CPU, 메모리,
CPU
메모리 디스크,
디스크 I/O 통합
가상화 소프트웨어 vSMP 제품 소개자료
㈜삼부시스템
Aggregate. Scale. Simplify. Save.
회사 소개
• 회사명: ScaleMP Inc
Inc.
•본
사:미국(Cupertino CA)
사:미국(Cupertino,
• 가상화 솔루션 전문 업체(x86
업체( 86 시스템)
Aggregate. Scale. Simplify. Save.
회사 소개
• 2003년도 설립
전세계적으로 300여 사이트 사용
• 2006년 제품 출시
High-end
g
computing
p
g 환경을 위한 가상화
낮은 투자 비용으로 고성능 제공
다수의 X86 서버로
Virtual shared-memory multi-processor(SMP)
구현
Aggregate. Scale. Simplify. Save.
vSMP Foundation란?
산업 표준의 x86시스템 여러대를 하나의 시스템으로 통합시켜주는 가상화
시스템 솔루션
Aggregate. Scale. Simplify. Save.
vSMP Foundation란?
Multiple Computers with Multiple
O
Operating
i Systems
S
Multiple Computers
with
i a Single
Si
Operating
O
i System
S
Aggregate. Scale. Simplify. Save.
vSMP Foundation란?
N x Servers
1 VM
16
xx
16
16
xx
16
OS
16
x
OS
16
OS
16
xxx
OS
16
OS
16
OSx
OS
N x OS
OS
OS
1 OS
Virtualization software for aggregating
gg g
g multiple
p off
off--the
the--shelf systems
y
into a single virtual machine,
providing improved usability and higher performance
Aggregate. Scale. Simplify. Save.
vSMP Foundation Specifications
Processor Supported:
Node
Specifications
Max. sockets per Node:
M
Max.
CPU1 Speed:
S d
Memory (GB) per Node (Min. / Max):
VM
Specifications
U li i d
Unlimited
8 / 2,048
2 / 128
M
Max.
Sockets
S k per VM
VM:
1 024
1,024
Max. CPUs1 per VM:
16,384
64
Fa lt Tolerant2 Support:
Fault
S pport:
Incl ded
Included
Backplane Redundancy3 Support:
Included
Backplane Load Balancing4 Support:
Optional
Partitioning5 Support:
Additional
Software
8
Nodes per VM (Min. / Max):
Max. Memory per VM (TB):
VM
Features
Intel Xeon 5xxx, 6xxx and 7xxx series
Optional (vSMP Foundation for SMP / Cluster)
Direct Connect6 Support:
Included
vSMP Productivity
y Pack7:
Included
Profiler:
Included
Aggregate. Scale. Simplify. Save.
vSMP Foundation Notes
• CPUs1: Represented by a processor core or a logical processor.
• Fault Tolerant2: Automatic restart of the VM upon node failure, isolating the failed node.
• Backplane Redundancy3: Use of 2 parallel connection to one or two InfiniBand switches (single
HCA).
• Backplane Load Balancing (BLB)4: Use up to 4 HCAs in parallel for increased VM bandwidth (up to
160 Gbps per node)
node). BLB also provides Backplane Redundancy
Redundancy.
• Partitioning5: Split of a VM to multiple smaller VMs. Each VM should have atleast 2 nodes.
• Direct Connect6: Connecting up to 4 nodes without InfiniBand switch (back-to-back).
• vSMP Productivity Pack7: Automatic tool for OS tunning and installation of essentials for HPC
operation environments. Available for RHEL and CentOS only.
• Flash drive8: Media loaded with vSMP Foundation software provided by ScaleMP. Available in USB,
CompactFlash or SD card format. Note: USB drive supported only if connected directly to a controller
and
d not via
i USB-hub.
h b
Aggregate. Scale. Simplify. Save.
vSMP 솔루션 사용시 필요 장비
• Compute Blades
Compute Blades
‐ 산업표준 x86 시스템
(HPC 노드, 1U 서버, 등등)
• InfiniBand HCAs, cables a d s tc
and switch
•vSMP Foundation™
산업 표준 x86
x86시스템들을
86시스템들을
86
시스템들을 이용하여 고성능 컴퓨팅 파워 제공
Aggregate. Scale. Simplify. Save.
