An Unofficial Look at the Current State of the Storage Industry from the Fall SNW 2012

At this fall's Storage Networking World (SNW) 2012 I purposely kept my briefings and attendance at the keynote events to a minimum so I could catch up with a number of industry insiders and engineers in the field to get an unofficial look at the current state of the storage industry that rarely gets publicly reported. What I learned was that many commonly held perceptions about the storage industry may be in fact on shaky ground and should prompt users to closely examine their assumptions about how they should proceed with their current and future storage plans.

In preparing this blog entry, it should be noted that none of this information should be considered "official" nor is any of it necessarily verifiable. However it was all shared with me by individuals who I have known for years, generally view as reliable sources for information and would have the background, contacts and expertise to have an informed opinion based on their intimate knowledge of the storage industry. Here were the three most noteworthy pieces of information shared with me.

• Cisco is pushing the bounds of its allegiance to the VCE alliance.  It was in early 2010 that with great fanfare VMware, Cisco and EMC announced the creation of the VCE Company and its associated vBlock architecture to simplify the deployment and scaling of virtualized data centers. At that time and even through today, it is viewed as a risk-averse way for enterprises to effectively build and scale up their virtualized infrastructure.

The problem that has emerged is that flash-based storage arrays are very disruptive to the EMC component of this VCE alliance. The speed, smaller footprint and extreme energy efficiency of these flash-based storage arrays are apparently prompting many enterprises to question the logic of bringing in a big, disk-based storage array when a smaller flash-based storage array can more efficiently (from a energy saving and footprint perspective) deliver the performance and scalability that enterprises need.

This is why Cisco is apparently starting to pull back (not necessarily pull out) of the VCE alliance and more strongly partner with emerging flash memory and solid state disk providers with the acquisition of one of these providers allegedly in the works. Nimble Storage has apparently caught Cisco's eye as a potential acquisition target. Nimble's announcement yesterday that it has provided a reference architecture in partnership with Cisco and VMware would serve to further confirm this rumor circulating at SNW that something may be in the works between Cisco and Nimble.

• Proceed with caution if native ZFS is the storage system OS. Installing open source ZFS on off-the-shelf hardware has in recent years emerged as an appealing alternative to create a high performance storage system at a fraction of the cost of using storage systems that have proprietary vendor storage operating systems installed on them. ZFS (or variations of it) has already been adopted by storage suppliers such as Aberdeen, IceWEB and Nexenta with others using some variation of it.

What users have to watch out for is the enterprise-readiness of the features natively found in ZFS. Its native CIFS and deduplication features were two that were specifically called out by the individual with which I spoke. This person noted that the CIFS protocol has "some issues" in its implementation in ZFS (though he could not provide me with any specifics) and the native deduplication feature when enabled becomes a real drag on storage system performance.

One should also not assume that its performance works equally well across all types of applications. ZFS has had a strong legacy of working with Oracle but in the case of Microsoft apps - not so much. As a result performance may do just fine when running Oracle on a ZFS-based storage array but not so well when Microsoft SQL Server or Exchange is hosted on it.

Finally, one needs to beware of the difficulty in supporting ZFS implementations in production environments. For instance, if one uses SuperMicro servers running ZFS OS on it, end users need to quantify who will be configuring, installing and supporting this storage configuration. If it is them and/or their value added reseller and something goes wrong, the finger pointing starts pretty quickly with the system administrator often the individual left holding the bag in terms of reconfiguring the storage array.

In short, ZFS-based storage arrays may be suitable for production business use cases but verify that the storage provider has addressed ZFS's native shortcomings in the CIFS protocol and deduplication features it ships and offers a viable support plan where the storage provider assumes ownership of support issues when and if they should occur.

• Converged infrastructures are re-introducing inflexibility back into the data center. On the surface, converged infrastructures such as what are being offered by Dell, HP and others (to include VCE mentioned earlier) sound great as they make servers, networking and storage simple to install and configure initially and then simple to manage long term. The problem with this converged architecture approach is that from a technical perspective both server and storage performance and capacity grow at different rates over time.

Further, every company has its own quirks in terms of its own internal server and storage growth rates. In some environments, server performance grows faster exceeding the upper limits of the converged infrastructure. In others, storage capacity growth outstrips what the converged infrastructure can again provide.

Companies get the simplicity they initially want from converged infrastructures. However they over time lose the flexibility that today's independently configurable (though arguably more complex) servers and storage provide. Additionally, they may end up paying a premium to have the vendor deliver to them this converged infrastructure solution.