I have read a number of posts regarding the 3750x series switches and performance issues. I have also read the comparison white papers that HP puts out (with a grain of salt, mind you) listing possible over subscription being a production characteristic of this model. As one of the harder performance issues to document, has anyone with this stack model fought this battle?
I have a pair of switch clusters
(This one has a bug that messes up the output drop counters, have to use the asic counters)
* 1 54 WS-C3750X-48P 12.2(58)SE2 C3750E-UNIVERSALK9-M
2 54 WS-C3750X-48P 12.2(58)SE2 C3750E-UNIVERSALK9-M
3 54 WS-C3750X-48P 12.2(58)SE2 C3750E-UNIVERSALK9-M
This one runs clean, no display bugs.
1 54 WS-C3750X-48P 12.2(53)SE2 C3750E-UNIVERSALK9-M
* 2 54 WS-C3750X-48P 12.2(53)SE2 C3750E-UNIVERSALK9-M
And one 2960G
* 1 24 WS-C2960G-24TC-L 12.2(35)SE5 C2960-LANBASE-M
The remote admin of this location has been doing a 550Meg .iso file transfer between two systems, moving them between the switches, using SMB/CIFS as the transfer protocol.
The 3750x stacks are on the order of 3-5 minutes slower to complete the transfers.
I am going back and working with the Admin on the site, to get a better test matrix so I have indivudual ports and what was done. But I did clear the counters all around, and the switches do run clean.. not an obvious duplex mismatch issue.
Looking to add heads and ideas on what I should be testing/looking at. I'll post what I learn.
Nick
