[OpenAFS] calculating memory

Benjamin Kaduk kaduk@MIT.EDU
Tue, 1 Feb 2011 00:58:11 -0500 (EST)


On Mon, 31 Jan 2011, Harald Barth wrote:

>
>> I am in charge of several afs servers in our college. Right now
>> there are 5 afs servers running on 5 SPARC based servers. We are
>> ditching Solaris since it sucks so bad and are going to move to
>> Linux VM's running inside of VMware.
>
> Before you do the move into a VM, check that your VM is up to the
> number of IO/sec that an AFS server can make. I know that su.se
> moved all there AFS servers into a VM solution and then out again
> because of performace issues under load from many concurrent users.

I don't administer the athena cell, but I do know that its servers are VMs 
(presumably amply resourced).  I believe they are backed by a SAN, and not 
virtualized disk, though.  This was done at the same time as moving the 
servers to an datacenter that is not on campus proper; qualitatively it 
seems that we have experienced a slight increase in AFS "blips", but it is 
very hard to quantify or track down.  It may just be a result of 
congestion on the link between main campus and this datacenter.

Just adding a data point that it can be done, if the conditions are right.

-Ben Kaduk