This syndicated post originally appeared at Zeus Kerravala's blog.

I attended Brocade’s annual “Tech Day” at its headquarters in San Jose on yesterday. This event is meant for Brocade to share its vision of where the industry is going and what product plans it has to support the vision. I’ve been to a few tech days now, and this was one of the better ones in recent history. Here are the main highlights.

The company defined its vision of what the terms Ethernet Fabric and Software Defined Networks mean, and then explained the interplay between the two. In Brocade’s terms, an Ethernet Fabric is defined by:

  • A network that is optimized for server virtualization and cloud architectures
  • An efficient network that has higher throughput and lower latency
  • The ability to scale out as needed to improve the flexibility and agility of the network
  • Automation to simplify network operations

The characteristics of a software-defined network (SDN) are:

  • Cloud orchestration using open stack
  • Programmatic control over the network
  • Virtualization of network resources

The SDN is used to solve several key data center challenges, such as creating multi-tenant networks, simplification of network operations and then bringing a level of IT agility to the network that other areas of IT, such as server operations, have today. The Ethernet Fabric would be the underlying supporting network that would enable the actual operationalizing of the SDN.

I’m in agreement with most of what Brocade articulated for both Fabrics and vision, although I disagree with the notion that cloud orchestration must be open-stack. I think cloud-stack is coming on strong. I also think another component of both fabrics and SDNs, which Brocade failed to address at the event, was that both need to be open and standard-based. The data center is moving to a war of “stacks,” and if you’re Cisco, IBM, VMware and Oracle, you can create a closed stack and get away with it (although most of the above are building some level of openness), but if you’re only part of the stack, the key is to work with as many other ecosystem vendors as possible. I think Brocade believes this but just didn’t mention it in the talk track (or maybe it did and I was too busy tweeting!).

Brocade launched a few new products to support the vision as well, including the long-awaited, widely rumored VDX 8770 Data Center Fabric Switch (formerly known as code name Mercury).

The 8770 is a beast of a switch and can be thought of as a massive backbone for a large, fast fabric. There’s a 4-slot and 8-slot version of it with a capacity of 4 TBs per slot, meaning customers should have this product in place for many years to come. With a capacity of 4 TBs per slot, the 8770 could support a 40×100 Gig-E line card. The VDX is powered by custom-built ASICs that give the product ultra low latency and enables it to scale up to 8,000 server-facing ports and 384,000 virtual machines. The 8770 is a nice complement to the 6700 series line that Brocade has in place today. Although meant for small scale, the 6700 series seems to have served its purpose. Brocade started small, enabling customers to try it at a lower price point, yielding an install base of over 700 production customers running fabrics.

Juxtapose this with Juniper’s “start big” strategy, and the company has a customer base generously estimated around 200, many of whom are not actually running a fabric. Latency, number of server-facing ports, number of VMs supported – these are things network managers need to evaluate data center fabrics on, and forget about legacy competitive criteria like cost per port, backplane capacity, etc.

I think the launch of products like the 8770 built on custom silicon is a good proof point that SDNs do not commoditize the network. I’ve heard the analogy that SDNs will do to the network what virtualization did to servers. I think this analogy is actually accurate but misunderstood. Did VMware commoditize legacy servers? Sure, it commoditized the legacy stuff that was already in existence. However, it also opened the door for a company like Cisco to ask how it can build a better server that allows for greater return on virtualization dollars spent. Hence the incredible growth of Cisco’s UCS server, which is a premium-price server.

Similarly, SDNs open the door for all network vendors to look at the problem and try and build a better switch capable of enabling all of the things that are possible with SDNs. That’s the concept behind the 8770. Market transitions create opportunities and the data center transition now is towards a more efficient, agile, simpler-to-operate data center. This doesn’t mean communization; it means innovation, and that’s what we saw at this year’s Tech Day.

The following two tabs change content below.

Zeus Kerravala

Zeus Kerravala is the founder and principal analyst with ZK Research. Kerravala provides a mix of tactical advice to help his clients in the current business climate and long term strategic advice.
Share This Post:
No Interactions

Be the first to comment!

Post a Comment:

You must be signed in to post a comment.

Insight and Influence Through Social Media
ZK Research: Home
RSS Feed
ZK Research is proudly powered by WordPress | Entries (RSS) | Comments (RSS) | Custom Theme by The Website Taylor