SlideShare a Scribd company logo
1 of 16
Download to read offline
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
11
11
IO Management with IME1.1
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
12 What are the IO challenges?
Application
Developer
approach
IO Characterisation IO Challenge
Multi-physics Non-trivial mix of IO behaviours Filesystem must be optimised for all
IO types, not just a small subset
Workflows and
Ensembles
Coordination of files/objects across many jobs. Different
IO access patterns from different elements of the
workflow
Requires a global namespace with
strong consistency
Adaptive mesh
refinement
Very difficult to manage and maintain logical application
“block” sizes with underlying filesystem
Non-optimal access to files, incurs
RMW, fs lock contention
Metadata-
orientation
Small, random IOs, Shared file formats Limited by filesystem locking
Machine learning High read activity Tough for HDD: disk thrashing
Higher Concurrency
and heterogeneous
CPUs
Large thread counts make sequential IO look like
random IO
avoids all assistance from Filesystem
and Storage caches
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
13
DDN Confidential
How do I make my application go faster?
► Optimal/Natural behaviour already exist within the algorithms
► Difficult to optimise for multiple aspects simultaneously (GPU, multi-
threading, network changes, caches)
► Difficult to optimise our application for new environments
► Difficult to maintain efficient porting even between different node counts
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
14
DDN Confidential
How do I make my application go faster?
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
15 DDN | IME DataFlow
Fast Data
NVM & SSD
Persistent
Data (Disk)
Diverse, high
concurrency
applications
Compute
Application issues IO
to IME client.
Erasure Coding
applied
IME client sends
fragments to IME
servers
IME servers write buffers
to NVM and manage
internal metadata
IME servers write
aligned sequential
I/O to SFA backend
Parallel File system
operates at
maximum efficiency
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
16
DDN Confidential
Applying Flash-Cache Ecomonics
► Real World IO sub-system activity
• 99% of the time <30%
• 70% of the time <5%
► Build a cache for the peaks, PFS for the capacity
► 10x Performance/Watt improvement for flash-cache over PFS
Argonne: P. Carns, K. Harms et al., Understanding and Improving Computational Science Storage Access through Continuous
Characterization
Parallel File System
Scale-Out Flash Cache
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
17
problem > client cache
problem(s) with Life Science IO
ftp ftp-mouse.sanger.ac.uk: anonymous
331 Anonymous login ok, send your complete email address as your password
Password:
230 Anonymous access granted
ftp> cd REL-1210-BAM
ftp> ls –lh
-rw-rw-r-- 1 proftpd proftpd 138.5G Dec 7 2009 129P2_OlaHsd.bam
-rw-rw-r-- 1 proftpd proftpd 182.3G Oct 9 2012 129S1_SvImJ.bam
-rw-rw-r-- 1 proftpd proftpd 60.0G Dec 10 2009 129S5SvEvBrd.bam
-rw-rw-r-- 1 proftpd proftpd 107.0G Oct 9 2012 A_J.bam
-rw-rw-r-- 1 proftpd proftpd 112.8G Oct 9 2012 AKR_J.bam
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
18 IME1.1 | Protected, Fast, Flexible IO
IME
Server0
IME
Server1
IME
Server2
IME
Server3
streaming writes
streaming reads
random reads
random writes
shared file
high concurrency
Single Shared
Namespace to
all clients
Tackle Tough IO dial-in erasure coding
blisteringly fast IOPs
wirespeed throughput
adaptive IO
insane rebuild speeds
FILE CACHE
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
19 IME240 Performance (EDR)
► 100% of peak SSD performance to clients
► Consistently high performance across IO sizes and SSF/FPP
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
20
DDN Confidential
IME1.1 | Blistering Sequential Performance
File Per Process
1M Seq IO4k Seq IO
► IME240 Sequential performance
to file per process over 23GB/s
per server (POSIX IO)
► >21GB/s with 4k IOs (MPIIO)
► consistent read and write
performance across IO sizes
► POSIX IO performance hits max
values at ~32K IO sizes
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
21
DDN Confidential
IME1.1 | Blistering Sequential Performance
Single Shared File
1M Seq IO4k Seq IO
► IME240 Sequential
performance for Single
Shared File over 22GB/s per
server (POSIX IO)
► 15GB/s with 4k IOs (MPIIO)
► consistent read and write
performance across IO
sizes
► POSIX IO performance hits
max values at ~32K IO sizes
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
22
DDN Confidential
IME | WRF with I/O quilting
one I/O process per node
Measurement Speed Up
IO Wallclock 17.2x
Total Execution Wallclock 3.5x
► Application wall time
reduced by 3.5 using IME
► I/O time reduces by x17.2
► Move from I/O bound to
compute bound
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
23 IME1.1 | True Dial-In Erasure Coding
IME
Server0
IME
Server1
IME
Server2
IME
ServerN
FILE CACHE
8+38+1
6+0 8+2
6+1
4+1
4+14+1
8+0
1+1
► IME1.1 supports multiple resilience
levels through flexible, adaptive
erasure coding
► System Wide Default up to 15+3
► Applications can overide defaults
and select a specific Erasure
Coding Scheme
DEFAULT: 8+3
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
24 IME1.1 Node Failure Management
► IME1.1 allows continued IO Service through IME server failure(s)
► Following server failure, IME commences background rebuild of all
missing extents for each Parity Group Instance (PGI) : 3+1 across
the 3 remaining servers
• 1 server will be overloaded with 2 blocks for each PGI
• Following a server failure but prior to rebuild completion, immediate read requests may be
satisfied through on-the-fly reconstruction of missing file extents
► Service interruption will occur with a second IME server failure in
this case (but not with NVM device failures that occur after rebuild)
► New writes still maintain 3+1 (with overloading)
► IME restart is needed to re-introduce failed node and restore
original redundancy. (requires full sync to the BFS)
4 IME Servers, Default Parity Schema= 3+1
Continued Production
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
25 IME1.1 Node Failure Management
► With higher node counts (>4) and using an erasure coding geometry of
higher parity count (>1), IME 1.1 can sustain >1 node failures
► In the below case after one node fails, reads which rely upon missing
extents are reconstructed from parity. Writes can continue to be written with
6+2 with full redundancy since we still have 8 nodes
► If another server fails, service continues, with writes at 6+2, but one server
will contain 2 objects in each PGI
9 IME Servers, Default Parity Schema= 6+2
Continued Production
ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others.
Any statements or representations around future events are subject to change.
26
Apply Scale-Out FLASH-CACHE Economics with
Best Performance Density 500GB/s per rack
Adaptive IO route around IO bottlenecks
Insane Rebuild Speeds minutes not hours
Flexible Erasure Coding SW-defined, adaptive N+M
Flash-Native minimise write amplification. maximise SSD lifetime
IME240 | IME1.1SCALE-OUT | OPTIMISED DATAPATH | FLASH-CACHE | BURST-BUFFER
EXTREME IOPS | MASSIVE THROUGHPUT | SOFTWARE DEFINED

