infrastructure optimization for seismic processing (eng)
TRANSCRIPT
Performance Boost forSeismic Processingwith right IT infrastructure
Vsevolod Shabad [email protected] and founder +7 (985) 765-76-03
NetProject at a glance
System integrator with strong oil&gas focus:we build and expand the IT infrastructurefor Geological and Geophysical applications
2
How to make the HPC right balanced?
The applications performance must be limited only by the licenses:
•Paradigm ES 360, GeoDepth, Echos•Schlumberger Omega•Landmark SeisSpace ProMAX•CGG Geocluster
with minimal overall cost
3
Cluster components utilization
30%
40%
50%
60%
70%
80%
90%
100%
Optimal
(limited only by CPU)
Suboptimal
(nothing limits - excessive resources)
Suboptimal
(limited by Network but not CPU)
Cluster performance must be limited only by the CPU!
CPU RAM Network Local HDD Shared File System
4
Key ideas for optimization
Massive reduction:•CPU downtime for compute nodes•CPU load by auxiliary tasks•operating costs
5
The typical sample of HPC CPU load during workday
(real case)
Candidates for optimization
Compute nodes
File nodes and storage
Management nodes and resource schedulers
Cluster interconnect
Backup and restore subsystem
Monitoring and management subsystem
Cooling subsystem
6
Compute nodes optimization
CPU & RAM: optimal choice
CPU offload with RDMA
CPU offload with GPU
Right form factor of servers
7
CPU & RAM: optimal choice sample
Part Haswell-EP (Xeon E5-2600 v3) Broadwell-EP (E5-2600 v4)
Server type HPE ProLiant XL230a Gen9
CPU 2x Xeon E5-2650 v3
(20 cores, 2.3 GHz)
2x Xeon E5-2680 v4
(28 cores, 2.4 GHz)
RAM 160 GB (8 GB/core)
PC4-2133
224 GB (8 GB/core)
PC4-2400
HDD 2x 1.8 TB SAS 10K RPM
Network 1x Infiniband FDR, 4x GigE
List Price, USD 17 456 18 366
Performance, TFLOPS 0.74 1.08
List Price per core, USD 872.8 655.9 (–24.8%)
List Price per TFLOPS, USD 23 713.4 17 081.5 (–28.0%)
8
Typical compute node for Paradigm ES 360
CPU & RAM: non-optimal choice sample
Part Broadwell-EP (E5-2600 v4) Broadwell-EP (E5-2600 v4)
Server type HPE ProLiant XL230a Gen9
CPU 2x Xeon E5-2680 v4
(28 cores, 2.4 GHz)
2x Xeon E5-2699 v4
(44 cores, 2.2 GHz)
RAM 224 GB (8 GB/core)
PC4-2400
352 GB (8 GB/core)
PC4-2400
HDD 2x 1.8 TB SAS 10K RPM
Network 1x Infiniband FDR, 4x GigE
List Price, USD 18 366 36 396
Performance, TFLOPS 1.08 1.55
List Price per core, USD 655.9 827.2 (+26.1%)
List Price per TFLOPS, USD 17 081.5 23 499,5 (+37,6%)
9
Typical compute node for Paradigm ES 360
Some non-obvious restrictions
HPE Proliant XL230a Gen9:•can‘t work with single CPU
•two same processors required•can‘t use the 32GB RDIMM for DDR4 RAM
•only LRDIMM for 32GB DDR4 allowed•…
Fujitsu CX400 M1:•can‘t use two Infiniband HBAs
with liquid cooling•only one Infiniband HBA allowed
•…
…
10
CPU offload with RDMA
Direct buffer-to-buffer data move without CPU and operating system
CPU offload with RDMA
RDMA usage in HPC:•between compute nodes•between compute node and file node (storage)
12
Sample of RDMA efficiency
13
Source: http://cto.vmware.com/wp-
content/uploads/2012/09/RDMAonvSphere.pdf
RDMA must be supported by the apps!
