss, high performance
beyond metro distances is
essentially impossible

More Related Content

What's hot

Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...Tal Lavian Ph.D.
 
Dynamic Data Center concept
Dynamic Data Center concept  Dynamic Data Center concept
Dynamic Data Center concept Miha Ahronovitz
 
Characterization and prediction of resource availability in grids
Characterization and prediction of resource availability in gridsCharacterization and prediction of resource availability in grids
Characterization and prediction of resource availability in gridsIAEME Publication
 
Efficient node bootstrapping for decentralised shared-nothing Key-Value Stores
Efficient node bootstrapping for decentralised shared-nothing Key-Value StoresEfficient node bootstrapping for decentralised shared-nothing Key-Value Stores
Efficient node bootstrapping for decentralised shared-nothing Key-Value StoresHan Li
 
Challenges and Issues of Next Cloud Computing Platforms
Challenges and Issues of Next Cloud Computing PlatformsChallenges and Issues of Next Cloud Computing Platforms
Challenges and Issues of Next Cloud Computing PlatformsFrederic Desprez
 
Practical Considerations for Deploying a Java Active Networking Platform
Practical Considerations for Deploying a Java Active Networking PlatformPractical Considerations for Deploying a Java Active Networking Platform
Practical Considerations for Deploying a Java Active Networking PlatformTal Lavian Ph.D.
 
DOE Magellan OpenStack user story
DOE Magellan OpenStack user storyDOE Magellan OpenStack user story
DOE Magellan OpenStack user storylaurabeckcahoon
 
Janet Network R&D Innovation - HEAnet / Juniper Innovation Day
Janet Network R&D Innovation - HEAnet / Juniper Innovation DayJanet Network R&D Innovation - HEAnet / Juniper Innovation Day
Janet Network R&D Innovation - HEAnet / Juniper Innovation DayMartin Hamilton
 
Provisioning Janet
Provisioning JanetProvisioning Janet
Provisioning JanetJisc
 
Synergy 2014 - Syn122 Moving Australian National Research into the Cloud
Synergy 2014 - Syn122 Moving Australian National Research into the CloudSynergy 2014 - Syn122 Moving Australian National Research into the Cloud
Synergy 2014 - Syn122 Moving Australian National Research into the CloudCitrix
 
Archiving data from Durham to RAL using the File Transfer Service (FTS)
Archiving data from Durham to RAL using the File Transfer Service (FTS)Archiving data from Durham to RAL using the File Transfer Service (FTS)
Archiving data from Durham to RAL using the File Transfer Service (FTS)Jisc
 
Welcome ndm11
Welcome ndm11Welcome ndm11
Welcome ndm11balmanme
 
Network-aware Data Management for High Throughput Flows Akamai, Cambridge, ...
Network-aware Data Management for High Throughput Flows   Akamai, Cambridge, ...Network-aware Data Management for High Throughput Flows   Akamai, Cambridge, ...
Network-aware Data Management for High Throughput Flows Akamai, Cambridge, ...balmanme
 
A location based least-cost scheduling for data-intensive applications
A location based least-cost scheduling for data-intensive applicationsA location based least-cost scheduling for data-intensive applications
A location based least-cost scheduling for data-intensive applicationsIAEME Publication
 
Update on the Exascale Computing Project (ECP)
Update on the Exascale Computing Project (ECP)Update on the Exascale Computing Project (ECP)
Update on the Exascale Computing Project (ECP)inside-BigData.com
 
Data-intensive bioinformatics on HPC and Cloud
Data-intensive bioinformatics on HPC and CloudData-intensive bioinformatics on HPC and Cloud
Data-intensive bioinformatics on HPC and CloudOla Spjuth
 

What's hot (20)

Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
 
ApacheCon NA 2013
ApacheCon NA 2013ApacheCon NA 2013
ApacheCon NA 2013
 
Dynamic Data Center concept
Dynamic Data Center concept  Dynamic Data Center concept
Dynamic Data Center concept
 
Characterization and prediction of resource availability in grids
Characterization and prediction of resource availability in gridsCharacterization and prediction of resource availability in grids
Characterization and prediction of resource availability in grids
 
Efficient node bootstrapping for decentralised shared-nothing Key-Value Stores
Efficient node bootstrapping for decentralised shared-nothing Key-Value StoresEfficient node bootstrapping for decentralised shared-nothing Key-Value Stores
Efficient node bootstrapping for decentralised shared-nothing Key-Value Stores
 
Challenges and Issues of Next Cloud Computing Platforms
Challenges and Issues of Next Cloud Computing PlatformsChallenges and Issues of Next Cloud Computing Platforms
Challenges and Issues of Next Cloud Computing Platforms
 
Practical Considerations for Deploying a Java Active Networking Platform
Practical Considerations for Deploying a Java Active Networking PlatformPractical Considerations for Deploying a Java Active Networking Platform
Practical Considerations for Deploying a Java Active Networking Platform
 
DOE Magellan OpenStack user story
DOE Magellan OpenStack user storyDOE Magellan OpenStack user story
DOE Magellan OpenStack user story
 
Janet Network R&D Innovation - HEAnet / Juniper Innovation Day
Janet Network R&D Innovation - HEAnet / Juniper Innovation DayJanet Network R&D Innovation - HEAnet / Juniper Innovation Day
Janet Network R&D Innovation - HEAnet / Juniper Innovation Day
 
Provisioning Janet
Provisioning JanetProvisioning Janet
Provisioning Janet
 
Synergy 2014 - Syn122 Moving Australian National Research into the Cloud
Synergy 2014 - Syn122 Moving Australian National Research into the CloudSynergy 2014 - Syn122 Moving Australian National Research into the Cloud
Synergy 2014 - Syn122 Moving Australian National Research into the Cloud
 
Archiving data from Durham to RAL using the File Transfer Service (FTS)
Archiving data from Durham to RAL using the File Transfer Service (FTS)Archiving data from Durham to RAL using the File Transfer Service (FTS)
Archiving data from Durham to RAL using the File Transfer Service (FTS)
 
Network security
Network securityNetwork security
Network security
 
Map reducecloudtech
Map reducecloudtechMap reducecloudtech
Map reducecloudtech
 
Welcome ndm11
Welcome ndm11Welcome ndm11
Welcome ndm11
 
Network-aware Data Management for High Throughput Flows Akamai, Cambridge, ...
Network-aware Data Management for High Throughput Flows   Akamai, Cambridge, ...Network-aware Data Management for High Throughput Flows   Akamai, Cambridge, ...
Network-aware Data Management for High Throughput Flows Akamai, Cambridge, ...
 
