Catalyst 3750 L3 Routing and RDP Issues

[MetroMan]

Hey all!
I have a brand new Catalyst 3750 at a client's place, we have 10 & 100mb LAN extensions to different offices that have been subnetted to different 10.0.x.x subnets...four in total.

We added one recently that directly over the LAN Extention...the thin client was pinging at 62ms over 10mb link, with the router/switch only 63ms....

Here's the prob: RDP seesions with heavy graphics are 300% slower when routed subnet vs. on the same subnet over the fiber...to me that doesn't make sense.

The Switch is has each of the first 5 ports divvied into different VLANs, and each has a GW address of 10.0.x.10, with a 0.0.0.0 address set to the mian office

http://WWW gateway...

NO routing protocals have been enabled as it is the ONLY L3 device on the network in all actuality, other than the Netgear gateway. I'll see if i cant post a running config....

Thanks!

 

[vipergg]

If this is a new setup check all interfaces for errors. Do a show interface counter errors and make sure all are clean . Verify speed and duplex settings are correct , both ends are set exactly alike , either both ends set as auto for speed and duplex or hardcode both switch and nic cards to the specific speed and duplex , cannot have one end hardcoded and the other end as auto as this will kill performance.. Things like this are usually physical layer problems . It shouldn't be 300% slower as most traffic is hardware switched at wire speeds in these new l2/3 switches including the 3750.

 

[ADB100]

Do you have QoS enabled? There are some issues with TCP traffic and default QoS settings that affect performance.

Andy

 

[MetroMan]

Hey guys, thanks for the replys. Yeah, I did have some errors on the switch ports...but I did set everything to fixed full duplexe at a mix of 10 and 100mb..and the errors continued, so I tried crossover cables, same thing....I'll run Q-Check and see how many errors come up....

does VLAN routing vs. traditional routing affect proformance at all?

Thanks!

 

[vipergg]

Remember if you hardcode the switchports you must hardcode pc and server nics to match otherwise the errors will continue . You should see no errors unless you are running half duplex for some reason which very few devices require half duplex anymore .

 

[MetroMan]

The switch ports go from the 3750 to ProCurve 24-port fully managed L2 switches that have the Duplex and Speed hard coded from the CLI and verified with a Microtest LAN tester...so I think that's not an issue....

hmmmmm....i'm still perplexed...

 

[vipergg]

If they are going to procurves ?I believe hp uses cdp like cisco , you could try doing a "show cdp neighbor detail" on each side and see if it thinks both sides are at full duplex , other than that if you are still seeing errors on the connecting links it is usually a physical problem or you are getting interference from somewhere like wire routing near flourescent lights or soemthing like that .

 

[MetroMan]

hmmmm...i'll give that a shot, and I'll try some netgear managed switches...and maybe an unmanaged switch or two....

I'll post back with results.

Thanks!