Cisco 3750-X vs Cisco 3750G

[halogensoftware]

We are looking to replace our pair of 3750G switches (Stacked) by the new 3750-X. The problem that

we are having is that we are having packets drops on our 1Gb trunk to our Firewall/Router. We had a

consultant last year that mentioned that the 3750G had a limitation on the dataplane, I can't remember

exactly waht he said, but was something like when a packet arrived on port it would be put in the stack ring,

until the destination port removed that packet from the ring, and only one packet could be in the ring at a time.

he then said that the 3750-X would not have this limitation, and was way much faster.

Could somone give me some insight on the differences of the two switches ? I also wonder if The X model have

burst buffer like the 4948 switches ?

Thanks in advance,

Alex.

[Wilfred Todd Smith]

It would be interesting to learn about the stack limitations and logic.

What troubleshooting have you done that leads you to believe it is a stack ring issue?

[rtjensen4]

I'm not an expert on this but here's how I understand it:

The 3750G uses stackwise and 3750X uses stackwise+. With Stackwise, EVERY packet has to go across the stack ring even if the destination port is on the local switch. With Stackwise+, the switches can do local switching.

Also, if you mix STackwise and Stackwise+ capable switches in the same stack, the Stackwise+ switches will still be able to switch locally and the Stackwise switches must continue to operate the same way.

Here's a link:

http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps5023/prod_white_paper09186a00801b096a.html

HTH

[halogensoftware]

Thanks for pointing out that document, I believe that this is
what the consultant said about the stack difference between these two models.

I was thinking about the problem that we're having w/ the drop packets and I was also wondering

if This Stackwise behaviour, source stripping,  could be the cause of our drop packets. Let me elaborate

my question:

Our 3750 is not doing any L3 routing, all the routing is being done by our Router/Firewall connected

to the stack via a 1Gb Trunk Link active-passive (one link to each member of the stack), all other switches

are Cross-Stack EtherChannel Connected to the Core (2Gb Links). Since all traffic is being routed on the Router/Firewall,

all traffic has to transverse the 1Gb Trunk, thus has to be put in the Stack ring. This is an old design, and I'm sure

you guys can already see the bottleneck, but my question is, why I only see drop packets in the active trunk port ? Shouldn't

it be on the source/originating port ? I would imagine that when all etherchannels are busy, full traffic, and all that traffic must be

sent to the trunk port to be routed, the buffer on the stack ingress would not handle this much traffic and start to

drop, and the originating port would see the drop packet ? Please correct me because I must be wrong :-)

Thanks in advance !