I don't know what Jason's customer's reasons are, but at least ones that we
have considered are:

1. More flexibility in resource allocation, both in terms of adjusting
sizing and security isolation

An example I have is that we ended up over-speccing our current Sun Ray
infrastructure.  We have 180 DTUs or so deployed, but most of the time the
highest utilization we've seen is about 50 concurrent users.  One server is
sufficient to support that, while we had originally specified 5 in order to
support the full number of users with redundancy.  We have since knocked
that down to 3 but don't want to reduce it further because we would lose too
much redundancy.  If we virtualized that workload, we could half the size of
the SRSS VMs so that the other half of each of the 3 servers could be used
for other services.  At the same time we maintain the same level of
redundancy as before.  Running those services on the Sun Ray servers
themselves wouldn't be desirable for security reasons, among others.  We
currently also have a pre-production testbed Sun Ray server that is heavily
underutilized most of the time, but for testing reasons is
hardware-identical to the production systems that can each support 50 users.
In addition, sometimes we have various syadmins that want to be able to run
multiple tests but require separate environments to do so.  If we virtualize
that system, we can have multiple test environments with the same amount of
hardware, and could even potentially leverage the production server
resources during off hours in order to facilitate parallel testing. 

2. Better mobility and availability (live migration of VM to perform
physical maintenance or accommodate other downtime)

Live migrate an SRSS VM to another VM host.  Users notice nothing while the
physical server is taken down to replace a failed DIMM or upgrade the host
OS/hypervisor.  Or install a new server to replace the 5-year old one and
simply plop the old VM onto the new hardware.

William Yang

> -----Original Message-----
> From: [email protected] [mailto:sunray-users-
> [email protected]] On Behalf Of John Francis
> Sent: Thursday, February 04, 2010 10:11 PM
> To: SunRay-Users mailing list
> Subject: Re: [SunRay-Users] Sizing SRSS on VMWare ESX
> 
> On 5 February 2010 12:28, William Yang <[email protected]> wrote:
> >
> >> -----Original Message-----
> >> From: [email protected] [mailto:sunray-users-
> >> [email protected]] On Behalf Of Ceri Davies
> >> Sent: Thursday, February 04, 2010 5:36 PM
> >> To: [email protected]; SunRay-Users mailing list
> >> Subject: Re: [SunRay-Users] Sizing SRSS on VMWare ESX
> >>
> >> On Thu, Feb 04, 2010 at 03:36:23PM -0400, Jason Doyle wrote:
> >> > Looking for some feedback from the field ....
> >> >
> >> > I have a customer with a deployment of a couple dual-socket Xeon
> servers
> >> to
> >> > host 100 DTU's and they're looking to grow this environment to
> support
> >> an
> >> > additional 350. Instead of deploying more physical servers they'd
> like
> >> to
> >> > deploy SRSS on Solaris VM's (VMware ESX) and add these to the FOG. My
> >> > questions are:
> >> >
> 
> Since you are looking to support a large number of DTUs/users, and by
> the sound of it you plan on offering native Solaris desktop sessions,
> I have to ask why is virtualising a good idea?  It doesn't sound like
> your servers would be under utilised.
> 
> Not flaming, just genuinely interested in hearing thoughts on this topic.
> 
> --
> Kind regards,
> 
> John Francis
> _______________________________________________
> SunRay-Users mailing list
> [email protected]
> http://www.filibeto.org/mailman/listinfo/sunray-users

_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users

Reply via email to