Many vendors and users have been focusing on how to improve storage capacity, as data has seen enormous growth rates in the past few years. But how does storage performance measure up? What options do vendors have for improving and assessing storage performance, and are users taking the technology seriously?
In his latest Storage magazine column, Jeff Boles provides us with a storage performance analysis. Small improvements have been made to storage performance solutions, and Boles believes that can lead to even better improvements. At the same time, users need to be aware of the caveats that sometimes come with these solutions.
As we approach the second half of 2011, it looks like we're going to see a much different data storage industry for the rest of the year. That's a good thing because 2010 was a minefield for strategic planning -- a roller coaster of excitement with the unpredictable future that goes with it. Game-changing technologies made their real first forays into the market, including cloud storage service providers, cloud on-ramps, cloud backup, cloud disaster recovery (DR) and others. In a recent emerging market forecast, Taneja Group identified storage products associated with the cloud as growing into a $10 billion market by 2014. As we see it, we're out of the gate with new product innovations and now engaged in a steady march.
In 2010, there was turmoil among the vendor ranks like we've never seen, with major vendor strategy shifts, a number of small vendor failures and mind-boggling acquisitions. Whew! Surely, all that vendor turmoil must be tapering off. We need a breather and we need to look at what's coming.
There's definitely one thing that needs more attention: storage performance. For too long we've been fixated on capacity, and that preoccupation has sapped some of the innovation energy that should have been directed toward solving storage performance problems. And because of that neglect, many of us are finding we need better performance than we can get from today's standard approaches. It seems most of the activity in 2008 and 2009 was centered on capacity, and in 2010 that capacity discussion shifted to the cloud. Lately, it seems performance only comes up when the subject is solid-state storage, an alternative ill-suited to many of the storage systems most of us have in our data centers. That might be the future, but performance needs to be addressed now.
But I'm beginning to see performance solutions that have some real promise. Discussions around solid-state technology have made us acutely aware of the massive limitations of many traditional storage systems. And that seems to have inspired some innovators to do something entirely different.
Despite what some vendors are peddling, it won't be the next generation of ridiculously complex, application-specific stacks of hardware that will speed up your database or your number crunching. And don't expect bigger and cheaper appliances to deliver on that promise either. Those and similar solutions have been anathema for strategic planning. How are you supposed to build a storage performance strategy from a proliferation of one-trick ponies that can't adapt to changing demands or the massive shared access required within today's consolidating data centers?
The cost of delivering scalable performance
My requirements for doing better aren't that stringent. Fundamentally, a solution should give you real performance and promise to scale well beyond what you need today. Simultaneously, it shouldn't look entirely different than your existing storage. It should still store data in a permanent, reliable way and be just as serviceable. We're seeing vendors like Alacritech, Avere, Kaminario and others think hard about how to deliver performance that fits into today's storage infrastructure without a massive integration effort, and with the right capabilities to deliver future scalable performance without compromises in storage capabilities.
One result of these emerging, improved architectures for storage performance is a dramatic change in what we pay for I/O -- when assessed on the basis of dollars per I/O, such systems have massively better costs than rotational disk. But those figures only tell part of the story. Bad architectures come at a price way beyond the difference in upfront costs. The upfront costs of any performance solution with a bad architecture may well be compounded by the operational costs incurred in the course of coping with and offsetting the bad architecture, such as:
- The cost of integration. Integrating solutions that address a subset of data can require enormous effort.
- The cost of data protection. Few solutions are built for primary long-term storage, so the cost of data protection can increase significantly. Worse yet are the considerations for DR.
- The cost of replacement. Solutions built without long service life as a key design criterion may reach a premature end of life. The cost here isn't just about replacement, but dealing with the effects of potential failure, often with strategies such as maintaining idle spare equipment.
- Cost of service. Some products aren't built for in-place service, so they may require specialized expertise for service or disruptive total replacements if a failure occurs.
- Cost of scale. A product that's attractively priced for a single app may easily be a mismatch for other demands. This can result in apps requiring multiple storage systems, capacity underutilization or compromising on performance.
- Cost of management. Inadequate scaling can lead to sprawl and result in having to deploy multiple devices to satisfy different application needs. The cost of managing isolated systems can become enormous.
And the bigger things get, the more costs add up. I'm not suggesting your primary storage vendor doesn't have answers that address these issues, as some are thinking pretty innovatively about how to extend controller capabilities. But you need to make sure you have the right primary data storage vendor if you're buying into their answer for performance. For some, their current investments or intended new investments may never go far enough. There may well be a significant deficiency in both the resulting value you're getting per dollar and in the competitive business capabilities of your resulting IT systems.
It may be a year of only incremental improvement, but when it comes to performance, even small improvements can transform how we do storage. With a few innovators on the horizon, it looks like the rest of 2011 could see even greater strides.
BIO: Jeff Boles is a senior analyst at Taneja Group. He can be reached at firstname.lastname@example.org.
This article was previously published in Storage magazine.