Disclaimer
The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.
Liability Disclaimer
In no event shall Author be liable for any damages whatsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.
Posting
Yes I have heard they are not good at high PPS applications, but saying that... SoftLayer seem to use 3750X with much success. I surpose it depends on the traffic levels really.
It's not so much that can't handle high PPS, but often by default, they don't do well when ports are oversubscribed due to less buffer resources than say perhaps a 4900 series. Of course if you set bounds on oversubscription ratios, by design, this problem is less likely to arise.
David's post raises an interesting point about a 3750-X being particularly sensitive to microbursts because their ports share a common buffer pool. I've read how they can easily run out of port buffers because there's insufficient buffer resources to provide every port a large dedicated buffer pool. This can be mitigated by yielding port dedicated buffers back to the common pool so a busy port can drawn from it. Doing this, I could see, might lead to a microburst issues, i.e. the port can't allocate buffers from the common pool fast enough. Again, though, this might be avoided by design or perhaps other mitigation techniques (for example, I wonder if "shut" inactive ports still reserve buffer resources, if not fewer active ports on the 3750-X might allow those buffers to be reserved by other active ports).
Thanks for the guides, I have already implemented those, so besides that - is there anything else that should be considered?
Those guides should cover about 99% for hardening the device. What always helps my thinking is keeping in mind the difference between traffic transiting a device vs. traffic directed to that device. I start with the premise all traffic is allowed to transit the device but no traffic is allowed to the device and then amend as necessary.
