Looking for recommendation on 10G Ethernet switch

Maxim Khitrov max at mxcrypt.com
Fri Nov 2 23:30:04 UTC 2012


On Fri, Nov 2, 2012 at 4:10 PM, Jeff Wheeler <jsw at inconcepts.biz> wrote:
> On Fri, Nov 2, 2012 at 11:13 AM, Eric Germann <egermann at limanews.com> wrote:
>> I'm looking for a recommendation on a smallish 10G Ethernet switch for a
>> small virtualization/SAN implementation (4-5 hosts, 2 SAN boxes) over
>> iSCSI with some legacy boxes on GigE.
>
>> 1Gbps.   Assessing whether it is better to go 10G now vs. multi-pathing
>> with quad GigE cards.  Trying to find the best solution for > 1G on a
>> trunk and < $50K per box.
>
> Certainly the days of doing NxGE to servers should be behind us.
> There are many good 10GE switch offerings.  The Juniper QFX and Arista
> (insert one of three good product lines here) have been excellent for
> us.  We like them for different reasons -- Arista is quite good if you
> want to integrate with a provisioning system; QFX is our choice when
> most provisioning is done manually.  Both are way under $50k per box.
>
> The biggest difference between the TOR-style switches and chassis
> offerings, aside from the obvious, is buffers.  All the TOR-type 10G
> switches have really small buffers and that can be a performance issue
> for iSCSI when utilization is high.  Most of the chassis-type switches
> have very generous buffers so you do not run into problems with
> micro-bursting, etc.  The vendors will all tell you about lossless
> ethernet, flow control, etc. and that crap sounds great on paper.  Try
> making it actually work.  You'll want those days of your life back. :)

I'm curious if anyone here has experience with the Extreme Summit X650
or X670 switches? I'm also currently looking for 10GBASE-T switch for
my first SAN, but one with 48 ports. The price for X670V-48T seems
reasonable, but I have no experience with 10 GbE over copper, so it's
difficult to figure out what criteria to use (other than price) when
comparing Extreme, Arista, Dell, and others. Any tips?

- Max




More information about the NANOG mailing list