More Related Content

What's hot

All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...
All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...
All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...Tony Pearson
 
DB2 Accounting Reporting
DB2  Accounting ReportingDB2  Accounting Reporting
DB2 Accounting ReportingJohn Campbell
 
Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS by Namik Hrle ...
Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS  by  Namik Hrle ...Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS  by  Namik Hrle ...
Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS by Namik Hrle ...Surekha Parekh
 
Episode 3 DB2 pureScale Availability And Recovery [Read Only] [Compatibility...
Episode 3  DB2 pureScale Availability And Recovery [Read Only] [Compatibility...Episode 3  DB2 pureScale Availability And Recovery [Read Only] [Compatibility...
Episode 3 DB2 pureScale Availability And Recovery [Read Only] [Compatibility...Laura Hood
 
Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - favero
 Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - favero Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - favero
Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - faveroWillie Favero
 
IBM's new Flashsystem 900
IBM's new Flashsystem 900IBM's new Flashsystem 900
IBM's new Flashsystem 900Stefan Lein
 
DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...
DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...
DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...John Campbell
 
DB2 11 for z/OS Migration Planning and Early Customer Experiences
DB2 11 for z/OS Migration Planning and Early Customer ExperiencesDB2 11 for z/OS Migration Planning and Early Customer Experiences
DB2 11 for z/OS Migration Planning and Early Customer ExperiencesJohn Campbell
 
IBM flash systems
IBM flash systems IBM flash systems
IBM flash systems Solv AS
 
A First Look at the DB2 10 DSNZPARM Changes
A First Look at the DB2 10 DSNZPARM ChangesA First Look at the DB2 10 DSNZPARM Changes
A First Look at the DB2 10 DSNZPARM ChangesWillie Favero
 
DB2 Pure Scale Webcast
DB2 Pure Scale WebcastDB2 Pure Scale Webcast
DB2 Pure Scale WebcastLaura Hood
 
Ibm db2 10.5 for linux, unix, and windows installing ibm data server clients
Ibm db2 10.5 for linux, unix, and windows   installing ibm data server clientsIbm db2 10.5 for linux, unix, and windows   installing ibm data server clients
Ibm db2 10.5 for linux, unix, and windows installing ibm data server clientsbupbechanhgmail
 
DB2 for z/OS - Starter's guide to memory monitoring and control
DB2 for z/OS - Starter's guide to memory monitoring and controlDB2 for z/OS - Starter's guide to memory monitoring and control
DB2 for z/OS - Starter's guide to memory monitoring and controlFlorence Dubois
 
