Why	
  Tradi*onal	
  Databases	
  Fail	
  so	
  
Miserably	
  to	
  Scale	
  
With	
  E-­‐Commerce	
  Growth	
  	
  
Dragon	
  Slayer	
  Consul*ng	
  –	
  Marc	
  Staimer	
  CDS	
  
Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   2	
  
marcstaimer@me.com
Analyst/Consultant	
  
IT	
  Industry	
  Experience	
  
Publish	
  w/TT	
  &	
  GigaOM	
  
Speak	
  @	
  Trade	
  Shows	
  
Improve	
  Vendors	
  Marke9ng	
  
Help	
  EUs	
  w/Problems	
  
Agenda	
  
Storage	
  Decisions	
  Conference	
  	
  	
  	
  |	
  	
  	
  	
  ©	
  TechTarget	
  	
  
It’s	
  What	
  We	
  Know	
  Is	
  True	
  When	
  It’s	
  Not	
  That	
  Causes	
  
The	
  Most	
  Difficult	
  Problems	
  
Storage	
  Decisions	
  Conference	
  	
  	
  	
  |	
  	
  	
  	
  ©	
  TechTarget	
  	
  
If	
  I	
  Catch	
  You,	
  
You’re	
  Mine	
  
SQL	
  Database	
  E-­‐Commerce	
  Performance	
  Problem	
  
!   E-­‐Commerce	
  mission	
  cri9cal	
  
!   Slow	
  performance	
  and/or	
  outages	
  =	
  
!   Lost	
  sales	
  
!   Lost	
  revenue	
  	
  
!   Lost	
  profits	
  
8/5/14	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   5	
  
When	
  The	
  IOPS	
  BoOleneck	
  is	
  Storage	
  
!   >	
  DRAM	
  	
  
!   Fast	
  with	
  lots	
  of	
  random	
  IOPS	
  
!   Expensive	
  &	
  vola9le	
  
!   >	
  HDDs	
  
!   Rela9vely	
  inexpensive	
  
!   Not	
  as	
  many	
  random	
  IOPS	
  so	
  need	
  lots	
  of	
  them	
  
!   >	
  suppor9ng	
  infrastructure	
  
!   >	
  exper9se	
  required	
  (short-­‐stroking)	
  	
  
8/5/14	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   6	
  
The	
  New	
  Darling:	
  Flash	
  Storage	
  
!   DDR	
  DIMM	
  
!   PCIe	
  
!   SAS/SATA	
  SSD	
  
!   Hybrid	
  array	
  
!   All	
  flash	
  array	
  (AFA)	
  
8/5/14	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   7	
  
Flash	
  ShiRs	
  SQL	
  DBMS	
  Performance	
  BoOleneck	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   8	
  
Making	
  SQL	
  DBMS	
  Performance	
  A	
  “Scale”	
  Problem	
  
!   Complicated	
  
!   Requiring	
  a	
  lot	
  of	
  exper9se	
  
!   Costs	
  are	
  excessive	
  
!   Many	
  tasks	
  are	
  con9nuously	
  manually	
  labor-­‐intensive	
  
!   Reduced	
  Up9me	
  
Common	
  SQL	
  Database	
  Scaling	
  Workarounds	
  
!   Scale-­‐Up	
  
!   Scale-­‐Out	
  
!   Sharding	
  
!   Read-­‐only	
  Slaves	
  
!   Peer-­‐to-­‐Peer	
  Replica9on	
  
!   Linked	
  Servers	
  &	
  Distributed	
  Queries	
  
!   Distributed	
  Par99on	
  Views	
  
!   Data	
  Par99on	
  Rou9ng	
  
!   NoSQL	
  Databases	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   10	
  
Scale-­‐Up	
  
!   More	
  cores	
  
!   More	
  DRAM	
  
!   More	
  system	
  IO	
  bandwidth	
  
!   More	
  power	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   11	
  
Scale-­‐Up	
  Pros	
  &	
  Cons	
  
Pros	
  
!   Simpler	
  –	
  easier	
  	
  
!   No	
  sodware	
  changes	
  
!   Nothing	
  addi9onal	
  to	
  manage	
  
!   Or	
  train/learn	
  
Cons	
  
!   Cost	
  $$$$	
  
!   >	
  Server	
  HW	
  CapEx	
  
!   >	
  Server	
  HW	
  OpEx	
  
!   (Servers	
  oden	
  proprietary)	
  
