Building Network Automation Solutions
6 week online course starting in September 2017

Interop Product Launch Craze

As expected, we’ve experienced a product launch craze just prior to Interop Las Vegas. I try to avoid marketing announcements, but the blogosphere exploded in hard-to-ignore posts ... and as always, it was great fun separating marketing fluff from reality. Here’s a grumpy take on the above-mentioned press releases.

HP

In their Interop press release, HP announced:

  • Virtual Switch 5900v (which seems to be a competitor to Cisco’s Nexus 1000V and IBM’s DVS 5000V);
  • Virtual Services Router (Comware-in-a-VM, seems to be a competitor to Cisco’s Cloud Services Router and Vyatta);
  • HP FlexFabric 12900 switch – a 10/40/100 GE switch which seems like a competitor to Arista’s 7500E. Like all core switches HP announced in the last 2 years, this one will eventually support TRILL and SPB (you got the hint, right?);
  • HP FlexFabric 11908 switch – a somewhat smaller 10/40 GE switch, yet again with OpenFlow (actually, it’s OpenFlow 1.3 ready – and my garage is Ferrari ready for a decade), TRILL and SPB support (yeah, I’ll believe the last two when I see them).

And now for a dose of reality – apart from 11908 (without TRILL, SPB, or MDC), which should be available by the end of June, HP claims everything else will ship in Q3/Q4 or even early next year. Who knows, maybe I’ll have something to include in November’s Data Center Fabric Update webinar ;)

Brocade

Brocade fares little better than HP. They announced a virtual ADX load balancer (joining F5, A10, Riverbed and almost every other load balancing vendor), DMVPN for Vyatta’s vRouter, 4x40GE modules for MLX routers, and OpenStack plugin for VCS fabric (read: glue between Quantum API and NETCONF on VCS fabric). For a longer overview, read the blog post by Pete Welcher.

I couldn’t find anything on the virtual ADX in Brocade’s documentation and there’s still no Vyatta release 6.6 documentation – when you pay proper respect to the regwall, you’re faced with release 6.5 documentation from October 2012.

It seems 4x40GE module is already in the MLX data sheet (or have they launched another 40GE module?) but I have a problem understanding what’s the real news there.

Summary: another announcement full of futures.

Arista

Arista launched a monster 100/40/10GE switch with hardware VXLAN gateway functionality and ridiculous pricing including $10K for a 100GE port (or 12x10GE ports) or under $600 for a 10GE port. Read Greg Ferro’s blog post for details. The best part of the announcement: the box is ACTUALLY SHIPPING. YAY!

Keeping in line with doing things differently than other vendors (like announcing shipping products), Arista introduced another cool EOS feature: Rapid Automated Indication of Link-Loss (RAIL) that helps kill TCP sessions to a lost IP host, drastically improving TCP convergence (which usually relies on TCP timeouts with default values in minutes). You probably know what I would do if I had to choose between TRILL, SPB, EVB and VEPA in a bright (and fuzzy) future or a set of features that solve real-life problems and challenges today ;)

Oh, and like everyone else in the industry, Arista announced OpenFlow 1.0 support, OpenFlow extensions (there’s nothing like embrace-and-extend strategy – but this time it actually makes sense), JSON API interface, and OpenStack Quantum plugin.

The only downside: it seems you have to wait for EOS release 4.12 to get most of the cool new features … too bad, I just lost some of my enthusiasm.

More information

For a real-life update of shipping (as in: being publicly documented in configuration guides available on vendors’ web sites) boxes and features, register for the Data Center Fabrics Update webinar.

4 comments:

  1. Arista Networks 4.12 EOS has been in EFT for months and will ship officially very shortly, no vapor-ware....

    ReplyDelete
  2. Assuming a conversion efficiency of 86% for the 7500E Switch Mode power supplies and a fully populated chassis with 1152 - 10G ports, the current draw should max at 48.7A @ 110VAC or 25A@220VAC, not including start-up surge or the load from supervisory and option cards. This is 18282+ BTU, or 1.56 Ton of cooling so no hiding this bad boy in a little closet. Given that it takes more than 1W of power to cool 1W of heat (otherwise perpetual motion would be a reality) the total power consumption to operate one 7500E would be greater than 50A@220VAC. Perhaps Not so cool (pun intended). But Ferarris run warm don't they.
    Unless I am way out to lunch here

    ReplyDelete
    Replies
    1. hi Anonymous,
      The system has 4 x 2900W PSUs (each are 80+ "gold "standard and operate typically at 92% efficiency) and runs at under 4W/10G port typical power.
      The system can operate grid redundant, i.e. a fully loaded chassis can run on 2 x PSUs.
      Its around ~15800 BTUs with typical power draw of <5KW for a fully loaded chassis, or 2 x 220V 16A feeds.

      Bear in mind this is front-back cooling (not side to side) so it is forseeable in hot-isle/cold-isle you'd have no problem cooling it.
      (we have many customers with upwards of 20KW of equipment/rack (servers/storage) and this fits nicely in their environment.)

      Perhaps best to compare apples/apples:
      Cisco N6004: "typical" power 2800W / maximum 3300W for up to 96x40G, or ~29.1W/40G port 'typical' and 34.375W/40G port 'maximum'.
      Arista 7508E scales to 3x that number (288x40G ports) at a power number that is less than half that on a per port basis.

      Delete
    2. Thank you Lincoln Dale for the Clarification. Keep in mind I was never inferring that some other vendor is providing an equivalent product that consumes less power.

      Delete

You don't have to log in to post a comment, but please do provide your real name/URL. Anonymous comments might get deleted.