Certified Hardware
Appro:
GreenBlade: gB222X
D ll
Dell:
•PowerEdge Rack Servers: 1950 III, 2950, R410, R610, R710, R910
•PowerEdge
P
Ed C S
Servers: C6100
•PowerEdge M1000e: M600, M610, M710
HP:
•ProLiant BladeSystem: BL460c, BL280c G61, BL460c G61, BL490c G61
•ProLiant DL: DL380 G7, DL580 G72
IBM:
•System x x3450, x3550 M2, x3650 M2, x3850 X53
•System x iDataPlex: dx340, dx360, dx360 M2, dx360 M3
•BladeCenter H: HS21XM, HS22
Intel:
•S5400SF: SR1560SF / SR1560SFHS
•S5520US: SR1600UR, SR1625UR, SR2600UR, SR2625UR
Systems
Oracle:
•Sun Blade 6000 Family: X6250, X6270, X6275
•Sun Fire Servers: X2270, X4170
Quanta:
QSSC-S4R
•X7DWT-INF / X7DWT-INF+
•X8DTN+
X8DTN+
Supermicro:
•X8DTT series
•SuperBlade: SBI-7125W-S6
Aggregate. Scale. Simplify. Save.
Certified Hardware
Backplane
Infiniband HCA:
Mellanox InfiniHost III, ConnectX, ConnectX-2
InfiniBand Switch: Mellanox switch
•Intel 1GigE
g ((8254x,, 8256x,, 82571-82576,, 80003ES2LAN))
•Broadcom NetExtreme II
Networking:
• 1GigE (5708, 5709, 5715, 5716)
• 10GigE (57710, 57711)
I/O Compatibility
List
Storage:
Others:
•ATA cotrollers
•SATA controllers: AHCI compatible
•LSI PCIe SAS and Fibre Channel HBAs (MPT):
• SAS/SATA: LSISAS3041E-R, LSISAS3081E-R, LSISAS380
1E LSISAS3442E-R
1E,
LSISAS3442E R
• FC: LSI7104EP-LC, LSI7204EP-LC, LSI7404EP-LC
•LSI PCIe Internal RAID adapters (MegaRAID®):
• SAS/SATA: 870xELP, 870xEM2, 8880EM2, 8888ELP, 92xx
• OEM devices:
• Dell: PERC 5, PERC 6, CERC 6, H700, H800
•Emulex PCIe Fibre Channel HBAs (LightPulse):
• LPe11xx, LPe12xx
Fi board
b d KVM/USB d
devices:
i
VGA
•First
•PS2
•USB 1.1 (UHCI only)
Aggregate. Scale. Simplify. Save.
Supported Operating Systems
• Linux 64 bit OS, kernel level 2.6.11 or later.
• ScaleMP’s standard support includes servicing users running
Red Hat Enterprise Linux 5 (RHEL5) and its clones as well as
Novell SuSE Linux Enterprise Server 11 (SLES11).
Aggregate. Scale. Simplify. Save.
vSMP vs VMM 비교
PARTITIONING
AGGREGATION
Subset of the physical resource
Concatenation of physical resources
Virtual Machines
Virtual Machine
App
OS
App
OS
App
OS
App
OS
Hypervisor or VMM
Hypervisor
or VMM
Aggregate. Scale. Simplify. Save.
Hypervisor
or VMM
Hypervisor
or VMM
Hypervisor
or VMM
Scale-Up vs Scale-Out 비교
Applications requiring li i
i i supersett
of the physical server resources
Scale-Up
Scale-Out
Fit the hardware to the problem size
Fit the hardware to the problem size
Break the problem to fit the hardware
Break the problem to fit the hardware
Aggregate. Scale. Simplify. Save.
Scale-Up vs Scale-Out 장단점 비교
Unix
Scale-up
Scale
up
X86
Scale-out
Scale out
장점:
단점:
간편한 단일 시스템 구성
설치 및 관리 비용 증가
• 복잡한 병렬 프로그래밍
• 멀티 O/S 시스템
• 클러스터 파일 시스템
• 쉽고 유연한 프로그래밍
• 단일 시스템 관리
• I/O 통합
단점:
단점
장점:
종속적인 하드웨어 디자인
표준화 된 디자인
• 고 비용
• 고정 된 하드웨어 아키텍쳐
• 낮은 비용
• Open 아키텍쳐
Aggregate. Scale. Simplify. Save.
vSMP 의 장단점 통합
Scale-Up
Scale-Out
간편한 단일 시스템 구성
A
Aggregation
ti
Virtual Machine
• 쉽고 유연한 프로그래밍
• 단일 시스템 관리
• I/O 통합
+
App
OS
표준화 된 디자인
Hypervisor or
VMM
Hypervisor or
VMM
Hypervisor or
VMM
Hypervisor or
VMM
• 낮은 비용
• Open
p 아키텍쳐
Aggregate. Scale. Simplify. Save.