!   >	
  DBMS	
  licensing	
  costs	
  
!   >	
  HA	
  costs	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   12	
  
Scale-­‐Out:	
  Sharding	
  
!   Breaks	
  up	
  the	
  database	
  	
  
!   Into	
  separate	
  instances	
  
!   Easy	
  to	
  understand	
  
!   Fairly	
  common	
  especially	
  with	
  
!   MySQL	
  
!   SQL	
  Server	
  
!   PostgreSQL	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   13	
  
Sharding	
  Pros	
  &	
  Cons	
  
Pros	
  
!   Data	
  split	
  along	
  natural	
  fault	
  lines	
  
!   Highly	
  scalable	
  
!   Many	
  DBAs	
  know	
  how	
  
!   Non-­‐DBAs	
  do	
  not	
  
Cons	
  
!   Mul9ple	
  DBMS	
  instances	
  >	
  license	
  $	
  
!   Requires	
  lots	
  of	
  exper9se	
  
!  App	
  &	
  SQL	
  DBMS	
  data	
  structures	
  
!   Time	
  consuming:	
  labor	
  intensive	
  
!  Error	
  prone	
  
!   Non-­‐dynamic	
  
!   Rela9onship	
  loss	
  
!   Too	
  many	
  SPOFs	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   14	
  
Scale-­‐Out:	
  Read-­‐Only	
  Slaves	
  
!   Master	
  database	
  that	
  replicates	
  
!   To	
  a	
  series	
  of	
  slave	
  databases	
  
!   Slaves	
  are	
  read-­‐only	
  
!   #	
  of	
  slaves	
  varies	
  
!   Limited	
  only	
  by	
  master’s	
  replica9on	
  ability	
  
!  Speed	
  
!  Affect	
  on	
  DBMS	
  performance	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   15	
  
Slaved	
  
SQL	
  DBMS	
  
Read-­‐Only	
  
Database	
  Files	
  
Read-­‐Only	
  Slaves	
  Pros	
  &	
  Cons	
  
Pros	
  
!   Nominal	
  setup	
  
!   Simple	
  
!   App/DBMS	
  	
  transparent	
  
!   Incredibly	
  Popular	
  
Cons	
  
!   Master	
  database	
  bolleneck	
  
!   Not	
  good	
  for	
  write	
  intensive	
  apps	
  
!   Master	
  database	
  SPOF	
  
!   Patches/upgrades/fixes	
  
!   Mul9ple	
  DBMS	
  instances	
  =	
  >	
  $$$$	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   16	
  
Scale-­‐Out:	
  Peer-­‐to-­‐Peer	
  w/Replica*on	
  
!   Mul9ple	
  database	
  copies	
  	
  
!   Each	
  database	
  engine	
  manages	
  and	
  maintains	
  its	
  own	
  DBMS	
  copy	
  	
  
!   Replica9on	
  updates	
  other	
  database	
  engines	
  when	
  there’s	
  a	
  change	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   17	
  
Peer-­‐to-­‐Peer	
  w/Replica*on	
  	
  
Pros	
  &	
  Cons	
  
Pros	
  
!   Mul9ple	
  database	
  copies	
  
!   No	
  SPOF	
  
!   Good	
  for	
  infrequent	
  updates	
  
Cons	
  
!   Complicated	
  opera9ons	
  
!   Conflict	
  resolu9on	
  is	
  difficult	
  
!   Stewardship:	
  labor-­‐intensive	
  
!   Merge	
  replica9on:	
  high	
  overhead	
  
!   Mul9ple	
  DBMS	
  instances	
  =	
  >	
  $$$$	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   18	
  
Scale-­‐Out:	
  Linked	
  Servers	
  &	
  Distributed	
  Queries	
  
!   U9lized	
  when:	
  
!   Databases	
  are	
  split	
  into	
  func9onal	
  areas	
  
!  Requiring	
  minimal	
  coupling	
  
!   Or	
  when	
  spliong	
  data	
  by	
  type	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   19	
  
