Brocade Announcements, 9/18/2013

Peter Welcher
Architect, Operations Technical Advisor

Brocade was kind enough to go over some new product / feature announcements with me. They are being formally announced this morning, Wednesday September 18, 2013. I’m going to pass along much of the press release information in condensed form, and then comment at length on the 5600 vRouter announcement. 

Here’s a summary of the info Brocade provided, with links to details:

VCS Virtual Fabric: a non-overlay alternative for multi-tenancy at scale, evolutionary, hypervisor-agnostic, API’s for cloud orchestration.

VCS AutoQoS: a VCS fabric capability that automatically recognizes and prioritizes storage traffic, greatly reducing capital and operational costs

Multitenancy Blueprint: a guide for multi-tenancy deployment, server to WAN at scale

Brocade VDX 6740 family of 10/40 GbE ToR switches:

  • VCS virtual fabric in new VDX ASIC
  • 40-160 Gig Ethernet trunks in a leaf switch
  • 32 Flex ports (what Cisco calls Unified Ports: FC/FCoE/Ethernet)
  • 24 MB “deep” buffers, dynamic buffering
  • ASIC support for OpenFlow 1.3
  • 10 Gig Ethernet, 10GBaseT support. 
  • Prices start at $15,995.

Here’s a picture of a VDX 6740:

Picture of VDX 6740

Also coming: new 100 Gig Ethernet line card for VDX 8770 chassis (available 1st half 2014).  

The Vyatta 5600 vRouter

A separate announcement covered the new Brocade Vyatta 5600 vRouter. It is positioned as a new high-performance virtual router for the Service Provider and Enterprise NFV (Network Function Virtualization) space. It is claimed to have up to 40 times the performance of competing virtual routers.

Brocade’s press release states: “The latest network-centric x86 servers deliver outstanding packet processing capabilities and are increasingly populated with 10 GbE Network Interface Cards (NICs) to accommodate data traffic growth. The Brocade Vyatta 5600 enables telcos and large service providers to harness this extremely cost-efficient networking power by leveraging software instead of deploying traditional purpose-built hardware.”

Brocade goes on to point out that this answers the NFV call to action by telcos and service providers. The new product is in limited availability now, general availability at the end of the year. It is said to complement the 5400 vRouter family. 

I’d like to thank the folks at Brocade who spent some time going over this product and answering my questions! Sometimes the pre-announcement briefings have a strong must-rush-through-reading-the-press-release-to-you-and-be-off-to-the-next-conference-call aspect.

The obvious question is “what did you do to boost the performance?” Apparently some of the improvement is due to improved Intel CPU networking capabilities. I’m told that the prior theoretical limit was 50 Gbps, Sandy Bridge processors upped that to 120 Gbps, and Ivy Bridge to about 180 Gbps. Probably not news to some of you readers, news to others.

Sidenote: I’m always glad to hear and pass along such numbers. I (and many others in networking) used to think servers struggled to put out 1 Gbps, hearing a while back about 50 Gbps was an eye-opener, and now clearly Intel chipsets can move a lot of data. Something we networking people need to get used to! Having said that, I do wish vendors of e.g. virtual firewalls such as VMware would publish some testing performance info, so we could evaluate how close their code comes to the theoretical Intel max numbers.


The other aspect to the much improved performance is due of course to better code. The prior code apparently scaled non-linearly as the number of CPU cores increased, due to cache misses. The code in this new vRouter has “vPlane technology” separating control and data planes. The data portion was re-architected to tie each vPlane to a core. Packet processing uses 100% of a core in non-interruptible fashion. This provides more linear scaling in cores, while also preserving packet ordering. 

Early lab testing is showing the 5600 as running at about 14.5 Mpps/core, or about 10 Gbps/core. (I didn’t ask, up to how many cores.) 

That means that a 16 core CPU might dedicate 2 cores to control, 14 to the data plane, for about 140 Gbps throughput. I asked about code stability. Quality of BGP implementation used to be touted by Cisco and Juniper. I was told the product is in testing by a large U.S. carrier. 

I like specific use cases. Brocade got specific. They have four use cases:

(1) BGP in a small scale core

(2) Virtual BGP route reflectors — noting that route reflectors are typically doing much more control than data plane activity. The key point cited was that you can handle millions of routes, limited only by RAM. This runs in software, freeing up costly hardware routers to move actual packets. 

(3) ACL offload — huge ACL processing. Early testing shows this operates in the 8-9 Mpps/core range. 

(4) L3 in the server, using high-speed but less costly L2 ToR switches to talk to the big core or distribution router (my terminology).

Download: Brocade Vyatta 5600 vRouter At-a-Glance:

The On-Demand Data Center Strategy from Brocade:


The vendors for Network Field Day 5 (#NFD5) paid for my travel expenses and perhaps small items, so I wish to disclose that in my blogs now. The vendors in question are: Cisco, Brocade, Juniper, Plexxi, Ruckus, and SolarWinds. I’d like to think that my blogs aren’t influenced by that. Yes, the time spent in presentations and discussion gets me and the other attendees looking at and thinking about the various vendors’ products, marketing spin, and their points of view. I intend to try to remain as objective as possible in my blogs. I’ll concede that cool technology gets my attention!

Stay tuned!

Twitter: @pjwelcher



Leave a Reply