Partition Manager 11 - Paragon Software
Partition Manager 11 - Paragon SoftwarePartition Manager 11 - Paragon Software
Partition Manager 11 - Paragon SoftwareParagon Software Group
 
Partition Manager 11 - Paragon Software
Partition Manager 11 - Paragon SoftwarePartition Manager 11 - Paragon Software
Partition Manager 11 - Paragon Softwareguestcf43641
 
IBMSystem x3620 M3 IBMSystems and Technology Data Sheet
IBMSystem x3620 M3 IBMSystems and Technology Data SheetIBMSystem x3620 M3 IBMSystems and Technology Data Sheet
IBMSystem x3620 M3 IBMSystems and Technology Data SheetIBM India Smarter Computing
 

What's hot (19)

All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...
All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...
All Flash is not Equal: Tony Pearson contrasts IBM FlashSystem with Solid-Sta...
 
DB2 Accounting Reporting
DB2  Accounting ReportingDB2  Accounting Reporting
DB2 Accounting Reporting
 
Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS by Namik Hrle ...
Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS  by  Namik Hrle ...Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS  by  Namik Hrle ...
Efficient Monitoring & Tuning of Dynamic SQL in DB2 for z/OS by Namik Hrle ...
 
Episode 3 DB2 pureScale Availability And Recovery [Read Only] [Compatibility...
Episode 3  DB2 pureScale Availability And Recovery [Read Only] [Compatibility...Episode 3  DB2 pureScale Availability And Recovery [Read Only] [Compatibility...
Episode 3 DB2 pureScale Availability And Recovery [Read Only] [Compatibility...
 
Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - favero
 Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - favero Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - favero
Universal Table Spaces for DB2 10 for z/OS - IOD 2010 Seesion 1929 - favero
 
IBM's new Flashsystem 900
IBM's new Flashsystem 900IBM's new Flashsystem 900
IBM's new Flashsystem 900
 
DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...
DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...
DB2 for z/OS Bufferpool Tuning win by Divide and Conquer or Lose by Multiply ...
 
DB2 11 for z/OS Migration Planning and Early Customer Experiences
DB2 11 for z/OS Migration Planning and Early Customer ExperiencesDB2 11 for z/OS Migration Planning and Early Customer Experiences
DB2 11 for z/OS Migration Planning and Early Customer Experiences
 
IBM flash systems
IBM flash systems IBM flash systems
IBM flash systems
 
A First Look at the DB2 10 DSNZPARM Changes
A First Look at the DB2 10 DSNZPARM ChangesA First Look at the DB2 10 DSNZPARM Changes
A First Look at the DB2 10 DSNZPARM Changes
 
DB2 Pure Scale Webcast
DB2 Pure Scale WebcastDB2 Pure Scale Webcast
DB2 Pure Scale Webcast
 
Ibm db2 10.5 for linux, unix, and windows installing ibm data server clients
Ibm db2 10.5 for linux, unix, and windows   installing ibm data server clientsIbm db2 10.5 for linux, unix, and windows   installing ibm data server clients
Ibm db2 10.5 for linux, unix, and windows installing ibm data server clients
 
FS900 Product Guide
FS900 Product GuideFS900 Product Guide
FS900 Product Guide
 
DB2 for z/OS - Starter's guide to memory monitoring and control
DB2 for z/OS - Starter's guide to memory monitoring and controlDB2 for z/OS - Starter's guide to memory monitoring and control
DB2 for z/OS - Starter's guide to memory monitoring and control
 
FS900 Data Sheet.PDF
FS900 Data Sheet.PDFFS900 Data Sheet.PDF
FS900 Data Sheet.PDF
 
Partition Manager 11 - Paragon Software
Partition Manager 11 - Paragon SoftwarePartition Manager 11 - Paragon Software
Partition Manager 11 - Paragon Software
 
Partition Manager 11 - Paragon Software
Partition Manager 11 - Paragon SoftwarePartition Manager 11 - Paragon Software
Partition Manager 11 - Paragon Software
 
IBMSystem x3620 M3 IBMSystems and Technology Data Sheet
IBMSystem x3620 M3 IBMSystems and Technology Data SheetIBMSystem x3620 M3 IBMSystems and Technology Data Sheet
IBMSystem x3620 M3 IBMSystems and Technology Data Sheet
 
IBM XIV Gen3 Storage System
IBM XIV Gen3 Storage SystemIBM XIV Gen3 Storage System
IBM XIV Gen3 Storage System
 

Similar to IO Management with IME 1.1

IME - Unlocking the Potential of NVMe
IME - Unlocking the Potential of NVMeIME - Unlocking the Potential of NVMe
IME - Unlocking the Potential of NVMeinside-BigData.com
 
DDN: Protecting Your Data, Protecting Your Hardware
DDN: Protecting Your Data, Protecting Your HardwareDDN: Protecting Your Data, Protecting Your Hardware
DDN: Protecting Your Data, Protecting Your Hardwareinside-BigData.com
 