Client	
  
App	
  	
  
Server	
  
DBMS	
  
Server	
  1	
  
DBMS	
  
Server	
  2	
  
Linked	
  Servers	
  &	
  Distributed	
  Queries	
  	
  
Pros	
  &	
  Cons	
  
Pros	
  
!   Makes	
  mul9-­‐DBMS	
  appear	
  as	
  1	
  
!   DBMS	
  update	
  frequency	
  	
  
!   Has	
  no	
  impact	
  
!   DBMS	
  split	
  into	
  func9onal	
  areas	
  
!   Minimal	
  coupling	
  	
  
!   When	
  spliong	
  data	
  by	
  type	
  
Cons	
  
!   Referen9al	
  integrity	
  constraints	
  
!   Queries	
  spanning	
  mul9ple	
  DBMS	
  	
  
!   =	
  >	
  latencies	
  <	
  performance	
  
!   Significant	
  exper9se	
  required	
  
!   Difficult	
  to	
  pull	
  off	
  successfully	
  
!   Mul9ple	
  DBMS	
  instances	
  =	
  >	
  $$$$	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   20	
  
Scale-­‐Out:	
  Distributed	
  Par**on	
  Views	
  (DPV)	
  
!   Table	
  data	
  par99oned	
  among	
  tables	
  in	
  numerous	
  distributed	
  databases	
  	
  
!  Based	
  on	
  a	
  par99oning	
  key	
  
!   Update-­‐intensive	
  apps	
  best	
  target	
  for	
  DPVs	
  
!  Generally	
  good	
  E-­‐Commerce	
  query	
  performance	
  
!  Few	
  rows	
  impacted	
  –	
  likely	
  processed	
  in	
  single	
  DBMS	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   21	
  
Distributed	
  Par**on	
  Views	
  (DPV)	
  	
  
Pros	
  &	
  Cons	
  
Pros	
  
!   Good	
  for	
  update	
  intensive	
  Apps	
  	
  
!   App	
  transparent	
  
!   Scale-­‐out	
  par99oned	
  data	
  
!   Table	
  data	
  par99oned	
  into	
  tables	
  
!   Mul9ple	
  DBMS	
  
Cons	
  
!   Data	
  movement	
  between	
  par99ons	
  
!   Slows	
  performance	
  
!   Ltd	
  by	
  table’s	
  par9onability	
  
!   Mul9ple	
  SPOFs	
  
!   Significant	
  exper9se	
  required	
  
!   Difficult	
  to	
  pull	
  off	
  successfully	
  
!   Mul9ple	
  DBMS	
  instances	
  =	
  >	
  $$$$	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   22	
  
Scale-­‐Out:	
  Data	
  Dependent	
  Rou*ng	
  (DDR)	
  
!   E-­‐Commerce	
  app	
  or	
  middleware	
  responsible	
  for	
  DBMS	
  query	
  rou9ng	
  
!   Premise:	
  they	
  have	
  deeper	
  understanding	
  of	
  the	
  data	
  
!   Popular	
  when	
  queries	
  distributed	
  across	
  hundreds	
  or	
  more	
  databases	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   23	
  
Client	
  
Middleware	
  or	
  	
  
data	
  layer	
  routed	
  
to	
  SQL	
  DBMS	
  	
  
w/data	
  requested	
  
Data	
  Dependent	
  Rou*ng	
  (DDR)	
  	
  
Pros	
  &	
  Cons	
  
Pros	
  
!   Good	
  when	
  queries	
  spread	
  
!   Across	
  100s	
  or	
  >	
  DBMS	
  
!   Apps	
  know	
  their	
  own	
  data	
  
!   Beler	
  query	
  decisions	
  
Cons	
  
!   Significant	
  exper9se	
  required	
  
!   Deep	
  understanding	
  of	
  apps	
  
!   Or	
  middleware	
  	
  
!   Strong	
  architectural	
  skills	
  
!   Massively	
  complicated	
  
!  Based	
  on	
  #	
  of	
  DBMS	
  servers	
  
!   ALL	
  apps	
  must	
  be	
  customized	
  
!   Mul9ple	
  DBMS	
  instances	
  =	
  >	
  $$$$	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   24	
  
Scale-­‐Out:	
  NoSQL	
  Databases	
  
!   Excellent	
  for	
  large	
  data	
  sets	
  and	
  analy9cs	
  in-­‐general	
  
!   OLAP	
  
!   Data	
  Warehousing	
  
!   Unstructured	
  data	
  analy9cs	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   25	
  
