I have found that with two citrix Metaframe XP servers on a farm running over frame relay to be unbearably slow, has any one had issues with poor performances with link speed or even to a point when the ICA client clicks or types something they don’t even see the effects until a couple of minutes afterwards? When we use it locally on 100Mbps Cat5 cable through a switch it doesn’t appear to have any issues. Our remote users run on the two servers fine in the morning, but when the local users get on in the morning, the remote users suffers from the performance. Some information is listed for your viewing: WAN link: 128KB Frame Relay from site 1 → site 2 → site 3 10/100Mbps switches internally on all sites Windows 2000 Terminal servers on both servers Two identitical servers with Citrix Metaframe XP 3 Gb of RAM plenty of harddrive space Dual Xeon 2.4Ghz processors on both machines Clients are all Windows 2000/XP Wierd that the transactions are done on the ICA client so no reason for a slow response time due to latency of packets arriving back to the clients perhaps? Any feedback is welcome. Regards, Yoong Luu
First question is “How many users on Citrix” Your “128KB Frame Relay from site 1 → site 2 → site 3” sounds very slow. Though the times you metion are excessive, I would say that a couple of areas to check would be: 1: how are you handling DHCP and where is your BDC? are they working across the WAN, if so fix it. 2: Look for a Navision client tht is directly connecting to the server over the WAN, with such a slow connection, a direct connect will kill your bandwidth. 3: Check out the printing you have setup. Hope it helps. PS you mention that your 100Mbs Cat5 is faster than your 128k WAN, … this is normal [;)]
Will your remote folks only use Navision or do they use other apps? Another thing that simply kills bandwith is remote printing. If you have this, try to install some third party to handle printing like EOL or something like it. Again watch out for other apps like explorer. Those web animations really kill bandwith. Vx
David, DHCP is internal to the WAN, but it is replicated again at the remote sites. My PDC is at site1 and the BDC which replicates stuff from the PDC is at site2. I noticed the WAN link is rated at CIR 32K and bandwidth of 128K, does this mean that the alleged bandwidth the vendor has to provide is 32K and the burstable amount is 128K? I believe that the link is saturated as in the vendor report the usage is at 70% on all the frame relay links, the slowness only happens when the local users at site1 come to work in the morning, else the site2 and site3 find it quite bearable to work ahead of site1 two hours in front. There are no clients over the WAN that uses Navision directly, they use everything in the citrix environment, ie. office, outlook exchange mail, navision and various other programs, not to exclude printing over citrix too. Printers are managed by queues setup on local servers at each site and the printers are then replicated across the two citrix servers, they are then installed on login for individual accounts depending on what is stored as a printer queue on that client locally. Any other information would be greatly appreciated, if you have any further questions or ideas let me know. The other question I have is, the frame relay is running off a 10 Base T hub because the ethernet port from the frame relay equipment NTU has a 10Mbps Rj45 connection, I assume it’s best to swap this to a switch 10/100Mbps which will allow traffic flow direct to each port rather than broadcast to all ports on the hub. Thanks for the responses. Yoong Luu
One thing you might want to look at is your Frame Discard Rates on the router ports of the Frame Relay link ? We were using Citrix over Frame and did initially have some issues as the protocol seems (educated guess) to have problems if frames from Citrix packets are discarded. Citrix published a rather detailed technical paper on the subject some years ago where they set out the ports used and indicated what should be done on the routers to protect those packets by clearing the ‘Discard Eligible’ bit. We ran Citrix over frame at half the speeds indicated for several years. We recently switched over to our local carrier’s switched IP network which is much better performance.
You really do need to get some bandwidth, why not put in some 1M internet links, and throw away the Frame Relay.
Internet bandwith is too much expensive. With 32K bandwith (that’s what the vendor garantees and when the network is available it will grow till the 128) you can’t have all that apps running. Again, you should upgrade you’r lines. Maybe get a VPN over that frame-relay, that would boost the perfomance. I would upgrade all the CIR’s to full 128k.