A location based least-cost scheduling for data-intensive applications
A location based least-cost scheduling for data-intensive applicationsA location based least-cost scheduling for data-intensive applications
A location based least-cost scheduling for data-intensive applications
 
Update on the Exascale Computing Project (ECP)
Update on the Exascale Computing Project (ECP)Update on the Exascale Computing Project (ECP)
Update on the Exascale Computing Project (ECP)
 
Data-intensive bioinformatics on HPC and Cloud
Data-intensive bioinformatics on HPC and CloudData-intensive bioinformatics on HPC and Cloud
Data-intensive bioinformatics on HPC and Cloud
 
Grid1
Grid1Grid1
Grid1
 

Viewers also liked

Science DMZ as a Service: Creating Science Super- Facilities with GENI
Science DMZ as a Service: Creating Science Super- Facilities with GENIScience DMZ as a Service: Creating Science Super- Facilities with GENI
Science DMZ as a Service: Creating Science Super- Facilities with GENIUS-Ignite
 
Internet2 Support for Biomedical Research
Internet2 Support for Biomedical ResearchInternet2 Support for Biomedical Research
Internet2 Support for Biomedical ResearchEd Dodds
 
Internet2 Support for Biomedical Research
Internet2 Support for Biomedical ResearchInternet2 Support for Biomedical Research
Internet2 Support for Biomedical ResearchEd Dodds
 
Experiences In Building Globus Genomics Using Galaxy, Globus Online and AWS
Experiences In Building Globus Genomics Using Galaxy, Globus Online and AWSExperiences In Building Globus Genomics Using Galaxy, Globus Online and AWS
Experiences In Building Globus Genomics Using Galaxy, Globus Online and AWSEd Dodds
 
2015 CDC Workshop on ScienceDMZ
2015 CDC Workshop on ScienceDMZ2015 CDC Workshop on ScienceDMZ
2015 CDC Workshop on ScienceDMZChris Dagdigian
 
Iris Ritter interconnection map
Iris Ritter interconnection mapIris Ritter interconnection map
Iris Ritter interconnection mapEd Dodds
 
Maximizing information and communications technologies for development in fai...
Maximizing information and communications technologies for development in fai...Maximizing information and communications technologies for development in fai...
Maximizing information and communications technologies for development in fai...Ed Dodds
 

Viewers also liked (7)

Science DMZ as a Service: Creating Science Super- Facilities with GENI
Science DMZ as a Service: Creating Science Super- Facilities with GENIScience DMZ as a Service: Creating Science Super- Facilities with GENI
Science DMZ as a Service: Creating Science Super- Facilities with GENI
 
Internet2 Support for Biomedical Research
Internet2 Support for Biomedical ResearchInternet2 Support for Biomedical Research
Internet2 Support for Biomedical Research
 
Internet2 Support for Biomedical Research
Internet2 Support for Biomedical ResearchInternet2 Support for Biomedical Research
Internet2 Support for Biomedical Research
 
Experiences In Building Globus Genomics Using Galaxy, Globus Online and AWS
Experiences In Building Globus Genomics Using Galaxy, Globus Online and AWSExperiences In Building Globus Genomics Using Galaxy, Globus Online and AWS
Experiences In Building Globus Genomics Using Galaxy, Globus Online and AWS
 
2015 CDC Workshop on ScienceDMZ
2015 CDC Workshop on ScienceDMZ2015 CDC Workshop on ScienceDMZ
2015 CDC Workshop on ScienceDMZ
 
Iris Ritter interconnection map
Iris Ritter interconnection mapIris Ritter interconnection map
Iris Ritter interconnection map
 
Maximizing information and communications technologies for development in fai...
Maximizing information and communications technologies for development in fai...Maximizing information and communications technologies for development in fai...
Maximizing information and communications technologies for development in fai...
 

Similar to Common Design Elements for Data Movement Eli Dart

The Science DMZ
The Science DMZThe Science DMZ
The Science DMZJisc
 
Tutorial: Maximizing Performance and Network Utility with a Science DMZ
Tutorial: Maximizing Performance and Network Utility with a Science DMZTutorial: Maximizing Performance and Network Utility with a Science DMZ
Tutorial: Maximizing Performance and Network Utility with a Science DMZGlobus
 
Data Mobility Exhibition
Data Mobility ExhibitionData Mobility Exhibition
Data Mobility ExhibitionGlobus
 
Don't Be Scared. Data Don't Bite. Introduction to Big Data.
Don't Be Scared. Data Don't Bite. Introduction to Big Data.Don't Be Scared. Data Don't Bite. Introduction to Big Data.
Don't Be Scared. Data Don't Bite. Introduction to Big Data.KGMGROUP
 
The Pacific Research Platform
The Pacific Research PlatformThe Pacific Research Platform
The Pacific Research PlatformLarry Smarr
 
ACES QuakeSim 2011
ACES QuakeSim 2011ACES QuakeSim 2011
ACES QuakeSim 2011marpierc
 
Big Data Analytics and Advanced Computer Networking Scenarios
Big Data Analytics and Advanced Computer Networking ScenariosBig Data Analytics and Advanced Computer Networking Scenarios
Big Data Analytics and Advanced Computer Networking ScenariosStenio Fernandes
 
Data processing in Cyber-Physical Systems
Data processing in Cyber-Physical SystemsData processing in Cyber-Physical Systems
Data processing in Cyber-Physical SystemsBob Marcus
 
A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...
A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...
A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...Ilkay Altintas, Ph.D.
 
_Cloud_Computing_Overview.pdf
_Cloud_Computing_Overview.pdf_Cloud_Computing_Overview.pdf
_Cloud_Computing_Overview.pdfTyStrk
 
Week 1 Lecture_1-5 CC_watermark.pdf
Week 1 Lecture_1-5 CC_watermark.pdfWeek 1 Lecture_1-5 CC_watermark.pdf
Week 1 Lecture_1-5 CC_watermark.pdfJohn422973
 
Network Management and Flow Analysis in Today’s Dense IT Environments
Network Management and Flow Analysis in Today’s Dense IT EnvironmentsNetwork Management and Flow Analysis in Today’s Dense IT Environments
Network Management and Flow Analysis in Today’s Dense IT EnvironmentsSolarWinds
 
Lecture 3.31 3.32.pptx
Lecture 3.31  3.32.pptxLecture 3.31  3.32.pptx
Lecture 3.31 3.32.pptxRATISHKUMAR32
 
Week 1 lecture material cc
Week 1 lecture material ccWeek 1 lecture material cc
Week 1 lecture material ccAnkit Gupta
 
Concepts of Distributed Computing & Cloud Computing
Concepts of Distributed Computing & Cloud Computing Concepts of Distributed Computing & Cloud Computing
Concepts of Distributed Computing & Cloud Computing Hitesh Kumar Markam
 
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...Tal Lavian Ph.D.
 
Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...
Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...
Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...Ed Dodds
 
Supporting Research through "Desktop as a Service" models of e-infrastructure...
Supporting Research through "Desktop as a Service" models of e-infrastructure...Supporting Research through "Desktop as a Service" models of e-infrastructure...
Supporting Research through "Desktop as a Service" models of e-infrastructure...David Wallom
 

Similar to Common Design Elements for Data Movement Eli Dart (20)

The Science DMZ
The Science DMZThe Science DMZ
The Science DMZ
 
Tutorial: Maximizing Performance and Network Utility with a Science DMZ
Tutorial: Maximizing Performance and Network Utility with a Science DMZTutorial: Maximizing Performance and Network Utility with a Science DMZ
Tutorial: Maximizing Performance and Network Utility with a Science DMZ
 
Data Mobility Exhibition
Data Mobility ExhibitionData Mobility Exhibition
Data Mobility Exhibition
 
Don't Be Scared. Data Don't Bite. Introduction to Big Data.
Don't Be Scared. Data Don't Bite. Introduction to Big Data.Don't Be Scared. Data Don't Bite. Introduction to Big Data.
Don't Be Scared. Data Don't Bite. Introduction to Big Data.
 
The Pacific Research Platform
The Pacific Research PlatformThe Pacific Research Platform
The Pacific Research Platform
 
try
trytry
try
 
ACES QuakeSim 2011
ACES QuakeSim 2011ACES QuakeSim 2011
ACES QuakeSim 2011
 
Big Data Analytics and Advanced Computer Networking Scenarios
Big Data Analytics and Advanced Computer Networking ScenariosBig Data Analytics and Advanced Computer Networking Scenarios
Big Data Analytics and Advanced Computer Networking Scenarios
 
Data processing in Cyber-Physical Systems
Data processing in Cyber-Physical SystemsData processing in Cyber-Physical Systems
Data processing in Cyber-Physical Systems
 
A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...
A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...
A Maturing Role of Workflows in the Presence of Heterogenous Computing Archit...
 
_Cloud_Computing_Overview.pdf
_Cloud_Computing_Overview.pdf_Cloud_Computing_Overview.pdf
_Cloud_Computing_Overview.pdf
 
Week 1 Lecture_1-5 CC_watermark.pdf
Week 1 Lecture_1-5 CC_watermark.pdfWeek 1 Lecture_1-5 CC_watermark.pdf
Week 1 Lecture_1-5 CC_watermark.pdf
 
Network Management and Flow Analysis in Today’s Dense IT Environments
Network Management and Flow Analysis in Today’s Dense IT EnvironmentsNetwork Management and Flow Analysis in Today’s Dense IT Environments
Network Management and Flow Analysis in Today’s Dense IT Environments
 
Lecture 3.31 3.32.pptx
Lecture 3.31  3.32.pptxLecture 3.31  3.32.pptx
Lecture 3.31 3.32.pptx
 
Week 1 lecture material cc
Week 1 lecture material ccWeek 1 lecture material cc
Week 1 lecture material cc
 
Concepts of Distributed Computing & Cloud Computing
Concepts of Distributed Computing & Cloud Computing Concepts of Distributed Computing & Cloud Computing
Concepts of Distributed Computing & Cloud Computing
 
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
Lambda Data Grid: An Agile Optical Platform for Grid Computing and Data-inten...
 
Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...
Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...
Creating a Climate for Innovation on Internet2 - Eric Boyd Senior Director, S...
 
Supporting Research through "Desktop as a Service" models of e-infrastructure...
Supporting Research through "Desktop as a Service" models of e-infrastructure...Supporting Research through "Desktop as a Service" models of e-infrastructure...
Supporting Research through "Desktop as a Service" models of e-infrastructure...
 
distributed system original.pdf
distributed system original.pdfdistributed system original.pdf
distributed system original.pdf
 

More from Ed Dodds

Updated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural America
Updated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural AmericaUpdated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural America
Updated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural AmericaEd Dodds
 
ILSR 2019 12 rural coop policy brief update page 8
ILSR 2019 12 rural coop policy brief update page 8ILSR 2019 12 rural coop policy brief update page 8
ILSR 2019 12 rural coop policy brief update page 8Ed Dodds
 
Inoversity - Bob Metcalfe
Inoversity - Bob MetcalfeInoversity - Bob Metcalfe
Inoversity - Bob MetcalfeEd Dodds
 
Distributed Ledger Technology
Distributed Ledger TechnologyDistributed Ledger Technology
Distributed Ledger TechnologyEd Dodds
 
UCX: An Open Source Framework for HPC Network APIs and Beyond
UCX: An Open Source Framework for HPC Network APIs and BeyondUCX: An Open Source Framework for HPC Network APIs and Beyond
UCX: An Open Source Framework for HPC Network APIs and BeyondEd Dodds
 
Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...
Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...
Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...Ed Dodds
 
Innovation Accelerators Report
Innovation Accelerators ReportInnovation Accelerators Report
Innovation Accelerators ReportEd Dodds
 
Strategy for American Innovation
Strategy for American InnovationStrategy for American Innovation
Strategy for American InnovationEd Dodds
 
Collaboration with NSFCloud
Collaboration with NSFCloudCollaboration with NSFCloud
Collaboration with NSFCloudEd Dodds
 
AppImpact: A Framework for Mobile Technology in Behavioral Healthcare
AppImpact: A Framework for Mobile Technology in Behavioral HealthcareAppImpact: A Framework for Mobile Technology in Behavioral Healthcare
AppImpact: A Framework for Mobile Technology in Behavioral HealthcareEd Dodds
 
