At Microsoft TechEd this year, I spent some time with Keith Hageman and Blair Parkhill of X-IO, a storage vendor that has rapidly risen in the storage ranks. X-IO's mission was to create a storage system unlike any other, which would deliver significant customer-valued, enterprise-grade advantages in ways that were previously unimaginable. Its hybrid storage technology, Hyper ISE, groups high-end HDD and SSD drives in a synergistic way to deliver unheard-of advantages. In 2011 and 2012, Windows IT Pro honored X-IO for its vision and its products' unique power, bestowing several Best of TechEd awards on the company.
This year, X-IO was showcasing a number of new technologies, including the File Storage Controller (FSC) 2200, a certified Cluster-in-a-Box appliance that brings together the very best of Windows Storage Server 2012 and the X-IO Intelligent Storage Element (ISE) family. It delivers an enterprise appliance that can serve a file-based shared storage solution with a cost-effective, highly reliable, high-performance storage service for critical applications.
The FSC 2200 covers what X-IO believes is the core enterprise application performance range: 400 to 8,000 IOPS for every 1TB of provisioned storage. This range addresses performance-sensitive data files, OLTP applications, data warehouses and business intelligence environments, and most Big Data environments. The benefits of this device include the following:
- Doesn't require a performance and capacity trade-off and is capable of delivering up to 10 times the performance of legacy storage systems at full capacity
- Offers maximum performace with reduced power and cooling requirements and a radical reduction in floor space
- Enables volumes to be configured with state-of-the-art data deduplication and compression, designed for primary data
- Easy to use and requires very little administration
With all the storage options on the market, it's easy to make mistakes in the buying process. X-IO has some advice for you. Only after you understand what your performance/capacity requirements are, empirically, are you ready to engage with storage vendors, and to test their paper-based claims against your pragmatic realities, using all the tried-and-true buying tools at your disposal—including that most powerful of tools, a pre-purchase proof-of-claim (PoC) that forces vendors to demonstrate that their proposed solution delivers both the IOPS and the capacity that your applications require. You're also ready to defend yourself, your team, and your organization against the most common storage purchasing errors, which include the following:
Buying more IOPS than the application requires
All storage vendors have SSD systems and will happily sell you these expensive systems, even if the IOPS delivered are far in excess of what your applications actually need (with a price tag far in excess of what you need to spend). Hybrid storage systems are explicitly designed to avoid this problem, as are today's high-capacity RAM servers. You can save significant amounts of money by using server memory instead of pure SSD storage and by purchasing hybrid storage, using the proper combination of solid-state and spinning media that best meets the balance of performance and capacity your application requires.
Buying an excessive amount of capacity to get the minimum amount of IOPS
Today, it's increasingly common for companies to replace racks of storage (or rows of racks of storage) with one or a few storage arrays, and improve application performance in the process. Doing much more with much less is possible today because the storage industry has, historically, vastly overprovisioned capacity in its products, in order to produce systems that delivered the IOPS required by particular classes of workload. This practice seriously overprices the cost of the storage system at time of purchase—the IOPS/capacity ratio is out of whack!—and creates a double-whammy effect over the lifetime of that system, since overprovisioned customers are forced to pay the power, cooling, space, and personnel costs associated with vastly more capacity than is required.
Buying storage based solely on first-order costs of acquisition
The first-order cost of acquisition is an important factor in choosing storage equipment. Once you've defined the number of IOPS required per unit of capacity, your purchasing goal is simple: Buy the least expensive system that meets your performance/capacity requirements. The question is: What does "least expensive" mean? Given that it's common today to employ a purchased storage system for five or six years, "least expensive" means not "least expensive to procure" but rather "least expensive to procure and operate for five years." And that means factoring in not only the cost of acquisition but also the cost of energy, floor space, personnel, warranty and service events, and so on. Any credible storage vendor can work with you on a TCO model that compares your options.
Failing to future-proof a storage purchase
Perhaps the most significant trend in storage today is the advancement of storage management features in applications, hypervisors, and OSs, leaving storage arrays to focus on traditional reliability, availability, configuration and change management, energy consumption, and density. The older model—in which storage management features resided in expensive, administratively intensive, performance-restricting controllers that sat between the server and the storage media—is already compromised in highly virtualized architectures and is generally trending down in large-scale data center architectures. Today, storage controllers should be seen as expensive remnants of yesterday's architecture, even if they are still present in your storage environment. In the next 36 months, the storage controller will disappear, into the server software stack, and the storage solution you purchase will need to operate effectively in such a "software-defined storage" environment.
Getting trapped in a buy-to-replace scheme
One of the most prevalent problems in enterprise storage purchasing is the thinking that a given storage system will last only three years before needing to be replaced. In many cases, that "three-year period" is even shorter, since it takes 6-12 months to install the vendor's product, and the vendor's sales team is trained to ask for either a new system order or expensive maintenance renewals. The facts today are straightforward:
- Today's most advanced storage systems are designed with enterprise componentry and software, built to last for at least five years of performance-rated service.
- Vendors, who have designed this level of quality into their systems, will warranty their equipment for five years, making exorbitent warranty-extension costs disappear.
- Capital savings associated with five-year-lifecycle storage products are staggering, when compared with the old-guard legacy three-year forced-replacement model.
An increasingly exciting company
We're proud that X-IO makes its home in Colorado, and in fact we'll be making a road trip soon to visit this increasingly exciting company. Check out their solutions for yourself at the X-IO website.