I've been surprised in the past few weeks as people I respect have been saying they're wary of buying switch gear from HP/3Com because they still have a bad taste in their mouths over 3Com's temporary exit from the enterprise market a decade ago. With server virtualization driving most data centers undergoing the most significant network redesign since the 20th century, eliminating a tier one vendor because previous management made a mistake 10 years ago seems short-sighted to me.
Data center network designs have been basically stable since the Cisco 6509 and comparable switches could manage enough traffic to collapse the backbone into a single pair of switches in the mid 1990s. Over the years, speeds have increased, we've increased the number of vLans and moved some routing to the edge, but the basic data network architecture has been pretty stable with data networks managing server security by port.
The rise of server virtualization is changing the whole landscape of the data center. Virtual server hosts bring expanded needs for bandwidth and shared storage. At the same time, the mobility of virtual servers from host-to-host even across data centers means we have to build flatter networks and find a way to move access control lists (ACLs) and security from port to port as VMs migrate between hosts. Add in the consolidation of storage and data networks and that Trill will let us discard the dual redundant core mode,l and the state of the art 2012 data center is going to look a lot less like the 2007 version than the 2007 version does to the 1997 one.
Getting back to 3Com, it's important to note that none of the people I've heard or read grousing about how they can't trust 3Com because they abandoned the enterprise market were actually using 3Com's CoreBuilder switches or routers. 3Com was losing ground at the high-end to Cisco, and their board made a foolish decision to get out of a money-losing business. Of course, the books that showed they were losing money on Corebuilders didn't figure in that core switch sales drives wiring closet switch sales and high end R&D trickles down through the product line. Just three years after selling their enterprise switch business to Extreme 3Com was back in the enterprise business in a joint venture with China's Huawei. By 2006 3Com bought out Huawei's share in H3C. Their flagship S12500 can have up to 512 10Gbps Ethernet ports and a pair of them can appear to the network as a single switch allowing trunking across the two switches and failover without spanning tree (which you all know by now I hate). Note this is about twice the size of a 6513 or Nexus 7000. We haven't been given any dates but HP/3Com's been clear that these switches will support DCB and FCoE at some point in the future.
The short story is that you shouldn't hold mistakes by management long gone against a company especially after that company's been acquired. HP bought 3Com for the enterprise network kit ProCurve couldn't yet deliver. Even if they never turn a profit on core switches, HP needs the products to have an alternative when the Cisco sales guys from present a Nexus/UCS solution to Proliant customers. Next time, I have to design a data center network HP/3Com will be on the short list with Cisco, Brocade/Foundry, Extreme and Juniper.