Report to the President and Congress Ensuring Leadership in Federally Funded ...
Report to the President and Congress Ensuring Leadership in Federally Funded ...Report to the President and Congress Ensuring Leadership in Federally Funded ...
Report to the President and Congress Ensuring Leadership in Federally Funded ...Ed Dodds
 
Data Act Federal Register Notice Public Summary of Responses
Data Act Federal Register Notice Public Summary of ResponsesData Act Federal Register Notice Public Summary of Responses
Data Act Federal Register Notice Public Summary of ResponsesEd Dodds
 
Gloriad.flo con.2014.01
Gloriad.flo con.2014.01Gloriad.flo con.2014.01
Gloriad.flo con.2014.01Ed Dodds
 
2014 COMPENDIUM Edition of National Research and Education Networks in Europe
2014 COMPENDIUM Edition of National Research and  Education Networks in Europe2014 COMPENDIUM Edition of National Research and  Education Networks in Europe
2014 COMPENDIUM Edition of National Research and Education Networks in EuropeEd Dodds
 
New Westminster Keynote - Norman Jacknis
New Westminster Keynote - Norman JacknisNew Westminster Keynote - Norman Jacknis
New Westminster Keynote - Norman JacknisEd Dodds
 
HIMSS Innovation Pathways Summary
HIMSS Innovation Pathways SummaryHIMSS Innovation Pathways Summary
HIMSS Innovation Pathways SummaryEd Dodds
 
Supporting high throughput high-biotechnologies in today’s research environme...
Supporting high throughput high-biotechnologies in today’s research environme...Supporting high throughput high-biotechnologies in today’s research environme...
Supporting high throughput high-biotechnologies in today’s research environme...Ed Dodds
 
Nationwide Interoperability Roadmap draft version 1.0
Nationwide Interoperability Roadmap draft version 1.0Nationwide Interoperability Roadmap draft version 1.0
Nationwide Interoperability Roadmap draft version 1.0Ed Dodds
 

More from Ed Dodds (20)

Updated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural America
Updated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural AmericaUpdated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural America
Updated Policy Brief: Cooperatives Bring Fiber Internet Access to Rural America
 
ILSR 2019 12 rural coop policy brief update page 8
ILSR 2019 12 rural coop policy brief update page 8ILSR 2019 12 rural coop policy brief update page 8
ILSR 2019 12 rural coop policy brief update page 8
 
Inoversity - Bob Metcalfe
Inoversity - Bob MetcalfeInoversity - Bob Metcalfe
Inoversity - Bob Metcalfe
 
Distributed Ledger Technology
Distributed Ledger TechnologyDistributed Ledger Technology
Distributed Ledger Technology
 
UCX: An Open Source Framework for HPC Network APIs and Beyond
UCX: An Open Source Framework for HPC Network APIs and BeyondUCX: An Open Source Framework for HPC Network APIs and Beyond
UCX: An Open Source Framework for HPC Network APIs and Beyond
 
Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...
Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...
Digital Inclusion and Meaningful Broadband Adoption Initiatives Colin Rhinesm...
 
Jetstream
JetstreamJetstream
Jetstream
 
Innovation Accelerators Report
Innovation Accelerators ReportInnovation Accelerators Report
Innovation Accelerators Report
 
Work.
Work.Work.
Work.
 
Strategy for American Innovation
Strategy for American InnovationStrategy for American Innovation
Strategy for American Innovation
 
Collaboration with NSFCloud
Collaboration with NSFCloudCollaboration with NSFCloud
Collaboration with NSFCloud
 
AppImpact: A Framework for Mobile Technology in Behavioral Healthcare
AppImpact: A Framework for Mobile Technology in Behavioral HealthcareAppImpact: A Framework for Mobile Technology in Behavioral Healthcare
AppImpact: A Framework for Mobile Technology in Behavioral Healthcare
 
Report to the President and Congress Ensuring Leadership in Federally Funded ...
Report to the President and Congress Ensuring Leadership in Federally Funded ...Report to the President and Congress Ensuring Leadership in Federally Funded ...
Report to the President and Congress Ensuring Leadership in Federally Funded ...
 
Data Act Federal Register Notice Public Summary of Responses
Data Act Federal Register Notice Public Summary of ResponsesData Act Federal Register Notice Public Summary of Responses
Data Act Federal Register Notice Public Summary of Responses
 
Gloriad.flo con.2014.01
Gloriad.flo con.2014.01Gloriad.flo con.2014.01
Gloriad.flo con.2014.01
 
2014 COMPENDIUM Edition of National Research and Education Networks in Europe
2014 COMPENDIUM Edition of National Research and  Education Networks in Europe2014 COMPENDIUM Edition of National Research and  Education Networks in Europe
2014 COMPENDIUM Edition of National Research and Education Networks in Europe
 
New Westminster Keynote - Norman Jacknis
New Westminster Keynote - Norman JacknisNew Westminster Keynote - Norman Jacknis
New Westminster Keynote - Norman Jacknis
 
HIMSS Innovation Pathways Summary
HIMSS Innovation Pathways SummaryHIMSS Innovation Pathways Summary
HIMSS Innovation Pathways Summary
 
Supporting high throughput high-biotechnologies in today’s research environme...
Supporting high throughput high-biotechnologies in today’s research environme...Supporting high throughput high-biotechnologies in today’s research environme...
Supporting high throughput high-biotechnologies in today’s research environme...
 
Nationwide Interoperability Roadmap draft version 1.0
Nationwide Interoperability Roadmap draft version 1.0Nationwide Interoperability Roadmap draft version 1.0
Nationwide Interoperability Roadmap draft version 1.0
 

Recently uploaded

Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCRCall Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCRlizamodels9
 
(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)
(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)
(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)riyaescorts54
 
Bioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptxBioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptx023NiWayanAnggiSriWa
 
REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...
REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...
REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...Universidade Federal de Sergipe - UFS
 
Microphone- characteristics,carbon microphone, dynamic microphone.pptx
Microphone- characteristics,carbon microphone, dynamic microphone.pptxMicrophone- characteristics,carbon microphone, dynamic microphone.pptx
Microphone- characteristics,carbon microphone, dynamic microphone.pptxpriyankatabhane
 
User Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather StationUser Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather StationColumbia Weather Systems
 
Pests of safflower_Binomics_Identification_Dr.UPR.pdf
Pests of safflower_Binomics_Identification_Dr.UPR.pdfPests of safflower_Binomics_Identification_Dr.UPR.pdf
Pests of safflower_Binomics_Identification_Dr.UPR.pdfPirithiRaju
 
