Curious if anyone's used the 7280 and wants to share their experience?
I'm looking at it primarily for three reasons, MLAG (i.e. multi-chassis
LACP), large ARP/MAC table (256k entries) and large IPv6 neighbor table
(256k entries). For the table sizes we would like out of one pair of
switches, we'd be into the Cisco 7000 series, but that's dramatically
more expensive and we don't need much of anything else that it offers.
Looked at Brocade too, but they don't have devices that can do the multi
chassis LACP, has the huge table sizes and has a reasonable number of
10gig ports. It was possible to construct a workable solution using
VDX's for switching and CER's for routing, but that's more complex than
Arista's option if it's a usable option.
Have you looked at the brocade MLXe line?
I love the MLXe as a platform. Especially for a campus style switch.
Also Cisco really isn't expensive. Not for this type of application. A 7606-S can be purchased refurbished for like 90% off list price. The market is seriously glutted with them.
Quoting Josh Reynolds <email@example.com>:
Not sure the OP was talking about 7600s. They're mostly End-of-Life, and
not in any way suited to the OP's requirements (MLAG missing, low
density, nothing like Arista 7280)
If the Cisco refurb route is available, N7K is probably obtainable
through those channels. You'd be surprised what companies lease and then
throw back for whatever reason.
And in relation to Brocade: I'd feel very uncomfortable throwing any
*new* money at MLXe, CER or CES. Strategy for those families seems to
have fallen off of a cliff.
Could you expand further on this?
Hardware is really nice.
Backplane, buffers, just basically “pumping” bandwidth. It’s really good.
However, mlag can show some bugs when having only 1 interface in an MLAG (only 1 side) they had issues with the ifindex numbering in software.
There were OSPF configuration options missing, etc.
In short, hardware is really nice, software needs more maturing.
Nice for distribution but not for core.
These are being implemented in production on many a bank network...so yes,
they are plenty good enough. You will obviously need to test them in a lab
to make sure the features you need to implement don't have any bugs that
need to be addressed first. Overall I've had good experiences with them
though in a spine/leaf topology in major data centers.
I've also been implementing Arista switches as core devices outside of the
data center with some pretty great results, but you need to be careful to
make sure the features you need are available on the platform you want to
buy. As with Cisco (and any other vendor) there are some hardware
limitations where some features will exist on one platform, but not another.