SGI UV 1000 Cluster Manager Problem

Open discussion on or around SGI; software/hardware related posts should go in the appropriate subforum.
Forum rules
Any posts concerning pirated software or offering to buy/sell/trade commercial software are subject to removal.
adlihajarat
Posts: 7
Joined: Sat Sep 20, 2014 1:20 pm

SGI UV 1000 Cluster Manager Problem

Unread postby adlihajarat » Wed Oct 18, 2017 4:05 pm

When I test the BMCs on the cluster, they show that they have 0B memory, although I installed 64 GB of RAM on the BASEIO blade and the same of the disk blade.
What could be wrong ??

User avatar
japes
Donor
Donor
Posts: 996
Joined: Thu Nov 08, 2007 4:35 pm
Location: Lynnwood, WA

Re: SGI UV 1000 Cluster Manager Problem

Unread postby japes » Wed Oct 18, 2017 10:14 pm

That's pretty new for most of us here. I'm not sure many people here have seen a UV system much less operated one.

Can you copy from your terminal some of the output that you're seeing? I'm assuming from the CMC console?

Do you have more blades in the system besides BASIO and Disk? I would expect you're going to need memory in each blade you have attached. I haven't touched one of these, but looking at some photos of blades on ebay I'm wondering what the different DIMMs are, I see ones on the main board of the blade, but a smaller set of DIMMs on a riser in photos. If I had to guess I'd think perhaps directory memory.
:O3000: :Fuel: :Tezro: :Tezro: :Octane2: :Octane: :Octane: :Indigo: :Indigo: :Indigo: :Indigo: :O2: :1600SW: :O2: :1600SW: :1600SW: :Indigo2: :Indigo2: :Indigo2: :Indigo2: :Indigo2IMP: :Indy: :Indy: :Indy: :Indy: :O3x0: :O3x02L: :O3x02L:

adlihajarat
Posts: 7
Joined: Sat Sep 20, 2014 1:20 pm

Re: SGI UV 1000 Cluster Manager Problem

Unread postby adlihajarat » Sat Oct 21, 2017 5:53 am

Dear Japes,

Yes you are right, there are two types of memory on each blade system board, one is the system memory and the other one is the cache memory for the CC-NUMA links.
I do have 2 BASEIO blades and 30 compute blades.
I am working on the issue by resetting the partitions on the blades and using the SGI Cluster Management Center.
Will keep you posted.

adlihajarat
Posts: 7
Joined: Sat Sep 20, 2014 1:20 pm

Re: SGI UV 1000 Cluster Manager Problem

Unread postby adlihajarat » Sun Oct 29, 2017 6:38 am

This might be interesting to some, the design of all the blades is the same. The only difference is the add-on card. This turns the blade into BASEIO if you have the BASEIO card, PCIE if you have the PCIE card, or boot disk blade if you have the card for it.

The most important drawbacks of the system are:
1- It is based of DDR3 ECC RAM PC3 8500R series
2- Based on intel Westmere chipset and the distinct E-7 Xeon series
3- The system is based on cache coherency, so if the cache is not good, then the system will not work properly
4- The system boot process takes very long time. In my case on the UV 100 takes about 5 mins to boot the BASEIO + one PCIE blade.

Advantages of the system:
If the system is working properly, then you will have the following advantages:
1- The Whole number of CPUs available will be as if you have a single CPU with all available cores. Upto 2560 cores/ 4096 Threads $$$$$$$$$
2- Same is for the system memory which can go up to 4TB per 42U rack, total system memory in 4 racks is 16 TB RAM. $$$$$$
3- The system can be used to solve huge size problems especially in the field of Genoms and CFD, Engineering and Numerical Analysis.
4- The system can work with compute intensive devices ie. TESLA and Intel PHI .
5- Can boot to Linux and Windows Server/Data Center.

Disadvantages:

1- The system is considered obsolete and has no support, you are on you own, or try to discuss it with HPE.
2- Parts are very expensive, and hard to get even over E-bay.

This is all I have to say, open to any comments or questions.


Return to “SGI: Discussion”

Who is online

Users browsing this forum: No registered users and 1 guest