FREE NURSING BUNDLE FOR NURSES.PDF by na
FREE NURSING BUNDLE FOR NURSES.PDF by naFREE NURSING BUNDLE FOR NURSES.PDF by na
FREE NURSING BUNDLE FOR NURSES.PDF by naJASISJULIANOELYNV
 
Pests of Blackgram, greengram, cowpea_Dr.UPR.pdf
Pests of Blackgram, greengram, cowpea_Dr.UPR.pdfPests of Blackgram, greengram, cowpea_Dr.UPR.pdf
Pests of Blackgram, greengram, cowpea_Dr.UPR.pdfPirithiRaju
 
BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.
BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.
BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.PraveenaKalaiselvan1
 
User Guide: Magellan MX™ Weather Station
User Guide: Magellan MX™ Weather StationUser Guide: Magellan MX™ Weather Station
User Guide: Magellan MX™ Weather StationColumbia Weather Systems
 
User Guide: Pulsar™ Weather Station (Columbia Weather Systems)
User Guide: Pulsar™ Weather Station (Columbia Weather Systems)User Guide: Pulsar™ Weather Station (Columbia Weather Systems)
User Guide: Pulsar™ Weather Station (Columbia Weather Systems)Columbia Weather Systems
 
Behavioral Disorder: Schizophrenia & it's Case Study.pdf
Behavioral Disorder: Schizophrenia & it's Case Study.pdfBehavioral Disorder: Schizophrenia & it's Case Study.pdf
Behavioral Disorder: Schizophrenia & it's Case Study.pdfSELF-EXPLANATORY
 
《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》rnrncn29
 
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptxLIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptxmalonesandreagweneth
 
Davis plaque method.pptx recombinant DNA technology
Davis plaque method.pptx recombinant DNA technologyDavis plaque method.pptx recombinant DNA technology
Davis plaque method.pptx recombinant DNA technologycaarthichand2003
 
Microteaching on terms used in filtration .Pharmaceutical Engineering
Microteaching on terms used in filtration .Pharmaceutical EngineeringMicroteaching on terms used in filtration .Pharmaceutical Engineering
Microteaching on terms used in filtration .Pharmaceutical EngineeringPrajakta Shinde
 
The dark energy paradox leads to a new structure of spacetime.pptx
The dark energy paradox leads to a new structure of spacetime.pptxThe dark energy paradox leads to a new structure of spacetime.pptx
The dark energy paradox leads to a new structure of spacetime.pptxEran Akiva Sinbar
 
Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝soniya singh
 

Recently uploaded (20)

Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCRCall Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
 
(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)
(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)
(9818099198) Call Girls In Noida Sector 14 (NOIDA ESCORTS)
 
Bioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptxBioteknologi kelas 10 kumer smapsa .pptx
Bioteknologi kelas 10 kumer smapsa .pptx
 
REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...
REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...
REVISTA DE BIOLOGIA E CIÊNCIAS DA TERRA ISSN 1519-5228 - Artigo_Bioterra_V24_...
 
Microphone- characteristics,carbon microphone, dynamic microphone.pptx
Microphone- characteristics,carbon microphone, dynamic microphone.pptxMicrophone- characteristics,carbon microphone, dynamic microphone.pptx
Microphone- characteristics,carbon microphone, dynamic microphone.pptx
 
User Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather StationUser Guide: Capricorn FLX™ Weather Station
User Guide: Capricorn FLX™ Weather Station
 
Pests of safflower_Binomics_Identification_Dr.UPR.pdf
Pests of safflower_Binomics_Identification_Dr.UPR.pdfPests of safflower_Binomics_Identification_Dr.UPR.pdf
Pests of safflower_Binomics_Identification_Dr.UPR.pdf
 
FREE NURSING BUNDLE FOR NURSES.PDF by na
FREE NURSING BUNDLE FOR NURSES.PDF by naFREE NURSING BUNDLE FOR NURSES.PDF by na
FREE NURSING BUNDLE FOR NURSES.PDF by na
 
Pests of Blackgram, greengram, cowpea_Dr.UPR.pdf
Pests of Blackgram, greengram, cowpea_Dr.UPR.pdfPests of Blackgram, greengram, cowpea_Dr.UPR.pdf
Pests of Blackgram, greengram, cowpea_Dr.UPR.pdf
 
BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.
BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.
BIOETHICS IN RECOMBINANT DNA TECHNOLOGY.
 
Hot Sexy call girls in Moti Nagar,🔝 9953056974 🔝 escort Service
Hot Sexy call girls in  Moti Nagar,🔝 9953056974 🔝 escort ServiceHot Sexy call girls in  Moti Nagar,🔝 9953056974 🔝 escort Service
Hot Sexy call girls in Moti Nagar,🔝 9953056974 🔝 escort Service
 
User Guide: Magellan MX™ Weather Station
User Guide: Magellan MX™ Weather StationUser Guide: Magellan MX™ Weather Station
User Guide: Magellan MX™ Weather Station
 
User Guide: Pulsar™ Weather Station (Columbia Weather Systems)
User Guide: Pulsar™ Weather Station (Columbia Weather Systems)User Guide: Pulsar™ Weather Station (Columbia Weather Systems)
User Guide: Pulsar™ Weather Station (Columbia Weather Systems)
 
Behavioral Disorder: Schizophrenia & it's Case Study.pdf
Behavioral Disorder: Schizophrenia & it's Case Study.pdfBehavioral Disorder: Schizophrenia & it's Case Study.pdf
Behavioral Disorder: Schizophrenia & it's Case Study.pdf
 
《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》《Queensland毕业文凭-昆士兰大学毕业证成绩单》
《Queensland毕业文凭-昆士兰大学毕业证成绩单》
 
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptxLIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
LIGHT-PHENOMENA-BY-CABUALDIONALDOPANOGANCADIENTE-CONDEZA (1).pptx
 
Davis plaque method.pptx recombinant DNA technology
Davis plaque method.pptx recombinant DNA technologyDavis plaque method.pptx recombinant DNA technology
Davis plaque method.pptx recombinant DNA technology
 
Microteaching on terms used in filtration .Pharmaceutical Engineering
Microteaching on terms used in filtration .Pharmaceutical EngineeringMicroteaching on terms used in filtration .Pharmaceutical Engineering
Microteaching on terms used in filtration .Pharmaceutical Engineering
 