NoSQL	
  Databases	
  Pros	
  &	
  Cons	
  
Pros	
  
!   Data	
  Ingest	
  
!   Unstructured	
  data	
  analy9cs	
  
!   OLAP	
  
!   Data	
  Warehousing	
  
!   Historical	
  data	
  
Cons	
  
!   Poor	
  transac9onal	
  performance	
  
!   Lack	
  real-­‐9me	
  dashboards	
  
!   &	
  Real-­‐9me	
  repor9ng	
  
!   Not	
  good	
  @	
  scanning	
  data	
  
!   As	
  it’s	
  changing	
  “NOW”	
  
!   Lacks	
  common	
  SQL	
  DBMS	
  tools	
  
!   Technology	
  s9ll	
  immature	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   26	
  
Conclusion	
  &	
  Summary	
  
!   Tradi9onal	
  scaling	
  methodologies	
  have	
  serious	
  issues	
  ranging	
  from	
  
!   Too	
  much	
  cost	
  $$$$	
  
!   Too	
  complicated	
  
!  Up	
  front	
  &	
  ongoing	
  
!   Too	
  labor	
  intensive	
  
!   Too	
  much	
  down9me	
  
!  Trouble	
  shoo9ng	
  excessively	
  difficult	
  
!   Too	
  much	
  frustra9on	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   27	
  
The	
  Test	
  of	
  Three	
  
Fall	
  2014	
   Dragon	
  Slayer	
  Consul9ng	
  All	
  Rights	
  Reserved	
  ©	
  2014	
  	
   28	
  
© 2014 CLUSTRIX
Software
launched
San Francisco
San Jose
Seattle
New York
London
Company
Offices Worldwide Top-Tier Venture Funding
1 Trillion+ transactions
per month
100s of installations worldwide
Clustrix
founded
Mission-critical
production
deployments
20142006 2010 2013
ClustrixDB Overview30
Without the right database,
e-commerce success
leads to website slowdown or failure
Slowdowns
Outages
Unhappy
Customers
Lost
Revenue
ClustrixDB Overview31
More
Uptime
More
Capacity
More
Business
The first e-commerce database
purpose-built for today’s online retailers
ClustrixDB Overview32
ClustrixDB Overview
Today’s E-Commerce Database
•  Readily handles e-commerce growth
•  Easy accommodates traffic spikes
•  Meets no-downtime demands
•  Supports on-the-fly updates
•  Enables live e-commerce reporting
•  Integrates with e-commerce stack
•  Replaces MySQL with minimal effort
ClustrixDB Overview33
Serve fast-growing mobile
and social channels
Business Advantages
Ready for
Revenue Growth
Peak Performance
at Peak Times
Capture More
Business
Scale effortlessly with
sales success
Avoid lost revenue
during traffic spikes
Readily manage
unpredictable demand
Analyze and act on
real-time sales data
Better orchestrate
omni-channel shopping
ClustrixDB Overview34
E-commerce leaders across retail, travel, digital services,
and social commerce rely on ClustrixDB
Customer Highlights
ClustrixDB Overview35
E-Retailer Takes 600% Sales Spike in Stride
“Three years ago, Black Friday was horrible. We wasted money
having database issues. With Clustrix, those issues are gone.”
Challenge ClustrixDB
•  Second-fastest growth rate in
Internet Retailer Top 500
•  Loads spike 20X or more
during holidays
•  Hours-long downtime incidents,
with $20K-$60K in lost revenue
per hour
•  Flexed up and down to
handle 600% Cyber Monday
sales spike with no downtime
•  Real-time catalog updates
during flash sales
•  Live reporting and analytics
on e-commerce business
Keith Bussey
VP Technology
ClustrixDB Overview36
Online Textbook Company Graduates from MySQL
“We wouldn’t be able to grow our business as easily
without ClustrixDB.”
Challenge ClustrixDB
•  3X revenue growth strained
chaotic MySQL backend
•  Significant administrative
overhead required for
redundancy and uptime
•  Large schema changes needed
system-wide downtime
•  100% increase in database
load with no performance hits
•  Simplified operations with
easy expansion and
seamless datacenter failover
•  Went from 10 hours of
downtime to no downtime
with on-the-fly changes
Paolo Resmini
VP of Engineering
ClustrixDB Overview37
Picture-Perfect Performance for Digital Photo Service
“Clustrix has given us one advantage over our competitors…
the ability to stay open at peak times with great service levels.”
Challenge ClustrixDB
•  Highly seasonal business
with 10X spike in December
•  Need to ensure uptime
during peak loads for
30 million members
•  Reached capacity limits of
conventional database
•  Gained performance and
scale with no app rewrite
•  89% reduction in December
service incidents
•  More time on innovation,
less time on database worries
Graham Hobson
CTO
ClustrixDB Overview38
Online Travel Site Increases Uptime, Lowers TCO
“MakeMyTrip frequently experiences 50% spikes in traffic that
no longer cause slow page load times and costly downtime.”
Challenge ClustrixDB
•  Two server MySQL config
with single point of failure
•  Slow page loads and costly
downtime
•  Sharding / MySQL Cluster
too expensive to implement
and maintain
•  Drop-in replacement with
no single point failure
•  $100K cost reduction with
easy expansion
•  Handles massive spikes in
traffic with faster web
response times
Sanjay Kharb
Director Technical Ops
ClustrixDB Overview39
Automatic,
100% fault tolerance
Flex up and down,
in minutes
Massive, linear scalability
Technical Advantages
Capacity Availability Productivity
Extreme concurrency
No single point of failure
Battle-tested performance
Eliminates re-architecting
the database
Plug-in MySQL compatibility
Self-managing operation
ClustrixDB Overview40
Clustrix Design!
Intelligent Data
Distribution
Massively Parallel
Query Processing
Shared Nothing
Architecture
SQL	
  	
