I dont believe so. We disabled all of the SNP features on our Citrix servers awhile ago. I had the same thought at the client level, but disabling TCP Offload didn't change the behavior.
- Sean On Aug 4, 2010, at 12:54 PM, "Webster" <[email protected]> wrote: > The hanging packets isn’t the TCP Chimney stuff is it? > > > > > > Webster > > > > From: Sean Martin [mailto:[email protected]] > Subject: Re: Tweaking Performance - Citrix Servers > > > > I can only assume so. I don't directly manage our Citrix > environment, I'm just offering assistance wherever possible. I know > we've had Citrix Consulting onsite to perform a specific health > check, we've had our networking team continually monitoring > everything possible (we have a CCIE on-staff, if that means anything > anymore), and our own Citrix admin team has been digging into all > areas of performance analysis, using the usual tools as well as > specific monitoring solutions we have such as EGManager. > > > > The initial response from the consultants is to plan migration to > XenApp 6 as soon as possible. We're also a little behind as we're > running CPS 4.5 hotfix rollup pack 2, so it's suggested we upgrade > to rollup pack 6(?). They confirmed the behavior we're seeing is due > to the use of session reliability (applictions hanging rather than > the sessions being disconnected), so they've indicated we need to > dig deeper into our network investigation as the problems seem to > indicate a loss of communication. However, the recent article I > linked in my original e-mail seems to indicate that the problems > we're seeing can often be incorrectly attributed to network > problems. So far we've been unable to identify any network > communication issues. The problems effect users on our local LAN as > well as branch offices over WAN connections. > > > > We have one of our managers using applications published off of a > specific server so that we can focus our troubleshooting efforts. We > have batch scripts published from this server that runs a debugging > tool given to us by a Microsoft developer. There's 3 scripts > configured for Outlook, Word, Excel. If the manager experiences a > significant application hang, he executes the respective script > which debugs the corresponding process for his session. Analysis, by > Microsoft, of several of those outputs has turned up nothing. > However, our Network Engineer has a sniffer setup to monitor all > communication from the manager's PC to the Citrix server. During the > perceived application hangs, our network engineer has identified > that both the NIC on the client PC and on the Citrix server appear > to be "hanging on to packets" for up to several hundred seconds > before releasing them. This was just recently identified so I don't > think we've dug into possible causes of that behavior. > > > > - Sean > > On Wed, Aug 4, 2010 at 11:51 AM, Webster <[email protected]> > wrote: > > Are you monitoring the standard TS/RDS/Ctx perfmon counters? > > > > > > Webster > > > > From: Sean Martin [mailto:[email protected]] > Subject: Re: Tweaking Performance - Citrix Servers > > > > Thanks for the response. > > > > Yeah, we've noticed a big difference between our older servers > without caching and newer ones with it. Unfortunately we can't > dedicate the time to our hardware refresh project that would > eliminate those older servers. We're in the process of > redistributing the newer hardware to the application silo running > the Office suite, since that's where most of our performance > complaints stem from. > > > > Even with caching enabled, we still receive complaints of > application hangs when users are connected to one of the newer > servers. We've spent the last several months investigating the issue > and trying to resolve it for good. We've made progress, we just > haven't eliminated the occurrences. > > > > If you're referring to the lack of fault tolerance with a RAID0 > setup, then we understand the cons very well. We have enough servers > in our farm that we can lose a handful of servers due to disk > failure and still support production (albeit performance may suffer > a little more). We have a fairly decent stock of spare parts as well. > > > > - Sean > > On Wed, Aug 4, 2010 at 10:40 AM, Webster <[email protected]> > wrote: > > I have seen issues with the blade RAID Controllers NOT having a > battery backup cache controller. That will seriously impact > performance. > > > > I don’t think RAID0 will negatively affect the page file. Just reme > mber the caveats of using RAID0. > > > > Before you go making changes, create a baseline to measure against. > > > > > > Webster > > Citrix Technology Professional > > > > From: Sean Martin [mailto:[email protected]] > > Subject: Tweaking Performance - Citrix Servers > > We use Blade servers in our Presentation Server 4.5 environment > (approximatley 100 server split between 8 application silos). The > hardware ranges from Dell PowerEdge 1955s to newer Dell M610s. We've > been toying with ways to increase performance of the servers and > recognize that disk performance on our Blades is usually the > bottleneck. In our standard configuration, the two drives (1955s = > 73GB/10k, M610s = 146GB/15K) are mirrored. We've configured a couple > of servers with no RAID (two independant disks) and tried balancing > resources across the two (page file, temp directories, spooler > directory, etc). While this did yield some positive results, it > wasn't that noticeable in the grand scheme of things. > > I just read the following article from Citrix, apparently published > on 7/21/2010. The issues referred to are currently what plague our > environment the most. One of the recommendations is to try a RAID 0 > configuration in Blade systems with only two drives. We've > historically written this off as a solution because of the > recommendation of not striping a page file. Do you think the > potential fragmentation issues would be considered an acceptable > risk when compared to the potential disk performance improvements? > We do have Diskkeeper installed on all of our servers which is > scheduled to run during the evening hours. I do believe it addresses > page file fragmentation but it has been awhile since I've looked > into the capabilities of the product. > > > > I should note that our M610s with 256MB cache (caching enabled) and > RAID1 do seem to perform much better than our older servers. > > > > http://support.citrix.com/article/CTX125882 > > > > ~ Finally, powerful endpoint security that ISN'T a resource hog! ~ ~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~