Infinite Memory Engine: HPC in the FLASH Era
Infinite Memory Engine: HPC in the FLASH EraInfinite Memory Engine: HPC in the FLASH Era
Infinite Memory Engine: HPC in the FLASH Erainside-BigData.com
 
Optimizing Lustre and GPFS with DDN
Optimizing Lustre and GPFS with DDNOptimizing Lustre and GPFS with DDN
Optimizing Lustre and GPFS with DDNinside-BigData.com
 
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...In-Memory Computing Summit
 
DDN and Intel: Partnered for Exascale
DDN and Intel: Partnered for ExascaleDDN and Intel: Partnered for Exascale
DDN and Intel: Partnered for ExascaleIntel IT Center
 
VMAX : répondez aux niveaux de services applicatifs les plus élevés
VMAX : répondez aux niveaux de services applicatifs les plus élevésVMAX : répondez aux niveaux de services applicatifs les plus élevés
VMAX : répondez aux niveaux de services applicatifs les plus élevésRSD
 
We4IT lcty 2013 - infra-man - domino run faster
We4IT lcty 2013 - infra-man - domino run faster We4IT lcty 2013 - infra-man - domino run faster
We4IT lcty 2013 - infra-man - domino run faster We4IT Group
 
A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...
A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...
A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...IBM India Smarter Computing
 
5 Things You Need to Know About Enterprise Fl
 5 Things You Need to Know About Enterprise Fl 5 Things You Need to Know About Enterprise Fl
5 Things You Need to Know About Enterprise FlWestern Digital
 
Presentation integration vmware with emc storage
Presentation   integration vmware with emc storagePresentation   integration vmware with emc storage
Presentation integration vmware with emc storagesolarisyourep
 
Z4R: Intro to Storage and DFSMS for z/OS
Z4R: Intro to Storage and DFSMS for z/OSZ4R: Intro to Storage and DFSMS for z/OS
Z4R: Intro to Storage and DFSMS for z/OSTony Pearson
 
Z109889 z4 r-storage-dfsms-jburg-v1909d
Z109889 z4 r-storage-dfsms-jburg-v1909dZ109889 z4 r-storage-dfsms-jburg-v1909d
Z109889 z4 r-storage-dfsms-jburg-v1909dTony Pearson
 
Z109889 z4 r-storage-dfsms-vegas-v1910b
Z109889 z4 r-storage-dfsms-vegas-v1910bZ109889 z4 r-storage-dfsms-vegas-v1910b
Z109889 z4 r-storage-dfsms-vegas-v1910bTony Pearson
 
DDN Strategic Vision Tour June 2015
DDN Strategic Vision Tour June 2015DDN Strategic Vision Tour June 2015
DDN Strategic Vision Tour June 2015inside-BigData.com
 
Hadoop Analytics on Isilon Deep Dive
Hadoop Analytics on Isilon Deep DiveHadoop Analytics on Isilon Deep Dive
Hadoop Analytics on Isilon Deep DiveClaudioFahey1
 
Controlling performance in the cloud: taking charge of your hosting environment
Controlling performance in the cloud: taking charge of your hosting environmentControlling performance in the cloud: taking charge of your hosting environment
Controlling performance in the cloud: taking charge of your hosting environmentDatabarracks
 
Wipro - FM Best Practices Showcase
Wipro - FM Best Practices ShowcaseWipro - FM Best Practices Showcase
Wipro - FM Best Practices ShowcaseSudhendu Bali
 

Similar to IO Management with IME 1.1 (20)

IME - Unlocking the Potential of NVMe
IME - Unlocking the Potential of NVMeIME - Unlocking the Potential of NVMe
IME - Unlocking the Potential of NVMe
 
DDN: Protecting Your Data, Protecting Your Hardware
DDN: Protecting Your Data, Protecting Your HardwareDDN: Protecting Your Data, Protecting Your Hardware
DDN: Protecting Your Data, Protecting Your Hardware
 
Infinite Memory Engine: HPC in the FLASH Era
Infinite Memory Engine: HPC in the FLASH EraInfinite Memory Engine: HPC in the FLASH Era
Infinite Memory Engine: HPC in the FLASH Era
 
DDN Product Update from SC13
DDN Product Update from SC13DDN Product Update from SC13
DDN Product Update from SC13
 
Optimizing Lustre and GPFS with DDN
Optimizing Lustre and GPFS with DDNOptimizing Lustre and GPFS with DDN
Optimizing Lustre and GPFS with DDN
 
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
IMC Summit 2016 Breakout - Brian Bulkowski - NVMe, Storage Class Memory and O...
 