  
SQL	
  	
  
SQL	
  	
  
SQL	
  	
  
SQL	
  	
  
ClustrixDB Overview41
Query Compiler
Data Map
Database Engine
Query Compiler
Data Map
Database Engine
Query Compiler
Data Map
Database Engine
Signs It’s Time for ClustrixDB
•  Capacity planning for a busy shopping season
•  Gearing up for mobile and social demand
•  Issues with website slowdowns or outages
•  Scaling is too costly, complex and time consuming
•  Database is a single point of failure
•  Business wants new e-commerce capabilities
ClustrixDB Overview42
Thank You.
facebook.com/clustrix
ecommerce.clustrix.com
@clustrix
linkedin.com/clustrix
ClustrixDB Overview43

Why Traditional Databases Fail so Miserably to Scale with E-Commerce Site Growth

  • 1.
    Why  Tradi*onal  Databases  Fail  so   Miserably  to  Scale   With  E-­‐Commerce  Growth    
  • 2.
    Dragon  Slayer  Consul*ng  –  Marc  Staimer  CDS   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     2   marcstaimer@me.com Analyst/Consultant   IT  Industry  Experience   Publish  w/TT  &  GigaOM   Speak  @  Trade  Shows   Improve  Vendors  Marke9ng   Help  EUs  w/Problems  
  • 3.
    Agenda   Storage  Decisions  Conference        |        ©  TechTarget    
  • 4.
    It’s  What  We  Know  Is  True  When  It’s  Not  That  Causes   The  Most  Difficult  Problems   Storage  Decisions  Conference        |        ©  TechTarget     If  I  Catch  You,   You’re  Mine  
  • 5.
    SQL  Database  E-­‐Commerce  Performance  Problem   !   E-­‐Commerce  mission  cri9cal   !   Slow  performance  and/or  outages  =   !   Lost  sales   !   Lost  revenue     !   Lost  profits   8/5/14   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     5  
  • 6.
    When  The  IOPS  BoOleneck  is  Storage   !   >  DRAM     !   Fast  with  lots  of  random  IOPS   !   Expensive  &  vola9le   !   >  HDDs   !   Rela9vely  inexpensive   !   Not  as  many  random  IOPS  so  need  lots  of  them   !   >  suppor9ng  infrastructure   !   >  exper9se  required  (short-­‐stroking)     8/5/14   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     6  
  • 7.
    The  New  Darling:  Flash  Storage   !   DDR  DIMM   !   PCIe   !   SAS/SATA  SSD   !   Hybrid  array   !   All  flash  array  (AFA)   8/5/14   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     7  
  • 8.
    Flash  ShiRs  SQL  DBMS  Performance  BoOleneck   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     8  
  • 9.
    Making  SQL  DBMS  Performance  A  “Scale”  Problem   !   Complicated   !   Requiring  a  lot  of  exper9se   !   Costs  are  excessive   !   Many  tasks  are  con9nuously  manually  labor-­‐intensive   !   Reduced  Up9me  
  • 10.
    Common  SQL  Database  Scaling  Workarounds   !   Scale-­‐Up   !   Scale-­‐Out   !   Sharding   !   Read-­‐only  Slaves   !   Peer-­‐to-­‐Peer  Replica9on   !   Linked  Servers  &  Distributed  Queries   !   Distributed  Par99on  Views   !   Data  Par99on  Rou9ng   !   NoSQL  Databases   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     10  
  • 11.
    Scale-­‐Up   !  More  cores   !   More  DRAM   !   More  system  IO  bandwidth   !   More  power   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     11  
  • 12.
    Scale-­‐Up  Pros  &  Cons   Pros   !   Simpler  –  easier     !   No  sodware  changes   !   Nothing  addi9onal  to  manage   !   Or  train/learn   Cons   !   Cost  $$$$   !   >  Server  HW  CapEx   !   >  Server  HW  OpEx   !   (Servers  oden  proprietary)   !   >  DBMS  licensing  costs   !   >  HA  costs   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     12  
  • 13.
