Data storage

Solid state storage: improving healthcare information systems Texas style

Neal Ekker, Texas Memory Systems

My mother managed a rural healthcare clinic for decades. As a child I ate supper around a kitchen table where the only doctor and only pharmacist within 50 miles sat and discussed the financial manoeuvres needed to keep the clinic open another month.

I fell asleep at night to the rhythmic chatter of an electric adding machine, the leading-edge information technology Mother used to do the clinic books — and perform the magic that ensured another season of polio shots, baby deliveries, and high school football physicals.

Now I live in a large Texas city where several of the most highly-respected healthcare organisations in the world are located. I had major surgery this year. My wife is currently battling cancer. Both my sisters have worked in healthcare professions their entire adult lives. To say that I have experience with and some understanding of healthcare delivery is an understatement.

I work as an executive in a long-established company that designs, builds, and sells solid state storage devices (SSDs). Every day of my working life I explain the benefits and liabilities of solid state storage to two types of audiences — people who want to use information technology (IT) to make more money or improve organisational mission success, and people who must use IT in order to do business but need it to cost less and do more.

The first group includes the financial services industry (banks and stock exchanges for example), telecommunications companies, and some government departments. They have been early adopters of solid state storage for exactly the reasons noted above — it really does help them increase profits and/or improve mission success.

Buying horsepower by the pound

A perfect example of the latter group that focuses primarily on cost is the healthcare industry — a glacially slow adopter of SSDs. Why? I believe this can be explained by a Texas-style analogy: healthcare IT managers, like most budget-constrained IT professionals and Texan ranchers, have always bought their horses by the pound.

Buying horses by the pound in the past was a successful strategy. If you’re shopping for kids’ ponies, it feels right to pay less for smaller animals and more for bigger ones. In centuries past, people needed horses to drag ploughs or pull wagons across the prairie; in both cases bigger was usually better.

But then along came the Pony Express, and long cattle drives, and calf roping at the rodeo, not to mention the on-going need for swift cavalry charges and race horses at county fairs. If it hadn’t been before, it became clear then that in some cases fast horses were valuable too, when used correctly.

The same holds true for buying and deploying enterprise storage. For 50 years the most recognized metric for making purchase decisions has been cost per capacity — $/GB.

Even today, in a presentation about SSD technology at one of the world’s premier analyst conventions, SSD cost was only considered and only compared to other solutions in terms of dollars per pound of horse flesh — dollars per gigabyte.

No matter how much faster SSDs are than mechanical spinning hard disk drives (HDDs); no matter how much less power, space, and maintenance they use; and no matter how much cost they can shave off your budget through operational efficiencies and leaner capital expenditures; the only value metric mentioned by otherwise well-informed analysts was $/GB.

If that’s the only way you know how to buy storage, or horses, then you deserve what you get — something big, slow, and only initially cheap. For those of us who have learned the hard lesson that performance has value too, let’s see what the fast horses offer.

Reliability

First, just forget about all the SSD reliability fears — that’s so last year. A quick glance at various SSD spec sheets shows endurance ranges stretching into decades.

In the past, concerns have been raised about the endurance and reliability of NAND Flash, the storage medium used in many current SSD products. Single-level cell (SLC) NAND Flash, the type used in enterprise-grade SSDs, will last for at least 100,000 write cycles.

The manufacturer’s spec actually states that less than 2% of the Flash cells should go bad after 100,000 writes. In fact, the majority of most SLC Flash cells last for many times longer. But the secret to SSD reliability resides in Flash controllers. Every enterprise Flash SSD has these miniature processors; they are where the majority of engineering innovation is occurring in the SSD industry. Thanks to innovative controller designs, it’s now truer than ever that you get what you pay for.

If reliability and data protection are key issues, as they always are in healthcare, then you can find excellent SSD products today that will meet your needs, especially when properly configured and deployed. Or you can pay less, and ride that horse too.

Cost and performance

So let’s focus in on what my experience has taught me is the key adoption criterion for budget constrained users such as those in healthcare — cost. But cost is not a single data point, it’s a journey, and the first step of the journey is to establish need.

Solid state storage has always been associated with a need for speed. In the past, your storage performance requirements were unusual or you couldn’t justify the cost of solid state storage. But the price of Flash SSDs has fallen dramatically, to the point where performance needs may almost be mainstream and SSDs will be cost effective.

Let’s look at some numbers: if you crunch data available on the Storage Performance Council’s (SPC) web site, high-performance HDD systems give you on average less than four inputs/outputs per second (IOPS) per gigabyte of storage capacity. Is this enough performance?

If you do some math using data available from the Microsoft Exchange Solution Review Program you’ll find that enterprise-level implementations of MS Exchange Server usually require around five IOPS/GB.

This means that even if you provision your email system with the highest performing hard disks available, in order to supply the application with enough storage performance to operate correctly you must over provision. On the other hand, many enterprise Flash SSDs offer well beyond 12 IOPS/GB. So you can use nearly three times less Flash to cover the same performance requirement.

With Flash prices plummeting, analysts at the Gartner Data Center Conference 2009 announced that within the year Flash SSDs would be cost equivalent to high-performance HDD under many common enterprise application workloads.

In the case of one of the most common applications on the planet — Exchange Server — Flash can be three times more expensive than 15K RPM disks and still be cost neutral for email storage.

Want to bet how soon the price of Flash will fall 3x below that of high-performance disk? Too late, that horse already ran by. Don’t believe me? Check the SPC web site again. A recent IBM all-HDD system achieved 26,000 IOPS with a total system cost of $40/GB. A comparable Flash SSD system from Texas Memory Systems listed on the same Web page of SPC-1 test results cost $58/GB, and galloped away at 250,000 IOPS.

The numbers above are impressive because we’re talking about storing all your email on SSDs, and even that’s cost competitive. But many studies suggest that the vast majority of your data is rarely accessed. No matter how far the price of Flash drops, it will never be wise to store inactive data on SSDs. Instead, a new inexpensive and highly efficient storage architecture is gaining acceptance — a small amount of SSD for the small amount of very active data, and a lot of slow, inexpensive, high-capacity SATA disks for all the rest.

How do systems configured in this way compare to storage solutions composed of only high performance disk? Studies done at our company suggest you can buy them for as little as one-seventh the initial purchase price. How’s that for a bargain?

In Texas, we learned long ago not to buy horses by the pound. A lot of horse flesh standing in the pasture eating a lot of feed doesn’t necessarily translate into a lot of useful work done. A similar concept must percolate its way into the decision processes of healthcare IT storage professionals.

The idea of evaluating and then purchasing and deploying storage in terms only of dollars per capacity is obsolete; you won’t save money this way, and you’re kidding yourself if you think you don’t need a certain minimum amount of storage performance, even for your common applications, less that application streaming high resolution medical images all around the hospital — and soon the globe.

I have confidence in the healthcare industry, because they’re my family, literally. Wise deployments of solid state storage will increase dramatically within healthcare information systems, soon. We can ride there together.

Neal Ekker, VP of channel sales, Texas Memory Systems

More articles >>