On 14-02-12 06:32 PM, Mark Jenkins wrote:
Thanks for letting me know that I should never bother, sticking with unix-like systems for true multi-user knowing this!
Well, sometimes that's not an option. Do keep in mind, however, that many non-windows-logo'd apps will not run correctly under Terminal Server, so even when you need Windows apps sometimes TS isn't the solution!

Attention Skullspace donors -- forget what I said -- don't throw your money away to the licensing on Terminal Server, as it sounds so broken that the admin labour will *never* be there to put it to use.
Pretty much, yeah. It's a specialized skillset. In Win2012, it got better, but still needs special skills.

Probably easier to scale a Windows lab up in the Skullspace setting by calling for with machines with hard drives and using broadcast ghosting.
Much.  Or use VMs.

Though, at the end of day, after all the admin work to do this, it seems to me there should be a startup performance advantage with Terminal Server over the multiple-VM approach?
Yes, that's true. Multiple VMs booting simultaneously creates a crapload of IOPS. Windows Server with TS boots... one VM. Multiple logins don't cause as much I/O as booting.

You mention a 4-5 hour deploy time -- is that due to the system working hard to find shared pages between VMs? That's what it seems like we're talking about with the KVM implementation, following your kernel doc link and looking at what it links to:
http://lwn.net/Articles/306704/
http://lwn.net/Articles/330589/
No, I meant it would take 4-5hrs to click a button and have the 180-day trial editions of Windows Server reinstall automatically, and then once that's finished, have it do WinXP or Win7 re-imaged 100% automatically, regardless of whether that's in VMs or on bare metal. Getting that completely automatic setup right takes a LOT of tweaking, though.

In running RHEL previously, the KSM stuff didn't seem to have any noticeable delay. Running VMware previously, there's also no noticeable delay. Possibly I never caused enough memory pressure to trigger the same-page merging, or possibly it just has low overhead that always occurs on the fly? Dunno.

--
-Adam Thompson
 [email protected]

_______________________________________________
SkullSpace Discuss Mailing List
Help: http://www.skullspace.ca/wiki/index.php/Mailing_List#Discuss
Archive: https://groups.google.com/group/skullspace-discuss-archive/

Reply via email to