    Scale-­‐Out:  Sharding   !  Breaks  up  the  database     !   Into  separate  instances   !   Easy  to  understand   !   Fairly  common  especially  with   !   MySQL   !   SQL  Server   !   PostgreSQL   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     13  
  • 14.
    Sharding  Pros  &  Cons   Pros   !   Data  split  along  natural  fault  lines   !   Highly  scalable   !   Many  DBAs  know  how   !   Non-­‐DBAs  do  not   Cons   !   Mul9ple  DBMS  instances  >  license  $   !   Requires  lots  of  exper9se   !  App  &  SQL  DBMS  data  structures   !   Time  consuming:  labor  intensive   !  Error  prone   !   Non-­‐dynamic   !   Rela9onship  loss   !   Too  many  SPOFs   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     14  
  • 15.
    Scale-­‐Out:  Read-­‐Only  Slaves   !   Master  database  that  replicates   !   To  a  series  of  slave  databases   !   Slaves  are  read-­‐only   !   #  of  slaves  varies   !   Limited  only  by  master’s  replica9on  ability   !  Speed   !  Affect  on  DBMS  performance   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     15   Slaved   SQL  DBMS   Read-­‐Only   Database  Files  
  • 16.
    Read-­‐Only  Slaves  Pros  &  Cons   Pros   !   Nominal  setup   !   Simple   !   App/DBMS    transparent   !   Incredibly  Popular   Cons   !   Master  database  bolleneck   !   Not  good  for  write  intensive  apps   !   Master  database  SPOF   !   Patches/upgrades/fixes   !   Mul9ple  DBMS  instances  =  >  $$$$   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     16  
  • 17.
    Scale-­‐Out:  Peer-­‐to-­‐Peer  w/Replica*on   !   Mul9ple  database  copies     !   Each  database  engine  manages  and  maintains  its  own  DBMS  copy     !   Replica9on  updates  other  database  engines  when  there’s  a  change   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     17  
  • 18.
    Peer-­‐to-­‐Peer  w/Replica*on     Pros  &  Cons   Pros   !   Mul9ple  database  copies   !   No  SPOF   !   Good  for  infrequent  updates   Cons   !   Complicated  opera9ons   !   Conflict  resolu9on  is  difficult   !   Stewardship:  labor-­‐intensive   !   Merge  replica9on:  high  overhead   !   Mul9ple  DBMS  instances  =  >  $$$$   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     18  
  • 19.
    Scale-­‐Out:  Linked  Servers  &  Distributed  Queries   !   U9lized  when:   !   Databases  are  split  into  func9onal  areas   !  Requiring  minimal  coupling   !   Or  when  spliong  data  by  type   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     19   Client   App     Server   DBMS   Server  1   DBMS   Server  2  
  • 20.
    Linked  Servers  &  Distributed  Queries     Pros  &  Cons   Pros   !   Makes  mul9-­‐DBMS  appear  as  1   !   DBMS  update  frequency     !   Has  no  impact   !   DBMS  split  into  func9onal  areas   !   Minimal  coupling     !   When  spliong  data  by  type   Cons   !   Referen9al  integrity  constraints   !   Queries  spanning  mul9ple  DBMS     !   =  >  latencies  <  performance   !   Significant  exper9se  required   !   Difficult  to  pull  off  successfully   !   Mul9ple  DBMS  instances  =  >  $$$$   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     20  
  • 21.
    Scale-­‐Out:  Distributed  Par**on  Views  (DPV)   !   Table  data  par99oned  among  tables  in  numerous  distributed  databases     !  Based  on  a  par99oning  key   !   Update-­‐intensive  apps  best  target  for  DPVs   !  Generally  good  E-­‐Commerce  query  performance   !  Few  rows  impacted  –  likely  processed  in  single  DBMS   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     21  
  • 22.
