You know what Cisco needs? A server SAN strategy
So squawks the Vulture storage desk
Comment Cisco has no SAN/filer legacy to escape from and, in a storage world where the hyper-converged server/storage/networking system is only getting trendier, it has a major opportunity on its hands.
Consider the mainstream storage vendors – Dell, EMC, Fujitsu, HDS, HP, IBM and NetApp – their storage product line heartland is the shared, networked array. Taking advantage of the server SAN technology – where scale-out clusters of servers use their own direct-attached storage to construct a server SAN – means escaping from their shared storage legacy.
Much of the tech titans' server SAN product development and marketing energy will be devoted towards protecting their shared storage array business and not cannibalising it.
This array business is under attack from new technology designs centred on all-flash arrays (Kaminario, Pure Storage, Violin Memory) and hybrid flash/disk arrays (Nimble, Tegile, Tintri). All the mainstreamers have entries in these product areas as well.
Cisco has an entry in one: all-flash arrays, with its Invicta product. We would argue that the shared array area is not the market in which Cisco should be expending its energies. It is a mature market with strongly competing, well-established and new entry players. Also, Cisco has existing partnerships with EMC (VCE) and NetApp (Flexpod) that need protecting.
It's main strategic interest here lies in growing its UCS server business. Making these servers better suited to server SANs would support that strategic aim. Your correspondent was underwhelmed by the fourth generation UCS server announcements as these represented incremental server developments and modest market extensions towards remote and branch office environments and the cloud.
Cisco needs to add server SAN hardware and software to its servers to make them the most attractive server base for server SANs and enabling it to develop its own hyper-converged server/storage/networking line.
What do we have in mind?
The hardware additions should include either or both FlashDIMMs and PCIe flash cards, together with an external PCI fabric or Infiniband for use as the scale-out server interconnect.
The software needs to be able to use a server's flash as both multi-server cache and storage memory. Cisco should buy either Atlantis, Nutanix or Simplivity to gain this software – that would be much faster than developing its own software. It already, by the way, has a reselling deal with Simplivity.
None of the mainstream server vendors – meaning Dell, HP and IBM – have an aggressive hyper-converged systems strategy, partially because of the need to protect their existing storage array product businesses.
Herein lies Cisco's opportunity. Its server competitors* are strategically weakened by their shared array business and, apart from its yet-to-be established Invicta array business, Cisco has no networked storage legacy to escape from.
A door is wide open in front of Cisco and it could march right through it while its competitors hobble along behind. So Cisco, will you heed the Vulture's strategic squawks? Or sniff in derision and ask rhetorically: "What do a bunch of hacks who've never run a business know?" ®
* Dell has an OEM deal with Nutanix to supply a hyper-converged system, and EMC is developing its ScaleIO-based serve SAN offering, placing these companies at the head of the pack competing with Cisco's UCS servers from a hyper-converged system, server SAN basis.