+
vSMP 사용시 장점
•
•
•
컴퓨팅 파워 향상(고속 연산과 고성능 제공)
- 가상의 하나의 시스템으로 통합되기 때문에 단일 운영체제(OS) 로 운영
- 저가형 x86 시스템들을 단일 시스템
시스템으로
로 통합
- 대용량 공유 메모리를 제공
복잡하고 어려운 클러스터 기반의 Application 개발 단순화와 혁기적인 기간 단축
TCO 비용 절감
- 고가의 유닉스 서버 대신 저가형 x86시스템으로 대체 가능.
- 단순하고 쉬운 관리(여러대의 x86시스템들을 하나의 시스템으로 통합하여 관리)
APP
APP
APP
APP
APP
OS
OS
OS
OS
OS
APP
APP
OS
OS
APP
APP
OS
OS
Aggregate. Scale. Simplify. Save.
vSMP 솔루션 구성 방법
Aggregate. Scale. Simplify. Save.
vSMP 솔루션 구성 방법
Aggregate. Scale. Simplify. Save.
vSMP 솔루션 구성 방법
Aggregate. Scale. Simplify. Save.
SPEC CPU2000(CPU 벤치 마크 툴)
Aggregate. Scale. Simplify. Save.
STREAM (memory-bandwidth 벤치마크 툴)
참조 사이트:http://www.cs.virginia.edu/stream/standard/Bandwidth.html
사이트:http://www cs virginia edu/stream/standard/Bandwidth html
Aggregate. Scale. Simplify. Save.
지원 어플리케이션
• Users seeking to simplify cluster complexities
• Applications that use large memory l
footprint (even with one processor)
• Applications that need multiple processors and
processors and shared memory
Manufacturing
CSM ((Computational p
Structural Mechanics)
ABAQUS/Explicit
ABAQUS/Standard
/
ANSYS Mechanical
LSTC LS‐‐DYNA
LSTC LS
ALTAIR Radioss
CFD (Computational Fluid Dynamics)
FLUENT
ANSYS CFX
STAR‐‐CD
STAR
AVL FIRE
Tgrid
Other
inTrace OpenRT
Life Sciences
Gaussian
VASP
AMBER
Schrödinger Jaguar
Schrödinger Glide
g
NAMD
DOCK
GAMESS
GOLD
mpiBLAST
GROMACS
MOLPRO
OpenEye FRED
p y
OpenEye OMEGA
SCM ADF
HMMER
Weather
Weather Forecasting
MM5
WRF
Aggregate. Scale. Simplify. Save.
Energy
Schlumberger ECLIPSE
Paradigm GeoDepth
3DGEO 3DPSDM
Norsar 3D
EDA
Mentor
Cadence
Synopsys
Finance
Wombat
KX
Others
The MathWorks
The MathWorks
MATLAB
R
Octave
Wolfram MATHEMATICA
ISC STAR‐‐P
ISC STAR
7/26/2012
FORMULA1 TEAM
•
Customer: Formula1 team
•
Current platform: Large
Large-memory
memory Itanium-based
Itanium based system
•
Problems:
– Need to generate large mesh as part of pre-processing of whole-car
simulation ((FLUENT TGrid))
– Mesh requirements are ~200GB in size
– Expect to grow significantly within 12 months after initial deployment
– Would like to standardize on x86 architecture due to lower costs and open
standards
•
Solution:
– 12 Intel dual-processor Xeon systems to provide 384GB RAM single virtual
system running Linux with vSMP Foundation
•
SCALEUP
AT SCALEOUT
Solution evaluated and found to be faster than PRICING
Benefits:
– Better performance:
alternative systems (x86 and non-x86)
– Cost: Significant savings compared to alternative system
– Versatility: Also being used to run FLUENT (MPI) as part of large cluster
– Investment protection: Solution can grow
Aggregate. Scale. Simplify. Save.