    Distributed  Par**on  Views  (DPV)     Pros  &  Cons   Pros   !   Good  for  update  intensive  Apps     !   App  transparent   !   Scale-­‐out  par99oned  data   !   Table  data  par99oned  into  tables   !   Mul9ple  DBMS   Cons   !   Data  movement  between  par99ons   !   Slows  performance   !   Ltd  by  table’s  par9onability   !   Mul9ple  SPOFs   !   Significant  exper9se  required   !   Difficult  to  pull  off  successfully   !   Mul9ple  DBMS  instances  =  >  $$$$   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     22  
  • 23.
    Scale-­‐Out:  Data  Dependent  Rou*ng  (DDR)   !   E-­‐Commerce  app  or  middleware  responsible  for  DBMS  query  rou9ng   !   Premise:  they  have  deeper  understanding  of  the  data   !   Popular  when  queries  distributed  across  hundreds  or  more  databases   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     23   Client   Middleware  or     data  layer  routed   to  SQL  DBMS     w/data  requested  
  • 24.
    Data  Dependent  Rou*ng  (DDR)     Pros  &  Cons   Pros   !   Good  when  queries  spread   !   Across  100s  or  >  DBMS   !   Apps  know  their  own  data   !   Beler  query  decisions   Cons   !   Significant  exper9se  required   !   Deep  understanding  of  apps   !   Or  middleware     !   Strong  architectural  skills   !   Massively  complicated   !  Based  on  #  of  DBMS  servers   !   ALL  apps  must  be  customized   !   Mul9ple  DBMS  instances  =  >  $$$$   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     24  
  • 25.
    Scale-­‐Out:  NoSQL  Databases   !   Excellent  for  large  data  sets  and  analy9cs  in-­‐general   !   OLAP   !   Data  Warehousing   !   Unstructured  data  analy9cs   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     25  
  • 26.
    NoSQL  Databases  Pros  &  Cons   Pros   !   Data  Ingest   !   Unstructured  data  analy9cs   !   OLAP   !   Data  Warehousing   !   Historical  data   Cons   !   Poor  transac9onal  performance   !   Lack  real-­‐9me  dashboards   !   &  Real-­‐9me  repor9ng   !   Not  good  @  scanning  data   !   As  it’s  changing  “NOW”   !   Lacks  common  SQL  DBMS  tools   !   Technology  s9ll  immature   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     26  
  • 27.
    Conclusion  &  Summary   !   Tradi9onal  scaling  methodologies  have  serious  issues  ranging  from   !   Too  much  cost  $$$$   !   Too  complicated   !  Up  front  &  ongoing   !   Too  labor  intensive   !   Too  much  down9me   !  Trouble  shoo9ng  excessively  difficult   !   Too  much  frustra9on   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     27  
  • 28.
    The  Test  of  Three   Fall  2014   Dragon  Slayer  Consul9ng  All  Rights  Reserved  ©  2014     28  
  • 29.
  • 30.
    Software launched San Francisco San Jose Seattle NewYork London Company Offices Worldwide Top-Tier Venture Funding 1 Trillion+ transactions per month 100s of installations worldwide Clustrix founded Mission-critical production deployments 20142006 2010 2013 ClustrixDB Overview30
  • 31.
    Without the rightdatabase, e-commerce success leads to website slowdown or failure Slowdowns Outages Unhappy Customers Lost Revenue ClustrixDB Overview31
  • 32.
    More Uptime More Capacity More Business The first e-commercedatabase purpose-built for today’s online retailers ClustrixDB Overview32
  • 33.
    ClustrixDB Overview Today’s E-CommerceDatabase •  Readily handles e-commerce growth •  Easy accommodates traffic spikes •  Meets no-downtime demands •  Supports on-the-fly updates •  Enables live e-commerce reporting •  Integrates with e-commerce stack •  Replaces MySQL with minimal effort ClustrixDB Overview33
  • 34.
