Interested in learning how you can achieve faster database I/O performance at 1/40th the price of a high-end SAN? Then you must read the March SQL Server Magazine article "Pare Down and Power Up," at http://www.sqlmag.com/Articles/ArticleID/49011/49011.html. Read it and you’ll learn how Rich Johnson and his colleagues at Microsoft Consulting Services leveraged the work of Dr. Jim Gray, a senior researcher and manager of Microsoft's Bay Area Research Center, to build a 24TB data warehouse capable of more than 2.2GB per second for sequential reads and 2.0GB for sequential writes.
Even if you don't have a strong background in hardware I/O, you’ll find that the 2.2GB-per-second throughput is probably as good as or better than any database throughput numbers you can achieve using a conventional SAN solution, regardless of how much money you want to spend. How did Rich do it? He used Serial ATA (SATA) drives in a direct-attached storage system.
This week’s editorial isn’t designed to fight the arguments that SAN advocates will raise regarding the availability and other manageability benefits of a SAN, and I’m certainly not going to attempt to address the issue of reliability of a SATA drive versus a SCSI drive. This week's column is simply about making you think. Sure, I’ll concede that a SAN is really nifty for all sorts of reasons. But let’s focus on the possibility of getting performance that equals or exceeds that of a SAN for 1/40th the price. Spending $50,000 rather than two million for a high-end, multiterabyte data warehouse will intrigue many people. All of a sudden, the extra benefits a SAN provides might not seem so attractive or cost-effective.
The information Rich presents in his article won't make SANs obsolete, but it might make readers realize that many customer solutions will be better served by considering a direct-attached SATA-based solution. Over the next year, I expect a tremendous amount of FUD from system vendors that have vested interests in protecting the SAN status quo and their large profit margins. I also expect a lot of innovative solution providers to find ways to deliver reliable, available, highly scalable database I/O solutions by using the ideas presented in Rich’s article. Read the article and let me know what you think.
P.S. You might want to also read the commentary from Douglas McDowell, "Are you ready for the "SQL Server Microwave"?," at http://www.sqlmag.com/Articles/ArticleID/49561/49561.html . This article describes Rich Johnson's new prototype BI Server and storage platform he's been testing.