Tuesday, July 13, 2010

It may be a slow boat to converged networks

July 14, 2010 – The hype behind converged networks (or converged fabrics) is heating up, leading one to believe that we’re on the cusp of this game-changer in IT infrastructure. However, I think this is going to be a very slow transition that will take place in two distinct phases.

First, IT organizations will migrate to 10Gigabit Ethernet (10GbE), which is already well underway. As they do so, the underlying equipment (network adapters, switches) may in fact support a converged fabric (Ethernet LAN + Fibre Channel SAN via FCoE) but most users won’t take advantage of that functionality in the first phase of the migration.

For now, end users are primarily concerned about network performance, particularly in I/O-intensive environments such as virtual servers and, soon, virtual desktops (see “Storage considerations for VDI implementations,” by the Evaluator Group’s Russ Fellows and John Webster).

Faster Ethernet is priority #1, not truly converged fabrics. Remember, almost 90% of the server network ports in a data center are Ethernet.

This a central point in an InfoStor blog post from Frank Berry, CEO and senior analyst with IT Brand Pulse (see “3G C-NICs Address Mass Migration to 10GbE”).

In an end-user survey conducted by IT Brand Pulse, more than 70% of the users said they were either too busy to investigate convergence or were not planning to converge their networks (although that does leave 30% with at least a plan to do so).

According to Frank’s blog: “LAN and SAN convergence is happening today, but it’s deployment of LANs and iSCSI SANs that are making it happen. In 2010, less than 100,000 FCoE-enabled host network ports will be deployed, while over 1 million iSCSI host ports will be installed.”

Migration to 10GbE will go quickly, but the transition to a truly converged, single-wire network will go much more slowly.

One reason for this is not technology-related but, rather, cultural. More than ever, a converged network will require very close cooperation between the network and storage teams, which in some cases is like asking the Red Sox to merge with the Yankees.

Who’s in charge of product evaluation and purchasing? Who will “own” the converged network?

These factors, perhaps more than technology-related issues, could delay adoption of converged networks.

Related article:
Fabric convergence: Changing the nature of fabric attach

2 comments:

Dale Wickizer said...

Dave,

Good article. I think you hit the nail on the head with the last portion of your article. The transition to converged networks is not limited technologically. The limitation is cultural.

Organizations have allowed their network teams to be siloed into the LAN/WAN team and the SAN team (often with an unhealthy distrust for each other). I work for NetApp, the leading unified (i.e., protocol agnostic) storage provider. We run into this all the time.

Often, when talking with the SAN team, we get a lot of pushback on FCoE or any protocols that run over Ethernet. The fallback is for them to say that Ethernet is not secure, which is bunk. There is nothing inherently secure about Fibre Channel SANs other than security by obscurity. This is not about security, it is about control.

Organizations are looking to consolidate onto virtualized shared infrastructure to drive down costs and improve service levels. Having a "one wire" network strategy is key to that endeavor, if they ever hope to streamline process and reduce costs in a reasonable time frame. One cannot have a "one wire" strategy without a "one team" approach. Therefore, organizations need to make the consolidation of these teams a top priority.

吳婷婷 said...
This comment has been removed by a blog administrator.