Network Computing is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

How Server Vendors Can Compete In Storage

8 Great Cloud Storage Services
8 Great Cloud Storage Services

(click image for larger view and for slideshow)

The last round of earning announcements from the major server vendors have not been spectacular. One of their key problems is that they are losing shares in their storage offerings to storage-focused companies.

Ironically, thanks to software-defined storage and server virtualization, server companies have a tremendous opportunity to dominate the storage market by making storage systems a thing of the past.

Server vendors such as Oracle, HP, Dell and IBM should embark on a massive converged infrastructure play. They should make this move courageously, dramatically -- and overnight. An over-thought, committee-based "let's not offend anyone" won't get the job done.

"The Server Is The Storage"

Sun, prior to its acquisition by Oracle, had a tag line that read, "The Network is the Computer." My suggestion to server vendors would be to change that tagline to "The Server is the Storage" and design servers that can support massively dense virtual machines per server while providing complete data services.

[ Software-defined storage, a way of providing a single point of data services on any disk hardware, sounds appealing, but is it for you? Read Is Software-Defined Storage Right For You? ]

These architectures merge the concept of the current virtual infrastructure with standalone scale-out storage. Scale-out storage is a cluster of nodes built from 1U or 2U servers. If these nodes are filled with hard disks or solid-state drives, then the capacity is aggregated across those nodes. All the big server vendors I mentioned have systems in this category; they just need software that would run on these servers under the control of a hypervisor so that other virtual machines could share the compute resources.

The good news for these server vendors is that this software is readily available from about half a dozen companies. They could OEM this software or outright purchase it for far less than other expenditures they have made. As I discussed in "Cloud Defined Storage," these software solutions are typically focused on the managed service provider or cloud service provider markets. But with the backing of a major server vendor they could easily attract the interest of data centers of all sizes.

What should be especially appealing to this strategy is that a single solution, with some additional development investment, should be able to scale from the smallest of businesses to enterprises to the hyper-scale data centers. As I discussed in "What is a Hyperscale Data Center?", these data centers are experiencing significant storage challenges meeting conflicting demands on capacity and performance while still offering competitive subscription rates. Not surprisingly, traditional data centers face a similar challenge.

After the technology is acquired or OEM'ed, these vendors could start working on modularizing their legacy storage assets and pulling specific features into their "server is the storage" strategy. Each owns some interesting capabilities such as deduplication and data tiering; adding those features to the now-virtualized storage software would give these vendors a way to differentiate themselves.

The Impact

Imagine if every server you bought already included all the storage capacity and performance you needed? With 2-TB 2.5" SSDs now becoming available, the basic 2U server could deliver 12 TB or more of affordable, high-performance storage. Of course, even higher capacity servers with hard disks could be added. The vendors would need to develop ways to manage the mixed server types.

Also imagine if that same server already had all the software needed to join an existing cluster of servers that were managing various virtualized workloads. Its storage capacity would be automatically aggregated to the existing cluster and its compute resources available to accept virtualized application workloads. There would be no need for a separate storage infrastructure. This would make things very difficult for storage system vendors.

The good news for the storage system vendors is that the server vendors probably won't do this. First, they all have acquired multiple existing storage companies that they need to justify spending money on. Second, it would take an overnight, all-in commitment that would be more along the lines of a startup than an established vendor. Unfortunately, it would probably take an established vendor to make the strategy successful.

If you don't want to wait for the server vendors to adopt this strategy, then -- as I covered in "Building The SAN-Less Data Center" -- there are a few companies delivering a turnkey version of this strategy today. You also, of course, have the option of designing the strategy yourself, leveraging available servers, software and storage media -- but you get to play manufacturer and inherit all the responsibility that entails.