One of the first things you notice about the recently released TPC-H benchmark from HP and Microsoft is that it’s big, REALLY big. It weighs in at over 500 pages and, since it’s the full disclosure document, contains every minute detail about the benchmarked system and environment you can imagine. For example, every hard disk in the system is fully describe – all 994 of them. Along with the hard disks, every hard disk controller (49), fiber channel controller (2), storage enclosures (99), and OS setting were fully described.
Some interesting and unusual things jump out at you right away as different from an industry standard hardware configuration, assuming you’re willing to take the time to read the specs. Here are a few examples:
1. The database was stored on raw partitions. (I’m used to seeing this on Oracle back in the day. But I haven’t seen this in production in years and, although I knew the feature existed, I’ve never seen it used on SQL Server.) The NTFS file system was used for backup files and flat files.
2. A custom function, COUNT_BIG, was used in place of COUNT.
3. SQL Server was started and ran using special startup switches.
Tomorrow, I’ll tell you more about the unusual startup switches that SQL Server ran under for its record breaking performances.