Seismic processing:•Paradigm ES 360, GeoDepth, Echos
•since version 15.5•Schlumberger Omega, Landmark ProMAX
•with MPI
File systems:• IBM Spectrum Scale (ex GPFS)•Lustre •BeeGFS (ex FhGFS)
14
CPU offload with GPU
Paradigm:•Echos RTM
Schlumberger:•Omega RTM, Kirchhoff
Tsunami:•Tsunami RTM
…
15
Full list: http://www.nvidia.co.uk/content/EMEAI/PDF/tesla-
gpu-applications/gpu-apps-catalog-eu.pdf
Sample of GPU efficiency
Statoil custom seismic processing application
NVIDIA GPU (GeForce GTX280)
Source: http://www.idi.ntnu.no/~elster/master-studs/owej/owe-johansen-master-
ntnu.pdf
16
Right form factor of servers
Rack-optimized 1U•HPE Proliant DL360 Gen9, …
Rack-optimized 2U•Fujitsu RX2540 M2, …
High-density servers•Lenovo NeXtScale M5, …
Blade servers•DELL M1000e, …
17
Typical high-density chassis
Chassis Number of nodes Chassis height, RU
Lenovo NeXtScale n1200 12 6
HPE Apollo 2000 4 2
HPE Apollo a6000 10 5 (+1.5)
Fujitsu Primergy CX400 S2 4 2
Huawei X6800 8 4
DELL PowerEdge FX2 8 2
DELL PowerEdge C6320 4 2
18
File nodes and storage
Right choice of file system with RDMA и QoS
High-density block storage arrays
Short IOPS caching with SSD
File system overhead reduction
Backup to Redirect-On-Write snapshots
Transparent data migration to tapes
19
Right choice of file system
IBM
Spectrum
Scale
(GPFS)
Lustre
(+ ZFS)
BeeGFS Panasas
PanFS
EMC Isilon
Huawei 9000
NetApp FAS(NFS)
RDMA support
at client side
Yes Yes Yes No No
Redirect-On-Write
snapshot support
Yes Yes
(with ZFS)
No Yes Yes
Transparent
migration to tape
Yes No No No No
Commercial support Yes “Yes”
(no for ZFS-
on-Linux)
Yes Yes Yes
Installation difficulty High High High Low Low
Single-Thread
performance
Moderate Low High Low Low
20
File system overhead sample
21
Source: http://wiki.lustre.org/images/b/b2/Lustre_on_ZFS-Ricardo.pdf
Right choice of block storage array
Requirements sample (moderate cluster for SLB Omega):
• I/O throughput — 3 GB/sec on sequential I/O•usable capacity — 300 TB
Solutions:•EMC VNX5400 — 174 spindles, 39 RU•NetApp E5600 — 69 spindles, 12 RU
Front-end I/O ports options:•SAS, Fibre Channel, 10GbE, Infiniband
NetApp E5600 disk array supports iSER connection via Infiniband fabric (without FC & SAS)
22
Short IOPS caching with SSD
Most applicable for:•metadata•trace headers
File system support:• IBM Spectrum Scale:
•Highly-available write cache (HAWC)•Local Read-Only Cache (LROC)
•Panasas ActiveStor:• internal Panasas ActiveStor capabilities
•Lustre:•no
•EMC Isilon:•SmartFlash
23
Backup to Redirect-On-Write snapshots
Traditional backup to tapes:•500 TB of data,
verification after write•four LTO-7 (300 MB/sec) drives:
•500 000 000 / (4 * 300 * 2 * 3600) = 58 hoursunder ideal conditions
Innovative backup (snapshots):•10 minutes
regardless of data volume
24
Transparent data migration to tapes
Middleware: IBM Spectrum Archive (LTFS)
Drives IBM TS1150: 360 MB/s, 10 TB per cartridge
25
Resource schedulers
Cluster usage alignment and optimization
Energy aware scheduling
Top-3 systems:• IBM Spectrum LSF
(Platform LSF)•Altair PBS Pro•Adaptive Computing
MOAB HPC Suite
26
Cluster interconnect
Key technology — RDMA
Two options:• Infiniband FDR/EDR•Ethernet 40G/56G/100G (with RoCE)
27
Backup and restore subsystem
Protection from logical data damage•with Redirect-On-Write snapshots
Protection from от physical data damageis not necessary
•with dedicated seismic data archive on LTO-7 tapes
28
Snapshot technologies comparative
Redirect-On-Write Copy-On-WriteNetApp FAS3240 IBM Storwize v7000
Source: NetProject’s comparative testing for JSC NOVATEK (2012)
29
Monitoring and management
Dedicated Gigabit Ethernet fabric
Management software:•Altair PBS Pro + HPE CMU• IBM Platform Cluster Manager
30
Cooling subsystem
Air conditioning:•applicable for any equipment (up to 15 кW/rack)•temperature range could be expanded
Liquid conditioning:•equipment adaptation
or “liquid coolingrack door” required
•high level of data center density
31
Temperature range expansion
4% operational savings from cooling for every 10 C increase in operating temperature (Intel, IDC, Gartner)
Incompatible:•with NVIDIA GRID•with CPU > 120W
ATD module price for Fujitsu Primergy is 26 USD!
32
Our product portfolio:
Servers:•Lenovo, HPE, DELL, Fujitsu, Cisco, Huawei, Intel,
Supermicro, Inspur, Sugon, …
Storage:•NetApp, IBM, HDS, Panasas, Huawei, EMC, …
Network:•Mellanox, Cisco, Lenovo, Huawei, Brocade
Tape libraries:• IBM, HPE, Quantum
Resource scheduling tools:• IBM Platform, Altair, Adaptive Computing
33
Company strengths
Strong industry focus on G&G IT infrastructure
Deep knowledge and experience:• Industry specifics and major applications:
•better than most system integrators and IT infrastructure vendors
• IT infrastructure products and technologies•better than most G&G software vendors
Advanced project management methodology
Structured knowledgebase of past projects
High engineering culture of staff
Deep customer involvement in projects
34
Company weaknesses
Strong industry focus on G&G IT infrastructure
Low level of company brand recognition
Lack of overseas market experience
A small number of employees
Limited financial resources
35
Company background in a brief
Technology experience:•Networking since 1996•Storage since 2004•Servers since 2006•VDI since 2010•HPC since 2014
Oil & Gas industry focus since 2012•metallurgy and banking focus in 1996-2014
Big Data experience since 2007
ISO 9001:2008 certified since 2012
36
Partner reference about NetProject
At the last time we are increasingly ask NetProject for
consulting about hardware configurations, clusters,
workstations, and infrastructure solutions.
This company's team has participated in the creation of the
various data center architectures for the wide spectrum of
Paradigm technologies and has proved itself from the best side.
Serge Levin,Paradigm Geophysical (Russia) Sales Director
37