DDN and Intel: Partnered for Exascale
DDN and Intel: Partnered for ExascaleDDN and Intel: Partnered for Exascale
DDN and Intel: Partnered for Exascale
 
VMAX : répondez aux niveaux de services applicatifs les plus élevés
VMAX : répondez aux niveaux de services applicatifs les plus élevésVMAX : répondez aux niveaux de services applicatifs les plus élevés
VMAX : répondez aux niveaux de services applicatifs les plus élevés
 
We4IT lcty 2013 - infra-man - domino run faster
We4IT lcty 2013 - infra-man - domino run faster We4IT lcty 2013 - infra-man - domino run faster
We4IT lcty 2013 - infra-man - domino run faster
 
A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...
A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...
A z/OS System Programmer’s Guide to Migrating to a New IBM System z9 EC or z9...
 
5 Things You Need to Know About Enterprise Fl
 5 Things You Need to Know About Enterprise Fl 5 Things You Need to Know About Enterprise Fl
5 Things You Need to Know About Enterprise Fl
 
Presentation integration vmware with emc storage
Presentation   integration vmware with emc storagePresentation   integration vmware with emc storage
Presentation integration vmware with emc storage
 
Z4R: Intro to Storage and DFSMS for z/OS
Z4R: Intro to Storage and DFSMS for z/OSZ4R: Intro to Storage and DFSMS for z/OS
Z4R: Intro to Storage and DFSMS for z/OS
 
Z109889 z4 r-storage-dfsms-jburg-v1909d
Z109889 z4 r-storage-dfsms-jburg-v1909dZ109889 z4 r-storage-dfsms-jburg-v1909d
Z109889 z4 r-storage-dfsms-jburg-v1909d
 
Z109889 z4 r-storage-dfsms-vegas-v1910b
Z109889 z4 r-storage-dfsms-vegas-v1910bZ109889 z4 r-storage-dfsms-vegas-v1910b
Z109889 z4 r-storage-dfsms-vegas-v1910b
 
DDN Strategic Vision Tour June 2015
DDN Strategic Vision Tour June 2015DDN Strategic Vision Tour June 2015
DDN Strategic Vision Tour June 2015
 
Mellanox Storage Solutions
Mellanox Storage SolutionsMellanox Storage Solutions
Mellanox Storage Solutions
 
Hadoop Analytics on Isilon Deep Dive
Hadoop Analytics on Isilon Deep DiveHadoop Analytics on Isilon Deep Dive
Hadoop Analytics on Isilon Deep Dive
 
Controlling performance in the cloud: taking charge of your hosting environment
Controlling performance in the cloud: taking charge of your hosting environmentControlling performance in the cloud: taking charge of your hosting environment
Controlling performance in the cloud: taking charge of your hosting environment
 
Wipro - FM Best Practices Showcase
Wipro - FM Best Practices ShowcaseWipro - FM Best Practices Showcase
Wipro - FM Best Practices Showcase
 

More from inside-BigData.com

Preparing to program Aurora at Exascale - Early experiences and future direct...
Preparing to program Aurora at Exascale - Early experiences and future direct...Preparing to program Aurora at Exascale - Early experiences and future direct...
Preparing to program Aurora at Exascale - Early experiences and future direct...inside-BigData.com
 
Transforming Private 5G Networks
Transforming Private 5G NetworksTransforming Private 5G Networks
Transforming Private 5G Networksinside-BigData.com
 
The Incorporation of Machine Learning into Scientific Simulations at Lawrence...
The Incorporation of Machine Learning into Scientific Simulations at Lawrence...The Incorporation of Machine Learning into Scientific Simulations at Lawrence...
The Incorporation of Machine Learning into Scientific Simulations at Lawrence...inside-BigData.com
 
How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...
How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...
How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...inside-BigData.com
 
Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...
Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...
Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...inside-BigData.com
 
HPC Impact: EDA Telemetry Neural Networks
HPC Impact: EDA Telemetry Neural NetworksHPC Impact: EDA Telemetry Neural Networks
HPC Impact: EDA Telemetry Neural Networksinside-BigData.com
 
Biohybrid Robotic Jellyfish for Future Applications in Ocean Monitoring
Biohybrid Robotic Jellyfish for Future Applications in Ocean MonitoringBiohybrid Robotic Jellyfish for Future Applications in Ocean Monitoring
Biohybrid Robotic Jellyfish for Future Applications in Ocean Monitoringinside-BigData.com
 
Machine Learning for Weather Forecasts
Machine Learning for Weather ForecastsMachine Learning for Weather Forecasts
Machine Learning for Weather Forecastsinside-BigData.com
 
HPC AI Advisory Council Update
HPC AI Advisory Council UpdateHPC AI Advisory Council Update
HPC AI Advisory Council Updateinside-BigData.com
 
Fugaku Supercomputer joins fight against COVID-19
Fugaku Supercomputer joins fight against COVID-19Fugaku Supercomputer joins fight against COVID-19
Fugaku Supercomputer joins fight against COVID-19inside-BigData.com
 