    Serve fast-growing mobile andsocial channels Business Advantages Ready for Revenue Growth Peak Performance at Peak Times Capture More Business Scale effortlessly with sales success Avoid lost revenue during traffic spikes Readily manage unpredictable demand Analyze and act on real-time sales data Better orchestrate omni-channel shopping ClustrixDB Overview34
  • 35.
    E-commerce leaders acrossretail, travel, digital services, and social commerce rely on ClustrixDB Customer Highlights ClustrixDB Overview35
  • 36.
    E-Retailer Takes 600%Sales Spike in Stride “Three years ago, Black Friday was horrible. We wasted money having database issues. With Clustrix, those issues are gone.” Challenge ClustrixDB •  Second-fastest growth rate in Internet Retailer Top 500 •  Loads spike 20X or more during holidays •  Hours-long downtime incidents, with $20K-$60K in lost revenue per hour •  Flexed up and down to handle 600% Cyber Monday sales spike with no downtime •  Real-time catalog updates during flash sales •  Live reporting and analytics on e-commerce business Keith Bussey VP Technology ClustrixDB Overview36
  • 37.
    Online Textbook CompanyGraduates from MySQL “We wouldn’t be able to grow our business as easily without ClustrixDB.” Challenge ClustrixDB •  3X revenue growth strained chaotic MySQL backend •  Significant administrative overhead required for redundancy and uptime •  Large schema changes needed system-wide downtime •  100% increase in database load with no performance hits •  Simplified operations with easy expansion and seamless datacenter failover •  Went from 10 hours of downtime to no downtime with on-the-fly changes Paolo Resmini VP of Engineering ClustrixDB Overview37
  • 38.
    Picture-Perfect Performance forDigital Photo Service “Clustrix has given us one advantage over our competitors… the ability to stay open at peak times with great service levels.” Challenge ClustrixDB •  Highly seasonal business with 10X spike in December •  Need to ensure uptime during peak loads for 30 million members •  Reached capacity limits of conventional database •  Gained performance and scale with no app rewrite •  89% reduction in December service incidents •  More time on innovation, less time on database worries Graham Hobson CTO ClustrixDB Overview38
  • 39.
    Online Travel SiteIncreases Uptime, Lowers TCO “MakeMyTrip frequently experiences 50% spikes in traffic that no longer cause slow page load times and costly downtime.” Challenge ClustrixDB •  Two server MySQL config with single point of failure •  Slow page loads and costly downtime •  Sharding / MySQL Cluster too expensive to implement and maintain •  Drop-in replacement with no single point failure •  $100K cost reduction with easy expansion •  Handles massive spikes in traffic with faster web response times Sanjay Kharb Director Technical Ops ClustrixDB Overview39
  • 40.
    Automatic, 100% fault tolerance Flexup and down, in minutes Massive, linear scalability Technical Advantages Capacity Availability Productivity Extreme concurrency No single point of failure Battle-tested performance Eliminates re-architecting the database Plug-in MySQL compatibility Self-managing operation ClustrixDB Overview40
  • 41.
    Clustrix Design! Intelligent Data Distribution MassivelyParallel Query Processing Shared Nothing Architecture SQL     SQL     SQL     SQL     SQL     ClustrixDB Overview41 Query Compiler Data Map Database Engine Query Compiler Data Map Database Engine Query Compiler Data Map Database Engine
  • 42.
    Signs It’s Timefor ClustrixDB •  Capacity planning for a busy shopping season •  Gearing up for mobile and social demand •  Issues with website slowdowns or outages •  Scaling is too costly, complex and time consuming •  Database is a single point of failure •  Business wants new e-commerce capabilities ClustrixDB Overview42
  • 43.