Re: [Wg-test-framework] Need to adjust hardware list.

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Marcos E. Matsunaga writes ("Re: [Wg-test-framework] Need to adjust hardware 
list."):
> 1. After thinking a little on the subject, I don't see why not. Having 
> two of each doesn't make much sense to me as we can have a wider variety 
> of environments/configuration having 24 different servers. For 
> infra-structure , I'd say that we must have some redundancy. They don't 
> have to be exact match, but similar.

Having two of each machine is really very helpful.  We often have (a)
hardware failures and (b) machine-specific bugs.  If every machine is
part of an identical pair then it is easy to distinguish these two
cases, which is important because the appropriate responses are very
different.

Also, when we have a machine-specific bug, the test system starts to
focus on that specific bug, which causes that specific machine to be
extra busy.  Having another instance of it makes that less of a
throughput problem.

> 4. I think custom building our own servers would bring a whole new 
> different beast in the mix. Someone would have to take time selecting 
> parts, building, testing and make sure it is all working properly.

We have to do a commissioning test anyway.  The risk is that we would
have to iterate in cases of problems.

> 5. Dell has the R415 that uses AMD processor 
> (http://www.dell.com/us/business/p/poweredge-r415/pd)
> HP ProLiant DL385p Gen8 Server also uses AMD processor 
> (http://www8.hp.com/us/en/products/proliant-servers/product-detail.html?oid=5249584)
> IBM System x3755 M3 is another one that uses AMD processor 
> (http://www-03.ibm.com/systems/x/hardware/rack/x3755m3/index.html)

Thanks for the suggestions.  I haven't looked at the pricing of these
compared to the other systems on the list.  But I will forward these
suggestions to All-Net.

> There are plenty of other not so known sources of 1U rack mount AMD 
> servers, but don't know if I'd trust them.

I have had fine experiences with no-name servers.  In practice they
don't seem to have a higher failure rate than big-name branded ones.

And anyway hardware failure is not a crisis for our test system, so
long as it doesn't happen at an unreasonable rate.  Failed machines
are pretty easy to spot and can simply be taken out of the test pool
until they are mended.

Ian.

_______________________________________________
Wg-test-framework mailing list
Wg-test-framework@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/wg-test-framework