Energy Efficient Computing using Dynamic Tuning
Energy Efficient Computing using Dynamic TuningEnergy Efficient Computing using Dynamic Tuning
Energy Efficient Computing using Dynamic Tuninginside-BigData.com
 
HPC at Scale Enabled by DDN A3i and NVIDIA SuperPOD
HPC at Scale Enabled by DDN A3i and NVIDIA SuperPODHPC at Scale Enabled by DDN A3i and NVIDIA SuperPOD
HPC at Scale Enabled by DDN A3i and NVIDIA SuperPODinside-BigData.com
 
Versal Premium ACAP for Network and Cloud Acceleration
Versal Premium ACAP for Network and Cloud AccelerationVersal Premium ACAP for Network and Cloud Acceleration
Versal Premium ACAP for Network and Cloud Accelerationinside-BigData.com
 
Zettar: Moving Massive Amounts of Data across Any Distance Efficiently
Zettar: Moving Massive Amounts of Data across Any Distance EfficientlyZettar: Moving Massive Amounts of Data across Any Distance Efficiently
Zettar: Moving Massive Amounts of Data across Any Distance Efficientlyinside-BigData.com
 
Scaling TCO in a Post Moore's Era
Scaling TCO in a Post Moore's EraScaling TCO in a Post Moore's Era
Scaling TCO in a Post Moore's Erainside-BigData.com
 
CUDA-Python and RAPIDS for blazing fast scientific computing
CUDA-Python and RAPIDS for blazing fast scientific computingCUDA-Python and RAPIDS for blazing fast scientific computing
CUDA-Python and RAPIDS for blazing fast scientific computinginside-BigData.com
 
Introducing HPC with a Raspberry Pi Cluster
Introducing HPC with a Raspberry Pi ClusterIntroducing HPC with a Raspberry Pi Cluster
Introducing HPC with a Raspberry Pi Clusterinside-BigData.com
 

More from inside-BigData.com (20)

Major Market Shifts in IT
Major Market Shifts in ITMajor Market Shifts in IT
Major Market Shifts in IT
 
Preparing to program Aurora at Exascale - Early experiences and future direct...
Preparing to program Aurora at Exascale - Early experiences and future direct...Preparing to program Aurora at Exascale - Early experiences and future direct...
Preparing to program Aurora at Exascale - Early experiences and future direct...
 
Transforming Private 5G Networks
Transforming Private 5G NetworksTransforming Private 5G Networks
Transforming Private 5G Networks
 
The Incorporation of Machine Learning into Scientific Simulations at Lawrence...
The Incorporation of Machine Learning into Scientific Simulations at Lawrence...The Incorporation of Machine Learning into Scientific Simulations at Lawrence...
The Incorporation of Machine Learning into Scientific Simulations at Lawrence...
 
How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...
How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...
How to Achieve High-Performance, Scalable and Distributed DNN Training on Mod...
 
Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...
Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...
Evolving Cyberinfrastructure, Democratizing Data, and Scaling AI to Catalyze ...
 
HPC Impact: EDA Telemetry Neural Networks
HPC Impact: EDA Telemetry Neural NetworksHPC Impact: EDA Telemetry Neural Networks
HPC Impact: EDA Telemetry Neural Networks
 
Biohybrid Robotic Jellyfish for Future Applications in Ocean Monitoring
Biohybrid Robotic Jellyfish for Future Applications in Ocean MonitoringBiohybrid Robotic Jellyfish for Future Applications in Ocean Monitoring
Biohybrid Robotic Jellyfish for Future Applications in Ocean Monitoring
 
Machine Learning for Weather Forecasts
Machine Learning for Weather ForecastsMachine Learning for Weather Forecasts
Machine Learning for Weather Forecasts
 
HPC AI Advisory Council Update
HPC AI Advisory Council UpdateHPC AI Advisory Council Update
HPC AI Advisory Council Update
 
Fugaku Supercomputer joins fight against COVID-19
Fugaku Supercomputer joins fight against COVID-19Fugaku Supercomputer joins fight against COVID-19
Fugaku Supercomputer joins fight against COVID-19
 
Energy Efficient Computing using Dynamic Tuning
Energy Efficient Computing using Dynamic TuningEnergy Efficient Computing using Dynamic Tuning
Energy Efficient Computing using Dynamic Tuning
 
HPC at Scale Enabled by DDN A3i and NVIDIA SuperPOD
HPC at Scale Enabled by DDN A3i and NVIDIA SuperPODHPC at Scale Enabled by DDN A3i and NVIDIA SuperPOD
HPC at Scale Enabled by DDN A3i and NVIDIA SuperPOD
 
State of ARM-based HPC
State of ARM-based HPCState of ARM-based HPC
State of ARM-based HPC
 