The dark energy paradox leads to a new structure of spacetime.pptx
The dark energy paradox leads to a new structure of spacetime.pptxThe dark energy paradox leads to a new structure of spacetime.pptx
The dark energy paradox leads to a new structure of spacetime.pptx
 
Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝
Call Girls in Munirka Delhi 💯Call Us 🔝8264348440🔝
 

Common Design Elements for Data Movement Eli Dart

  • 1. Common  Design  Elements  for   Data  Movement   Eli  Dart,  Network  Engineer   ESnet  Science  Engagement   Lawrence  Berkeley  Na<onal  Laboratory   Cosmology  CrossConnects  Workshop   Berkeley,  CA   February  11,  2015  
  • 2. Overview   2/10/15  2   •  Context   •  Design  paJerns   •  What  do  we  need  to  do?  
  • 3. Context   •  Data-­‐intensive  science  con<nues  to  need  high-­‐performance  data  movement   between  geographically  distant  loca<ons   –  Observa<on  (or  instrument)  to  analysis   –  Distribu<on  of  data  products  to  users   –  Aggrega<on  of  data  sets  for  analysis   –  Replica<on  to  archival  storage   •  Move  computa<on  to  data?    Of  course!    Except  when  you  can’t…   –  A  liquid  market  in  fungible  compu<ng  alloca<ons  does  not  exist   –  Users  get  an  alloca<on  of  <me  on  a  specific  compute  resource  –  if  the  data   isn’t  there  already,  it  needs  to  be  put  there   –  If  data  can’t  be  stored  long-­‐term  where  it’s  generated,  it  must  be  moved   –  Other  reasons  too  –  the  point  is  we  have  to  be  able  to  move  Big  Data   •  Given  the  need  for  data  movement,  how  can  we  reliably  do  it  well?   2/10/15  3  
  • 4. The  Task  of  Large  Scale  Data  Movement   •  Several  different  ways  to  look  at  a  data  movement  task   •  People  perspec<ve:   –  I  am  a  member  of  a  collabora<on   –  Our  collabora<on  has  accounts  with  compute  alloca<ons  and  data   storage  alloca<ons  at  a  set  of  sites   –  I  need  to  move  data  between  those  sites   •  Organiza<on/facility  perspec<ve:   –  ANL,  NCSA,  NERSC,  ORNL  and  SDSC  are  all  used  by  the  collabora<on   –  All  these  sites  must  have  data  transfer  tools  in  common   –  I  must  learn  what  tools  and  capabili<es  each  site  has,  and  apply  those   tools  to  my  task   •  Note  that  the  integra<on  burden  is  on  the  scien<st!   2/10/15  4  
  • 5. Service  Primi<ves   •  There  is  another  way  to  look  at  data  movement   •  All  large-­‐scale  data  movement  tasks  are  composed  of  a  set  of  primi<ves   –  Those  primi<ves  are  common  to  most  such  workflows   –  If  major  sites  can  agree  on  a  set  of  primi<ves,  all  large-­‐scale  data  workflows   will  benefit   •  What  are  the  common  primi<ves?   –  Storage  systems  (filesystems,  tape  archives,  etc.)   –  Data  transfer  applica<ons  (Globus,  others)   –  Workflow  tools,  if  automa<on  is  used   –  Networks   •  Local  networks   •  Wide  area  networks   •  What  if  these  worked  well  together  in  the  general  case?   •  Compose  them  into  common  design  paJerns   2/10/15  5  
  • 6. The  Central  Role  of  the  Network   •  The  very  structure  of  modern  science  assumes  science  networks  exist:  high   performance,  feature  rich,  global  scope   •  What  is  “The  Network”  anyway?   –  “The  Network”  is  the  set  of  devices  and  applica<ons  involved  in  the  use  of  a   remote  resource   •  This  is  not  about  supercomputer  interconnects   •  This  is  about  data  flow  from  experiment  to  analysis,  between  facili<es,  etc.   –  User  interfaces  for  “The  Network”  –  portal,  data  transfer  tool,  workflow  engine   –  Therefore,  servers  and  applica<ons  must  also  be  considered   •  What  is  important?    Ordered  list:   1.  Correctness   2.  Consistency   3.  Performance   ©  2014,  Energy  Sciences  Network   6 – ESnet Science Engagement (engage@es.net) - 2/10/15
  • 7. TCP  –  Ubiquitous  and  Fragile   •  Networks  provide  connec<vity  between  hosts  –  how  do  hosts  see  the   network?   –  From  an  applica<on’s  perspec<ve,  the  interface  to  “the  other  end”  is  a   socket   –  Communica<on  is  between  applica<ons  –  mostly  over  TCP   •  TCP  –  the  fragile  workhorse   –  TCP  is  (for  very  good  reasons)  <mid  –  packet  loss  is  interpreted  as   conges<on   –  Packet  loss  in  conjunc<on  with  latency  is  a  performance  killer   –  Like  it  or  not,  TCP  is  used  for  the  vast  majority  of  data  transfer   applica<ons  (more  than  95%  of  ESnet  traffic  is  TCP)   ©  2014,  Energy  Sciences  Network   7 – ESnet Science Engagement (engage@es.net) - 2/10/15
  • 8. A small amount of packet loss makes a huge difference in TCP performance Metro  Area   Local   (LAN)   Regional   Con<nental   Interna<onal   Measured (TCP Reno) Measured (HTCP) Theoretical (TCP Reno) Measured (no loss) With loss, high performance beyond metro distances is essentially impossible ©  2014,  Energy  Sciences  Network   8 – ESnet Science Engagement (engage@es.net) - 2/10/15
  • 9. Design  PaGern  –  The  Science  DMZ  Model   •  Design  paJerns  are  reusable  solu<ons  to  design  problems  that  recur  in  the  real   world   –  High  performance  data  movement  is  a  good  fit  for  this   –  Science  DMZ  model   •  Science  DMZ  incorporates  several  things   –  Network  enclave  at  or  near  site  perimeter   –  Sane  security  controls   •  Good  fit  for  high-­‐performance  applica<ons   •  Specific  to  Science  DMZ  services   –  Performance  test  and  measurement   –  Dedicated  systems  for  data  transfer  (Data  Transfer  Nodes)   •  High  performance  hosts   •  Good  tools   •  Details  at  hJp://fasterdata.es.net/science-­‐dmz/     2/10/15  9  
