Inside Cisco's Hotel California
A unified fabric should include storage
Comment If Cisco is to introduce its unified compute system, or California project, on March 16, what will the storage look like?
The overall Cisco scheme is that the virtualised network is the platform. A Cisco unified fabric switch, like the Nexus 5000, will be Shelob, the giant spider in Cisco's 'one network to bind them' concept, sitting at the centre of a web of interconnections. Cisco blade servers running virtual machines - VMware ones for now and Hyper-V ones tomorrow - will link via 10Gig Ethernet to this switch and talk through it to storage, both block and file, and to the outside world. LAN, WAN and cloud.
The likelihood is that storage arrays will be linked to the Nexus in this scheme by Ethernet and not Fibre Channel. MDS9000 Fibre Channel fabric switches won't appear in this setup because including Fibre Channel will be the very antithesis of a unified fabric. We can infer then that storage arrays will link to the Nexus by Ethernet only.
This is okay for file access as both NFS and CIFS work that way. But block access means either iSCSI or Fibre Channel over Ethernet. Like many others, Cisco has been pushing Fibre Channel over Ethernet (FCoE) as the enterprise block access method with iSCSI as the SME, non-data centre complement to this.
So, the odds are that Cisco will layout an FCoE block storage access route. It implies that only FCoE-interface arrays will qualify for a block access role. NetApp has a native FCoE interface capability for its storage arrays and can also provision both block and file storage from its FAS products.
EMC could add an FCoE interface to its Clariion CX4 arrays using its UltraFlex I/O technology. Clariions are block storage arrays. The latest Celerra can provide both block and file-access and also uses the UltraFlex I/O architecture, meaning that EMC could readily add an FCoE access door to it.
It will be in Cisco's interest to promote its unified computing scheme to work with independent storage vendors like EMC and NetApp. The presupposition is that storage-and-server vendors like Dell, HP, IBM, and Sun will not be as co-operative in linking their storage to Cisco's Nexus switch as it runs Cisco's server blades, competing with their own server blades.
Cisco sees its unified compute system as being completely applicable to cloud computing, to both internal and external cloud suppliers with ultimately an interchange of virtualised resources between private and public clouds. This echoes the EMC and VMware vision to a high degree.
We can further expect then that the storage arrays admitted to Shelob's lair in Cisco land will have vStorage API front ends and that VMware's vCloud ideas will feature broadly too, as will EMC's Atmos cloud storage products. If this FCoE-only block storage access route comes to pass in the Cisco announcement then it would be logical for EMC to simultaneously announce an FCoE interface capability for Atmos, Clariion, and Celerra. ®