Versal Premium ACAP for Network and Cloud Acceleration
Versal Premium ACAP for Network and Cloud AccelerationVersal Premium ACAP for Network and Cloud Acceleration
Versal Premium ACAP for Network and Cloud Acceleration
 
Zettar: Moving Massive Amounts of Data across Any Distance Efficiently
Zettar: Moving Massive Amounts of Data across Any Distance EfficientlyZettar: Moving Massive Amounts of Data across Any Distance Efficiently
Zettar: Moving Massive Amounts of Data across Any Distance Efficiently
 
Scaling TCO in a Post Moore's Era
Scaling TCO in a Post Moore's EraScaling TCO in a Post Moore's Era
Scaling TCO in a Post Moore's Era
 
CUDA-Python and RAPIDS for blazing fast scientific computing
CUDA-Python and RAPIDS for blazing fast scientific computingCUDA-Python and RAPIDS for blazing fast scientific computing
CUDA-Python and RAPIDS for blazing fast scientific computing
 
Introducing HPC with a Raspberry Pi Cluster
Introducing HPC with a Raspberry Pi ClusterIntroducing HPC with a Raspberry Pi Cluster
Introducing HPC with a Raspberry Pi Cluster
 
Overview of HPC Interconnects
Overview of HPC InterconnectsOverview of HPC Interconnects
Overview of HPC Interconnects
 

Recently uploaded

Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfsudhanshuwaghmare1
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)Gabriella Davis
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...apidays
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...apidays
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingEdi Saputra
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘RTylerCroy
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUK Journal
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processorsdebabhi2
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonAnna Loughnan Colquhoun
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MIND CTI
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAndrey Devyatkin
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businesspanagenda
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationSafe Software
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Drew Madelung
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherRemote DBA Services
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?Igalia
 

Recently uploaded (20)

Boost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdfBoost Fertility New Invention Ups Success Rates.pdf
Boost Fertility New Invention Ups Success Rates.pdf
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...Apidays New York 2024 - The value of a flexible API Management solution for O...
Apidays New York 2024 - The value of a flexible API Management solution for O...
 
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
Apidays Singapore 2024 - Building Digital Trust in a Digital Economy by Veron...
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
 
🐬 The future of MySQL is Postgres 🐘
🐬  The future of MySQL is Postgres   🐘🐬  The future of MySQL is Postgres   🐘
🐬 The future of MySQL is Postgres 🐘
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024
 
AWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of TerraformAWS Community Day CPH - Three problems of Terraform
AWS Community Day CPH - Three problems of Terraform
 
Why Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire businessWhy Teams call analytics are critical to your entire business
Why Teams call analytics are critical to your entire business
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
 
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
Strategies for Unlocking Knowledge Management in Microsoft 365 in the Copilot...
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?
 