  • 10. Context:  Science  DMZ  Adop<on   •  DOE  Na<onal  Laboratories   –  Both  large  and  small  sites   –  HPC  centers,  LHC  sites,  experimental  facili<es   •  NSF  CC-­‐NIE  and  CC*IIE  programs  leverage  Science  DMZ   –  $40M  and  coun<ng  (third  round  awards  coming  soon,  es<mate  addi<onal  $18M  to  $20M)   –  Significant  investments  across  the  US  university  complex,  ~130  awards   –  Big  shoutout  to  Kevin  Thompson  and  the  NSF  –  these  programs  are  cri<cally  important   •  Na<onal  Ins<tutes  of  Health   –  100G  network  infrastructure  refresh   •  US  Department  of  Agriculture   –  Agricultural  Research  Service  is  building  a  new  science  network  based  on  the  Science  DMZ  model   –  hJps://www.ro.gov/index?s=opportunity&mode=form&tab=core&id=a7f291f4216b5a24c1177a5684e1809b   •  Other  US  agencies  looking  at  Science  DMZ  model   –  NASA   –  NOAA   •  Australian  Research  Data  Storage  Infrastructure  (RDSI)   –  Science  DMZs  at  major  sites,  connected  by  a  high  speed  network   –  hJps://www.rdsi.edu.au/dashnet   –  hJps://www.rdsi.edu.au/dashnet-­‐deployment-­‐rdsi-­‐nodes-­‐begins   •  Other  countries   –  Brazil   –  New  Zealand   –  More     2/10/15  10  
  • 11. Context:  Community  Capabili<es   •  Many  Science  DMZs  directly  support  science  applica<ons   –  LHC  (Run  2  is  coming  soon)   –  Experiment  opera<on  (Fusion,  Light  Sources,  etc.)   –  Data  transfer  into/out  of  HPC  facili<es   •  Many  Science  DMZs  are  SDN-­‐ready   –  Openflow-­‐capable  gear   –  SDN  research  ongoing   •  High-­‐performance  components   –  High-­‐speed  WAN  connec<vity   –  perfSONAR  deployments   –  DTN  deployments   •  Metcalfe’s  Law  of  Network  U<lity   –  Value  propor<onal  to  the  square  of  the  number  of  DMZs?  n  log(n)?   –  Cyberinfrastructure  value  increases  as  we  all  upgrade   2/10/15  11  
  • 12. Strategic  Impacts   •  What  does  this  mean?   –  We  are  in  the  midst  of  a  significant  cyberinfrastructure  upgrade   –  Enterprise  networks  need  not  be  unduly  perturbed  J   •  Significantly  enhanced  capabili<es  compared  to  3  years  ago   –  Terabyte-­‐scale  data  movement  is  much  easier   –  Petabyte-­‐scale  data  movement  possible  outside  the  LHC  experiments   •  3.1Gbps  =  1PB/month   •  (Try  doing  that  through  your  enterprise  firewall!)   –  Widely-­‐deployed  tools  are  much  beJer  (e.g.  Globus)   •  Raised  expecta<ons  for  network  infrastructures   –  Scien<sts  should  be  able  to  do  beJer  than  residen<al  broadband     •  Many  more  sites  can  now  achieve  good  performance   •  Incumbent  on  science  networks  to  meet  the  challenge   –  Remember  the  TCP  loss  characteris<cs   –  Use  perfSONAR   –  Science  experiments  assume  this  stuff  works  –  we  can  now  meet  their  needs     2/10/15  12  
  • 13. High  Performance  Data  Transfer  -­‐  Requirements   •  There  is  a  set  of  things  required  for  reliable  high-­‐performance  data  transfer   –  Long-­‐haul  networks     •  Well-­‐provisioned   •  High-­‐performance   –  Local  networks   •  Well-­‐provisioned   •  High-­‐performance   •  Sane  security   –  Local  data  systems   •  Dedicated  to  data  transfer  (else  too  much  complexity)   •  High-­‐performance  access  to  storage   –  Good  data  transfer  tools   •  Interoperable   •  High-­‐performance   –  Ease  of  use   •  Usable  by  people   •  Usable  by  workflows   •  Interoperable  across  sites  (remove  integra<on  burden)   2/10/15  13  
  • 14. Long-­‐Haul  Network  Status   •  100  Gigabit  per  second  networks  deployed  globally   –  USA/DOE  Na<onal  Laboratories  –  ESnet   –  USA/.edu  –  Internet2   –  Europe  –  GEANT   –  Many  state  and  regional  networks  have  or  are  deploying  100Gbps  cores   •  What  does  this  mean  in  terms  of  capability?   –  1TB/hour  requires  less  than  2.5Gbps  (2.5%  of  100Gbps  network)   –  1PB/week  requires  less  than  15Gbps  (15%  of  100Gbps  network)   –  hJp://fasterdata.es.net/home/requirements-­‐and-­‐expecta<ons   –  The  long-­‐haul  capacity  problem  is  now  solved,  to  first  order   •  Some  networks  are  s<ll  in  the  middle  of  upgrades   •  However,  steady  progress  is  being  made   2/10/15  14  
  • 15. Local  Network  Status   •  Many  ESnet  sites  now  have  100G  connec<ons  to  ESnet   –  2x100G:  BNL,  CERN,  FNAL   –  1x100G:  ANL,  LANL,  LBNL,  NERSC,  ORNL,  SLAC   •  Capacity  provisioning  is  much  easier  in  a  LAN  environment   •  Security  requires  aJen<on  (see  Science  DMZ)   •  Major  DOE  compu<ng  facili<es  have  a  lot  of  capacity  deployed  to  their  data   systems   –  ANL:  60Gbps   –  NERSC:  80Gbps   –  ORNL:  20Gbps   •  Big  win  if  sites  use  Science  DMZ  model   2/10/15  15  
  • 16. Progress  So  Far   •  There  is  a  set  of  things  required  for  reliable  high-­‐performance  data  transfer   –  Long-­‐haul  networks     •  Well-­‐provisioned   •  High-­‐performance   –  Local  networks   •  Well-­‐provisioned   •  High-­‐performance   •  Sane  security   –  Local  data  systems   •  Dedicated  to  data  transfer  (else  too  much  complexity)   •  High-­‐performance  access  to  storage   –  Good  data  transfer  tools   •  Interoperable   •  High-­‐performance   –  Ease  of  use   •  Usable  by  people   •  Usable  by  workflows   •  Interoperable  across  sites  (remove  integra<on  burden)   2/10/15  16  
