Cisco Outlines Data Center 3.0

Cisco took the time to reiterate its data center strategy and outlined its definition of unified fabric networking, which encompasses all of the common storage networking protocols. Cisco also announced the Nexus 4000 blade switch destined for traditional blade server partners.  Long on vision, short on definitive announcements, the webcast was a non-event, but did highlight one thing: due to the convergence of data and networking along with consolidation of computing and storage, Cisco is facing, perhaps for the first time in a long time, real competition in the data center.

First, the product news. The Nexus 4000 supports 10Gb ports that can auto-sense to 1Gb ports. The Nexus 4000 runs Cisco's NX-OS with all of the reliability features support on the Nexus 5000 and 7000 switches. The ASICS in the 4000 are designed for high performance with a 1.2 micro second delay across the chip and 1.3 to 1.5 microseconds from port to port. Cisco achieves the low delay by performing all of the processing necessary to forward the FCoE frame to its next destination rather than relying on multiple chips. At the time of this writing, Cisco wouldn't announce which partners would be integrating the Nexus 4000 into their blade chassis, but did indicate announcements would be coming soon. Like Brocade, list pricing is also a secret.  The details on the Nexus 4000 are thin since the actual format of the card will be based on the requirements of the chassis vendors.

Cisco's vision for unified computing is to have a seamless fabric of inter-connectivity from server to storage, regardless of networking or storage protocol. That's great in theory, but that doesn't imply openness. Customers are still tied to the vendor-approved equipment lists. When asked the question "Storage vendors have approved lists of gear they work with.  When will that change to enable true customer choice?" Cisco answered, via email: "Storage vendor OEMs are the route to market for storage products today, and so all products have to be qualified through them.  We can't comment if or when that might change in the future."  We wouldn't expect Cisco to speak for other vendors, but as a storage networking vendor themselves, they had nothing to say. All the talk about how FCoE is the unifying technology is fine, provided you don't go off the reservation and try to use non-approved equipment. Whether FCoE will break the strangle hold storage vendors, including Cisco, have on their ecosystem is unclear.

Cisco did provide insight into their product plans, which impact their unified networking and storage  strategy.  While the Nexus 5000 is the focus new technologies, Cisco will be bringing FCoE blades to the MDS and Nexus 7000, expanding their fabric from the server to the SAN. New models of the Nexus Fabric Extender will sport 10Gb FCoE for the Nexus 5000. Cisco plans bringing the Fabric Extender to the Nexus 7000.  Additionally, Cisco is bring 8Gb Fibre Channel card for the Nexus 5000 and is developing a 16Gb Fibre Channel card for the MDS.

Here is the problem that Cisco faces in the data center: while the company's products are well respected by the majority of people purchasing networking equipment, Cisco doesn't have a track record in computing servers. Data center server incumbents like HP, IBM, and Dell, all of whom resell Cisco rack and blade switches, are reaching agreements with other vendors for switching equipment. Publicly, all the vendors are saying they will deliver what the customer demands, but thinking HP might push their own Procurve 6120G/XG or 6120XG (10Gb) for HP's c-Class Blade systems over Cisco's isn't much of  a stretch. Whether Cisco will have a harder time breaking in the server market than server vendors will have breaking into the networking market is anyone's guess.