In-Depth
Enterprise Storage -- SAN April Foolery: Be Sure That Your Storage Spending Is Practical, Not a Practical Joke
We all have foibles and April Fools Day is a special occasion set aside to celebrate them. This celebration typically takes the form of a practical joke that is sometimes humorous and sometimes painful or embarrassing - depending on whether you are giving or getting.
The immediate response to the April Fools Day joke ranges from cheerful acceptance and grudging admiration for a gag well played, to hostility toward the jokester and the desire for swift revenge. Depending on the nature of the joke and the personality of the target, it’s usually a good idea to deliver the punch line in a setting where there is no access to firearms, explosives or sharp pointed things. Practical jokes should be, as the name says, practical.
Regardless of the initial impact, however, the results of a practical joke are almost always informational. Upon reflection, considerable wisdom can be gleaned from them - once the pain has passed.
Many present day SANs represent a kind of practical joke - played by storage vendors upon their customers. Those who have embraced the concept of a SAN and rushed out to become the early adopters (or advocates) of current generation technology may actually be the unwitting victims. The question is how the consumers will respond to the realization that they’ve been played.
At issue is less the technology of storage networking itself - though in many respects the technology falls far short of being ready for "prime time" business - but, rather, the business rationale for SANs.
The business value proposition of any technology explains how the technology will enable a company to do one or more of three things: Save money, reduce risk or make money by exploiting opportunities that could not be exploited without the technology. Harvard Business School, with its penchant for triangles, explains all business investments in these terms.
The original business value proposition for SANs included arguments from all three categories. SANs were supposed to save a company money by providing a mechanism for consolidating storage, sharing storage without unnecessary data replication, scaling it without disruption, operating it without consuming LAN bandwidth (saving the need for LAN upgrades) and making it more manageable.
On the subject of risk reduction, SANs were portrayed as a panacea. Tape backups could be conducted across the SAN without impacting storage accessibility - effectively eliminating the issue of backup windows. High availability could be guaranteed through realtime mirroring across the SAN fabric for companies who wanted online fail over capability. And, storage technology investments generally would be protected by the ability of the SAN to interconnect all servers with all storage platforms - including legacy gear.
Finally, vendors presented a vague assertion that SANs provided the capability for companies to realize new business objectives. This value proposition derived from the presumed ability of a SAN to enable greater access to data by applications and decision makers, so that work could be done more efficiently and markets could be exploited more quickly.
Well, the business value proposition is proving to be more of an April Fools Day joke than a real business justification. Here’s how.
On the cost savings front, SANs have proven to be more expensive to deploy than server arrays - and much more expensive than NAS devices. Everything from cabling to personnel costs is more expensive in a switched Fibre Channel SAN than in switched LAN. Despite the cost advantage claimed by Fibre Channel SAN switch-makers in the area of gigabit speed networking, gigabit LAN port pricing is beating them senseless.
Average per port costs for Fibre Channel SANs were hovering at around $1,350 in 1999, and declined to about $1,250 in 2000 according to Cahners In-Stat Group analysts. Gigabit Ethernet ports, which averaged $1,400 per port in early 1999, fell to an average $750 per port in Fall 2000. This price trend within the LAN world continues in 2001 with higher GigE switch densities yielding per port costs of $300 or less. This is both a driver of NAS today and, potentially, of storage networks based on GigE in the future.
Current generation SANs also lack the ability to share data efficiently beyond comparatively simple and homogeneous configurations. Sharing storage resources is accomplished through the mapping of physical devices to virtual volumes. Most of the current cadre of SAN "volume virtualization" schemes cost more than standalone or network-attached array products, which are delivered with virtualization schemes embedded in the microcode of the array controller. Moreover, most SAN virtualization schemes are still in their infancy, whereas the array and NAS vendors have been doing logical mapping reliably for many years.
SANs do provide a consolidation strategy, but managing that consolidated storage remains elusive. The Fibre Channel community and Storage Networking Industry Association are both promising to deliver universal Fabric Services for SAN management over the next year or two, but, for now, management is basically a kludge of disparate proprietary configuration and monitoring tools.
The bottom line on SAN-based consolidation cost savings: Current SANs look and act more like a pack rat’s storage shed than a uniform, well-managed and highly accessible storage infrastructure.
As for risk reduction, the SAN value proposition is again inconsistent with reality. Yes, SAN fabrics enable realtime mirroring, but so do traditional disk arrays and NAS devices. Backup to tape without involving LANs is increasingly doable with SANs: a bright spot.
However, the flip side of backup is restore - and current generation SANs throw more wrenches in the gears of fast restoration of data from tape than just about any technology since RAID 5. Restoring files across the SAN requires that data restore software pass its tape streams through several "filters," including a file system, one or more volume virtualization software layers and who-the-heck-knows-how-many array controllers on SAN-attached storage devices. This imposes serious overhead before data reaches target physical devices.
Backup software vendors have never been particularly keen about discussing restore timeframes, says an engineer from BakBone Software, "Most are lucky if they realize 30 percent of the rated data transfer speeds of tape drives during a restore operation." Early SAN adopters are realizing that already poor performance in tape restores declines even more when data must be filtered through all of the SAN virtualization rigamarole.
To get to the bottom line on risk reduction, just ask yourself: Can my company withstand 100-plus hours of downtime to restore five terabytes of data from tape to a SAN platform? The chill up your back is reality kicking in.
I won’t be too hard on SAN vendors with respect to the business enablement area of their value proposition. According to documented research by expert Paul Strassmann (www.strassmann.com), technology investments have never had a direct impact on - or demonstrated a discernable correlation to - the profitability of a company as described by its stock price. Why expect more from SANs?
SAN sales have helped line the coffers of many storage vendors, of course, and also provide the core enabler for the yet-to-be-realized potential of Storage Service Providers (SSPs). For those not in the know, SSPs offer a futuristic vision of a storage utility - analogous to an energy utility - that delivers exactly the right kind of storage to applications at the flick of a switch.
Still, ask most Californians about the reliability of energy utilities. This winter they endured the most cruel April Fools Day joke of them all.
SANs are evolving, to be sure. One cannot help but wonder, however, what business problem they will be attempting to solve by the time they arrive, or whether NAS and storage clusters may have already beaten them to the punch. In the tradition of April Fools Day, perhaps, it is time for us to reflect on our original business case for SANs and to see whether it still applies. Practical jokes can lead to practical thinking.
Jon William Toigo is an independent consultant and author of The Holy Grail of Data Storage Management. He can be reached at jtoigo@intnet.net.