Download presentation
Presentation is loading. Please wait.
Published byPearl Phelps Modified over 9 years ago
1
High Performance Computing G Burton – ICG – Oct12 – v1.1 1
2
Agenda Commodity Clusters Compute Pool Interconnects and Networks Shared Storage Login Layer and Middleware
3
HPC alternatives
4
HPC Facts IBM Sequoia - Number 1 in top 500 with 1.572 million Cores. 20 petaFLOPS ( Floating Point Operations / Second ). Sciama 10 teraFLOPS. China has Number 5 plus 62 other systems in top 500, now ahead of Germany, UK, Japan and France. 75% of Top 500 use Intel processors.
5
Demystifying the Techno Babble
6
Demystifying the Techno Babble (2)
7
Commodity Clusters 7
8
Made from commodity (off-the-shelf) components (read PC’s). Consequently (relatively) cheap. Usually Linux based High availability storage (no single point of failure) Generic compute pool (cloned servers that can easily be replaced).
9
Cluster Concept
10
10 Compute Pool - Just a bunch of PC’s
11
In the “good-ol-days” things were simple ………. 11
12
In the “good-ol-days” things were simple ………. 12
13
… these days much more packed into the same space … but basically the same! 13 These are the building blocks of HPC similar to Sciama
14
Total ICG Compute Pool > 1000 Cores 14
15
Coke or Pepsi – Chalk & Cheese Only 2 remaining commodity CPU makers are Intel and AMD. Latest AMD “Bulldozer” architecture competing with Intel “Sandy Bridge” Architecture. Both architectures are multi core ( Intel 48 cores) Architectures use same memory / video cards / hard drives etc CPU speed constraints down to on-chip transmission delays and heat dissipation ( 22nm ).
16
Intel-AMD – Bangs per Buck
17
Graphical Processing Units (GPU’s actually GPGPU’s – General purpose ) CPU’s still in charge Special programming language. CUDA and OpenCL Three players:- Intel AMD Nvidia Cpu – multiple cores Gpu – 100’s of cores
18
Interconnects and Networks 18
19
Interconnects and Networks Moving away from the Processor towards the Internet you get slower and slower due to Increased Latency and Reduced Bandwidth
20
Processor Interconnects For processor running at 3.2GHz – QPI Bus would be 25GBytes / Second ( Kindle version of “War & Peace” is 2GBytes)
21
Peripheral Component Interconnect Express (PCIe) PCI bus is interconnect to the outside world
22
External Networks Interconnects are Parallel – Bytes / second Networks are serial – bits / second (shown here in B/s for comparison – eg. DAS is 6Gb/s)
23
Sciama Network
24
Sciama Traffic
25
Connecting to Sciama
26
26 Shared Storage
27
Raw Disks are Dumb Remember: PATA, IDE (Advanced Technology Attachment)
28
Intelligence is in the File System
29
HPC’s Require many disks
30
Use High Capacity Arrays
31
HPC’s require large chucks of storage
32
Many RAID options 1-6 / 10 /50
33
Directly Attached Storage (DAS)
34
Of limited use as cannot be shared.
35
Network Attached Storage (NAS)
36
NAS or Network Appliance
37
Network BW is often the bottleneck
38
NAS - Lustre File System Lustre is an example of a distributed file system. There are many more. Sometimes called a “Cluster” file system
39
NAS – Lustre Often used with an Infiniband fabric.
40
Storage Area Networks
41
Storage Area Network using iSCSI
42
Fibre Channel High Availability SAN
43
Sciama NAS Storage
44
Sciama Storage Hardware Storage is expensive. 250gbp / Tbytes No Backup
45
Highly Available Hardware
46
Two paths to most components
47
Additional Sciama Storage /mnt/astro(1-5) 10 GbE
48
Login Layer and Middleware 48
49
Why Login Servers Login servers will provide the gateway to the cluster. Users can remotely login into the servers using “ssh” or a Remote Desktop Client. A desktop client gives a full working desktop in the environment (can full screen)
50
Some users are at remote locations.. 50
51
Use of Remote Login Client 51
52
Executable and Jobscript setup in Login Layer
53
Jobs submitted to the queues > qsub –q Queue1 run_script.sh
54
Scheduler prioritises and deems a job ready to run. Job passed to resource manager
55
Resource manager (Torque) checks for available resources.
56
Job either runs in the compute pool or returns to queue
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.