The Storage Cocktail Is Blended, Not Shaken
February 13, 2012 Dan Burger
I’ll bet very few of you have ever heard of the George Crump strategy. Well, that’s about to change. At least for those of you who are curious what a storage analyst has to say about the IBM midrange market. Crump is president of a technology assessment company called Storage Switzerland, a company that evaluates and reports on products, the technology inside the products, and the organizations that are current and potential customers.
There are many storage strategies to consider, and Crump began our conversation last week by explaining that he believes in a blend. So you are not going to hear any bandwagon stories about solid state disk, storage area networks, or independent ASPs. And you can forget about advice on vaulting as the road to happiness.
Crump is mostly about the fundamentals. He considers the essential vitamins of a healthy storage strategy to be performance, cost, and reliability. Although he doesn’t specialize in IBM i, iSeries, or AS/400 environments, his fundamentalist approach is a pretty good fit. He says, in actuality, it’s a pretty good fit for all platforms.
People may argue, and sales reps will try to persuade, that disk performs better than tape or vice versa, but Crump’s analysis is in favor of tape. In the blended strategy he talks about, tape is the major ingredient.
“You can write data to tape faster, in the right circumstances, than you can write to disk,” he says without hesitation. “And you can read data from tape faster than you can read it from a disk system. The only negative is finding data on tape takes substantially longer than it does on disk. However, with the latest generations of LTO, the time it takes to scan an entire tape is down to about 30 to 45 seconds.”
For back up and archiving functions, this performance level is easy to live with. And in terms of reliability, Crump grades tape higher than disk. Stories of tape failures are a dime a dozen, but they are also as stale as last week’s donuts, according to Crump.
“Tape has actually become more reliable,” he swears. “With disk, the error rate increases as you jam more capacity on the platters.”
You should know that Crump is not anti-disk. Even though you know the cost comparison is another area where tape shines.
“You can buy 6 terabytes on tape for $100,” he points out while admitting there is a steep up front cost for purchasing the tape drive. “The cost advantage for tape is substantial. Even with de-duplication and compression on disk, all those things really do is soften the blow. Disk has never reached cost parity with tape.”
If you’re an IBM i advocate, you’re feeling pretty good about Crump’s assessment abilities about now. But this is where you need to start paying attention. Here’s where disk becomes important.
“My recommendation to end users is to buy as much disk as needed to store the immediate access recovery,” he says. “That amount is almost always significantly smaller than what first comes to mind. A relatively small disk system that does de-duplication can hold a month’s worth to a quarter’s worth of backups. And that is all you need in most cases. Everything else can go to tape.”
If the likelihood of a database crash that causes you to recover more than three months of data is beyond slim and is probably none, then why build a disk system that accomplishes something you’ll never use? Maybe a code error sneaks into an application causing you to roll back three months to get rid of it. But if it took you three months to figure that out, you have a problem. Even with that situation, restoring from tape should be accomplished in minutes. That’s Crump’s view.
“It’s different when you are talking about an online banking system. When it’s down, maybe a thousand people are unable to check their bank statements,” Crump says. “But that recovery isn’t going to be coming from a copy of the data set that is even a week old. You better be recovering from the most recent version you have and almost always that’s going to be on disk.”
A small business is a different story. The cost of the tape drive alone will buy two or three disks. But as the size of the business climbs into the midmarket, a tape-heavy strategy still makes a lot of sense and disk should play an augmentation role in the Crump strategy.
In IBM i dominant shops, Crump says he would most likely be recommending the purchase of some disk. In environments where the IBM i is not dominant, he’s talking them out of throwing away tape systems.
“The whole idea that tape is dead is continuously being proved wrong,” he says.