Tanton Gibbs wrote: >> a machine with 4 cpu and 3gb of memory is definitely hard wired to crunch >> number and boost i/o performance. not accounting for machine load, there > is >> very little reason your Perl script will run slower on a powerful >> machine. > > I've found this to be not true most of the time. The larger machines > usually come with slower CPUs for one, but they also come with SANs which > are optimized for heavy load, random access IO. Since he was doing > sequential IO, the odds are that the larger computer ran slower. This has > been verified by many different companies. Small linux boxes with fast > hard > drives will almost always outperform the larger, fiber SANs. In fact, > google found this to be true and stocks its servers with traditional hard > drives instead of using a SAN storage device like what is probably > attached to your larger computer. > > Tanton
i believe your statement is bit too far reaching. i don't work for google so i don't know exactly why they would stock their file system(flat file base you mean? or rdbms?) with traditional hard drives. if they believe that way, their disk i/o will be faster and performance outweight over factor of of a secure i/o system, i think they just put themselves in a dangerous situattion. as i said before, there are a lot of reasons why a Perl script runs slower on a more powerful machine. however, i do believe a machine that the original poster mention should run faster for a number of reason: 1. disk i/o (any kind, random access i/o or sequential i/o) relies on how fast, how often and how long the r/w head has to move. a server designed for i/o will usually have high speed rotation for it's r/w head. traditional hard drive can almost never match this speed. however, depends on how often and how much distance the r/w has to travel, a traditional hard drive can actually outperform the high speed r/w head for a small file. the reason for this is that if the file is small, the high speech r/w head will not have enought time to accelerate. i believe that one of the primary reason why many uers believe that i/o is slower on a high end server compare to their low end computer is simply because the file is small. on IBM's website, i believe they have an excellent article to explain all of those. 2. data path, internal buffer, disk/controller/bus speech are also much more efficient in a high end server. again, a traditonal hard drive can almost never match the performance of this design. a script that runs 3 minutes in a pc with 6 minutes in a server is a 50% performance degrade. i will be happy to see if the op can run the test several times to confirm this. david -- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