  • 17. Local  Data  Systems   •  Science  DMZ  model  calls  these  Data  Transfer  Nodes   –  Dedicated  to  high-­‐performance  data  transfer  tasks   –  Short,  clean  path  to  outside  world   •  At  HPC  facili<es,  they  mount  the  global  filesystem   –  Transfer  data  to  the  DTN   –  Data  available  on  HPC  resource   •  High-­‐performance  data  transfer  tools   –  Globus  Transfer   –  Command-­‐line  globus-­‐url-­‐copy   –  BBCP   •  These  are  deployed  now  at  many  HPC  facili<es   –  ANL,  NERSC,  ORNL   –  NCSA,  SDSC   2/10/15  17  
  • 18. Data  Transfer  Tools   •  Interoperability  is  really  important   –  Remember,  scien<sts  should  not  have  to  do  the  integra<on   –  HPC  facili<es  should  agree  on  a  common  toolset   –  Today,  that  common  toolset  has  a  few  members   •  Globus  Transfer   •  SSH/SCP/Rsync  (yes,  I  know  –  ick!)   •  Many  niche  tools   •  Globus  appears  to  be  the  most  full-­‐featured   –  GUI,  data  integrity  checks,  fault  recovery   –  Fire  and  forget   –  API  for  workflows   •  Globus  is  also  widely  deployed   –  ANL,  NERSC,  ORNL   –  NCSA,  SDSC  (all  of  XSEDE)   –  Many  other  loca<ons   2/10/15  18  
  • 19. More  Progress   •  There  is  a  set  of  things  required  for  reliable  high-­‐performance  data  transfer   –  Long-­‐haul  networks     •  Well-­‐provisioned   •  High-­‐performance   –  Local  networks   •  Well-­‐provisioned   •  High-­‐performance   •  Sane  security   –  Local  data  systems   •  Dedicated  to  data  transfer  (else  too  much  complexity)   •  High-­‐performance  access  to  storage   –  Good  data  transfer  tools   •  Interoperable   •  High-­‐performance   –  Ease  of  use   •  Usable  by  people   •  Usable  by  workflows   •  Interoperable  across  sites  (remove  integra<on  burden)   2/10/15  19  
  • 20. Mission  Scope  and  Science  Support   •  Resource  providers  each  have  their  own  mission   –  ESnet:  high-­‐performance  networking  for  science   –  ANL,  NERSC,  ORNL:  HPC  for  DOE  science  users   –  NCSA,  SDSC,  et.  al.:  HPC  for  NSF  users   –  Globus:  full-­‐featured,  high-­‐performance  data  transfer  tools   •  No  responsibility  for  individual  science  projects   –  Resource  provider  staff  usually  not  part  of  science  projects   –  Science  projects  have  to  do  their  own  integra<on  (see  beginning  of  talk)   •  However,  resource  providers  are  typically  responsive  to  user  requests   –  If  you  have  a  problem,  it’s  their  job  to  fix  it   –  I  propose  we  use  this  to  get  something  done   2/10/15  20  
  • 21. Hypothe<cal:  HPC  Data  Transfer  Capability   •  This  community  has  significant  data  transfer  needs   –  I  have  worked  with  some  of  you  in  the  past   –  Simula<ons,  sky  surveys,  etc.   –  Expecta<on  over  <me  that  needs  will  increase   •  Improve  data  movement  capability   –  ANL,  NERSC,  ORNL   –  NCSA,  SDSC   –  This  is  an  arbitrary  list,  based  on  my  incomplete  understanding   –  Should  there  be  others?   •  Goal:  per-­‐Globus-­‐job  performance  of  1PB/week   –  I  don’t  mean  we  have  to  transfer  1PB  every  week   –  But,  if  we  need  to,  we  should  be  able  to   –  Remember,  this  only  takes  15%  of  a  100G  network  path   2/10/15  21  
  • 22. What  Would  Be  Required?   •  We  would  need  several  things:   –  Specific  workflow  (move  dataset  D  of  size  S  from  A  to  Z,  frequency  F)   –  A  commitment  by  resource  providers  to  see  it  through   •  ESnet  (+  other  networks  if  needed)   •  Compu<ng  facili<es   •  Globus   •  Is  it  100%  plug-­‐and-­‐play?    No.   –  There  are  almost  certainly  some  wrinkles   –  However,  most  of  the  hard  part  is  done   •  Networks   •  Data  transfer  nodes   •  Tools   •  Let’s  work  together  and  make  this  happen!   2/10/15  22  
  • 23. Ques<ons  For  You   •  Would  an  effort  like  this  be  useful?  (I  think  so)   •  Does  this  community  need  this  capability?    (I  think  so)   •  Are  there  obvious  gaps?  (probably,  e.g.  performance  to  tape)   •  Which  sites  would  be  involved?   •  Am  I  crazy?  (I  think  not)   2/10/15  23  
  • 24. Thanks!   Eli  Dart   Energy  Sciences  Network  (ESnet)   Lawrence  Berkeley  Na<onal  Laboratory   hJp://fasterdata.es.net/   hJp://my.es.net/   hJp://www.es.net/  
  • 26. Support  For  Science  Traffic   •  The  Science  DMZ  is  typically  deployed  to  support  science  traffic   –  Typically  large  data  transfers  over  long  distances   –  In  most  cases,  the  data  transfer  applica<ons  use  TCP   •  The  behavior  of  TCP  is  a  legacy  from  the  conges<on  collapse  of  the  Internet  in   the  1980s   –  Loss  is  interpreted  as  conges<on   –  TCP  backs  off  to  avoid  conges<on  à  performance  degrades   –  Performance  hit  related  to  the  square  of  the  packet  loss  rate   •  Addressing  this  problem  is  a  dominant  engineering  considera<on  for  science   networks   –  Lots  of  design  effort   –  Lots  of  engineering  <me   –  Lots  of  troubleshoo<ng  effort   2/10/15  26  
  • 27. A small amount of packet loss makes a huge difference in TCP performance 2/10/15   Metro  Area   Local   (LAN)   Regional   Con<nental   Interna<onal   Measured (TCP Reno) Measured (HTCP) Theoretical (TCP Reno) Measured (no loss) With loss, high performance beyond metro distances is essentially impossible