7/26/2012
24
ENGINEERING SERVICES COMPANY
•
Customer: Mid-size Engineering Services Company
•
Current platform: Multiple 2-socket workstations
•
Problems:
– Existing models (Abaqus) grow fast and can’t fit the engineers workstation
– Interested in running apps in batch at night
– No in-house skills to run x86 InfiniBand cluster (although the application runs nicely on
InfiniBand cluster) . Can’t afford RISC systems
•
Solution:
– 4 Intel dual-processor Xeon systems to provide 128GB RAM, 8 sockets (16 cores) single
virtual system running Linux with vSMP Foundation
•
Benefits:
– Performance: Solution significantly faster than existing workstations. Performance is
comparable to cluster performance (using vendor benchmarks)
benchmarks).
INNOVATION – Low OPEX: No IT required for day-to-day operation
WITHOUT – Versatility: Batch mode at night. Daytime jobs are executed on the system
COMPLEXITY
while using the workstation for display only. Multi-user environment with
perfect scaling – and sharing without performance degradation.
– Investment protection: Expected to expand the system by adding additional
4 nodes (to a total of 256GB RAM, 32 cores)
Aggregate. Scale. Simplify. Save.
7/26/2012
25
GLOBAL ENERGY COMPANY
•
Customer: Global Energy Company
•
Current platform: x86 grid
•
Problems:
– Using in-house single-threaded simulation tools in throughput mode. Each
simulation memory footprint has grown over the years and sometimes (10%)
exceeds 32GB.
– Application runs on x86 only
– Used to reschedule failed runs on large-memory systems
•
Solution:
– 6 Intel dual-processor
dual processor Xeon systems to provide 192GB RAM
RAM, 12 sockets (48
cores) single virtual system running Linux with vSMP Foundation
•
Benefits:
– Versatility: Both large and small workloads used concurrently on the same
system
– Utilization: Higher utilization compared to grid due to lower infrastructure
fragmentation
– Investment protection: Solution expanded by 100% since initial installation
SINGLE INFRASTRUCTURE FOR HORIZONTAL AND
VERTICAL APPLICATION SCALING – PLUG & PLAY
Aggregate. Scale. Simplify. Save.
7/26/2012
26
MEDICAL RESEARCH INSTITUTE
•
Customer: Medical Research Institute
•
Current platform: IBM P-5 System
•
Problems:
– Need to perform high performance image processing on very large MRI scans
– Scanned data for a single run is currently over 200GB. Memory requirements are
expected to grow significantly with the introduction of full body scan with more
sensors
– Would like the ability to use OpenMP for faster prototyping
– Would like to standardize on x86 architecture due to lower costs and open
standards
•
Solution:
– 16 Intel dual-processor Xeon systems to provide 1TB RAM, 32 sockets (128
cores) single virtual system running Linux with vSMP Foundation
•
Benefits:
– Better performance: Solution evaluated and found to be faster than any other
alternative system
– Cost: Significant savings compared to alternative system (order of magnitude)
– Versatility: Also being used to run FLUENT (MPI) as part of large cluster
LARGE MEMORY FOR MULTI‐THREADED PROGRAMMING
Aggregate. Scale. Simplify. Save.
7/26/2012
27
FINANCIAL SERVICES
•
Customer: Hedge Fund
•
Current platform: Multiple 4-Socket Servers
•
Problems:
– A single 4-socket server did not provide enough performance required for
customer business targets
– Multiple 4-socket
4 socket servers required complex decomposition and introduced
challenges in transferring data between processes in a short and deterministic
time (low latency and small jitters)
• Ethernet based solution could not provide this / IB solution is too complex to manage
and program for
– Co
Co-location
location at exchanges for a solution comprised of multiple systems is
complicate
•
Solution:
– 16 Intel dual-processor Xeon systems to provide 0.5TB RAM, 32 sockets (128
cores) single virtual system running Linux with vSMP Foundation
•
Benefits:
– Reduced latency and latency variance
– Simpler solution: Deploy and management of a single system
– Better utilization: Having a single system reduces resources fragmentation
SIMPLIFYING INTER‐PROCESS COMMUNICATION
Aggregate. Scale. Simplify. Save.
7/26/2012
28
㈜삼부시스템
기술영업 부장 안대준
TEL:02-538-4001/017-346-8590
/
E-mail: [email protected]
Aggregate. Scale. Simplify. Save.