IO Management with IME 1.1

  • 1. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 11 11 IO Management with IME1.1
  • 2. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 12 What are the IO challenges? Application Developer approach IO Characterisation IO Challenge Multi-physics Non-trivial mix of IO behaviours Filesystem must be optimised for all IO types, not just a small subset Workflows and Ensembles Coordination of files/objects across many jobs. Different IO access patterns from different elements of the workflow Requires a global namespace with strong consistency Adaptive mesh refinement Very difficult to manage and maintain logical application “block” sizes with underlying filesystem Non-optimal access to files, incurs RMW, fs lock contention Metadata- orientation Small, random IOs, Shared file formats Limited by filesystem locking Machine learning High read activity Tough for HDD: disk thrashing Higher Concurrency and heterogeneous CPUs Large thread counts make sequential IO look like random IO avoids all assistance from Filesystem and Storage caches
  • 3. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 13 DDN Confidential How do I make my application go faster? ► Optimal/Natural behaviour already exist within the algorithms ► Difficult to optimise for multiple aspects simultaneously (GPU, multi- threading, network changes, caches) ► Difficult to optimise our application for new environments ► Difficult to maintain efficient porting even between different node counts
  • 4. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 14 DDN Confidential How do I make my application go faster?
  • 5. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 15 DDN | IME DataFlow Fast Data NVM & SSD Persistent Data (Disk) Diverse, high concurrency applications Compute Application issues IO to IME client. Erasure Coding applied IME client sends fragments to IME servers IME servers write buffers to NVM and manage internal metadata IME servers write aligned sequential I/O to SFA backend Parallel File system operates at maximum efficiency
  • 6. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 16 DDN Confidential Applying Flash-Cache Ecomonics ► Real World IO sub-system activity • 99% of the time <30% • 70% of the time <5% ► Build a cache for the peaks, PFS for the capacity ► 10x Performance/Watt improvement for flash-cache over PFS Argonne: P. Carns, K. Harms et al., Understanding and Improving Computational Science Storage Access through Continuous Characterization Parallel File System Scale-Out Flash Cache
  • 7. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 17 problem > client cache problem(s) with Life Science IO ftp ftp-mouse.sanger.ac.uk: anonymous 331 Anonymous login ok, send your complete email address as your password Password: 230 Anonymous access granted ftp> cd REL-1210-BAM ftp> ls –lh -rw-rw-r-- 1 proftpd proftpd 138.5G Dec 7 2009 129P2_OlaHsd.bam -rw-rw-r-- 1 proftpd proftpd 182.3G Oct 9 2012 129S1_SvImJ.bam -rw-rw-r-- 1 proftpd proftpd 60.0G Dec 10 2009 129S5SvEvBrd.bam -rw-rw-r-- 1 proftpd proftpd 107.0G Oct 9 2012 A_J.bam -rw-rw-r-- 1 proftpd proftpd 112.8G Oct 9 2012 AKR_J.bam
  • 8. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 18 IME1.1 | Protected, Fast, Flexible IO IME Server0 IME Server1 IME Server2 IME Server3 streaming writes streaming reads random reads random writes shared file high concurrency Single Shared Namespace to all clients Tackle Tough IO dial-in erasure coding blisteringly fast IOPs wirespeed throughput adaptive IO insane rebuild speeds FILE CACHE
  • 9. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 19 IME240 Performance (EDR) ► 100% of peak SSD performance to clients ► Consistently high performance across IO sizes and SSF/FPP
  • 10. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 20 DDN Confidential IME1.1 | Blistering Sequential Performance File Per Process 1M Seq IO4k Seq IO ► IME240 Sequential performance to file per process over 23GB/s per server (POSIX IO) ► >21GB/s with 4k IOs (MPIIO) ► consistent read and write performance across IO sizes ► POSIX IO performance hits max values at ~32K IO sizes
  • 11. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 21 DDN Confidential IME1.1 | Blistering Sequential Performance Single Shared File 1M Seq IO4k Seq IO ► IME240 Sequential performance for Single Shared File over 22GB/s per server (POSIX IO) ► 15GB/s with 4k IOs (MPIIO) ► consistent read and write performance across IO sizes ► POSIX IO performance hits max values at ~32K IO sizes
  • 12. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 22 DDN Confidential IME | WRF with I/O quilting one I/O process per node Measurement Speed Up IO Wallclock 17.2x Total Execution Wallclock 3.5x ► Application wall time reduced by 3.5 using IME ► I/O time reduces by x17.2 ► Move from I/O bound to compute bound
  • 13. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 23 IME1.1 | True Dial-In Erasure Coding IME Server0 IME Server1 IME Server2 IME ServerN FILE CACHE 8+38+1 6+0 8+2 6+1 4+1 4+14+1 8+0 1+1 ► IME1.1 supports multiple resilience levels through flexible, adaptive erasure coding ► System Wide Default up to 15+3 ► Applications can overide defaults and select a specific Erasure Coding Scheme DEFAULT: 8+3
  • 14. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 24 IME1.1 Node Failure Management ► IME1.1 allows continued IO Service through IME server failure(s) ► Following server failure, IME commences background rebuild of all missing extents for each Parity Group Instance (PGI) : 3+1 across the 3 remaining servers • 1 server will be overloaded with 2 blocks for each PGI • Following a server failure but prior to rebuild completion, immediate read requests may be satisfied through on-the-fly reconstruction of missing file extents ► Service interruption will occur with a second IME server failure in this case (but not with NVM device failures that occur after rebuild) ► New writes still maintain 3+1 (with overloading) ► IME restart is needed to re-introduce failed node and restore original redundancy. (requires full sync to the BFS) 4 IME Servers, Default Parity Schema= 3+1 Continued Production
  • 15. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 25 IME1.1 Node Failure Management ► With higher node counts (>4) and using an erasure coding geometry of higher parity count (>1), IME 1.1 can sustain >1 node failures ► In the below case after one node fails, reads which rely upon missing extents are reconstructed from parity. Writes can continue to be written with 6+2 with full redundancy since we still have 8 nodes ► If another server fails, service continues, with writes at 6+2, but one server will contain 2 objects in each PGI 9 IME Servers, Default Parity Schema= 6+2 Continued Production
  • 16. ddn.com© 2016 DataDirect Networks, Inc. * Other names and brands may be claimed as the property of others. Any statements or representations around future events are subject to change. 26 Apply Scale-Out FLASH-CACHE Economics with Best Performance Density 500GB/s per rack Adaptive IO route around IO bottlenecks Insane Rebuild Speeds minutes not hours Flexible Erasure Coding SW-defined, adaptive N+M Flash-Native minimise write amplification. maximise SSD lifetime IME240 | IME1.1SCALE-OUT | OPTIMISED DATAPATH | FLASH-CACHE | BURST-BUFFER EXTREME IOPS | MASSIVE THROUGHPUT | SOFTWARE DEFINED