Re: What determines system Page Size?

Bryan Hughes wrote: > > Hey Mickey! > > Have you had a chance to check out those Circle K's yet? Well, seems like I > actually have the answer to this. Believe it or not, our friends at tech support > answered this. Basically, the OS's page size is somewhat irrevelant, at least > with ODS. The page/buffer size is hard-coded to 2K. Depending upon the system, > your IO transfer size can be as high as 8K (as is the case with Solaris 2.5.1 - I > believe). Yes, you heard correctly, the internal page size is hard coded. Guess > at one time you could set in ONCONFIG but not any more. > > We thought this to be rather stupid. Especially, when we are loading our > statement blobs. The blobs that fit nicely on a page were loading at about 1/6th > of a second, those that spanned load up to 1.1 seconds per. We thought, well we > should just adjust the internal page size right...WRONG...so we live with it. Two things. Actually while pagesize used to be a config file parameter tech support would send a hitman after you if you ever changed it from its default on your platform. Seems that most of the code, even then, assumed a fixed page size at compile time. It is NOT HARDWIRED at 2K on all platforms, there have been several 4K pagesize versions over the years though the vast majority of platforms are 2K. The actual pagesize is determined at compile-time and can be either 2K or 4K and depends on negotiation between the OS/hardware vendor's port team and Informix R&D depending on several mysterious factors which noone I know will reveal. Second, you do realize that you CAN set the pagesize of BLOBPAGES in a BLOBSPACE if you create the BLOB column of your table IN <blobspace> larger blobs can take advantage of the larger pagesize. Art S. Kagel