Default page size and variants
Jacob Salomon asked about default page sizes. Art Kagel said Windows and AIX default to 4K, other platforms to 2K, and dbspace page sizes must be multiples of the default. Paul Watson noted XPS lets PAGESIZE be set in ONCONFIG (8K on his AIX system). The rest was reminiscence about VMS disk blocks and pages.
Auto-generated by Claude from the posts below — may be imperfect; read the full thread.
Hi Folks. I have a question regarding the default page size for a server. Actually a couple, though I think I have one answer already. All the Informix servers I have encountered in the past 15 years or so have all had a default page size of 2K. I recall that some Solaris systems and some old IBM workstations preferred a default page size of 4K. Are there ANY servers still around that use a 4K page size? If you have such a 4K-preferred system, I presume it is not possible to create a DBspace with a 6K page size; that all DBSpaces must necessarily have a page size that is a multiple of 4K. I'm just requesting a fact-check on this issue. I understand that in the more ubiquitous 2K-page system, the largest possible page is 16K; eight 2K pages. In a 4K-page server, is that limit still 16K or does it expand to 32K? In each page header there is a 3-bit bit-field that seems to indicate this page's size as a multiple of the default page size but I have no access to any 4K-page server to check this. I see that this forum has designated me a badge for "discussion party starter"; I am quite tickled at that! ------------------------------ Jacob Salomon --- Nobody goes there anymore, it's too crowded. --Attr: Yogi Berra ------------------------------
Jacob: All the Informix servers I have encountered in the past 15 years or so have all had a default page size of 2K. I recall that some Solaris systems and some old IBM workstations preferred a default page size of 4K. Are there ANY servers still around that use a 4K page size? Windows and AIX systems are 4K default pagesize systems (MAC used to be but it is no longer supported). All others, including Solaris, are 2K. If you have such a 4K-preferred system, I presume it is not possible to create a DBspace with a 6K page size; that all DBSpaces must necessarily have a page size that is a multiple of 4K. I'm just requesting a fact-check on this issue. Exactly. All dbspaces must have a pagesize that is a multiple of the base pagesize. I understand that in the more ubiquitous 2K-page system, the largest possible page is 16K; eight 2K pages. In a 4K-page server, is that limit still 16K or does it expand to 32K? In each page header there is a 3-bit bit-field that seems to indicate this page's size as a multiple of the default page size but I have no access to any 4K-page server to check this. Currently (note currently) the widest pagesize, regardless of the base pagesize, is 16K. ------------------------------ Art S. Kagel, President and Principal Consultant ASK Database Management Corp. www.askdbmgt.com ------------------------------
With XPS you can specify the PAGESIZE in the ONCONFIG, the AIX XPS system I look after is 8K
Paul said: > With XPS you can specify the PAGESIZE in the ONCONFIG, the AIX XPS system I look after is 8K Hi Paul. That's actually a pretty cool item to be aware of. While I was still at Informix, we were trying to adapt IDS (or was it still OnLine?) to VMS (anyone old enough to remember that great OS?) and getting horrid benchmark scores. Having come from a VMS background, I knew that the basic unit of disk I/O was an 8K page and suggested compiling that as PAGESIZE in their setups. I never did find out if they succeeded with the sale or of IDS ever got implemented on VMS. (According to an article in Wikipedia , VMS is still around, latest release in June 2023.) +----- Jacob Salomon --------------------------------------------------+ | The first lesson of economics is scarcity: there is never enough of | | anything to fully satisfy all those who want it. The first lesson of | | politics is to disregard the first lesson of economics. | +--------------- Thomas Sowell (Is Reality Optional?::Student Loans) --+
under VMS I thought was 512bytes at the underlying level but presented as 8K
Paul said: > under VMS I thought was 512 bytes at the underlying level but presented as 8K Essentially correct, Paul. I think. I suppose block addressing for disk I/O could still be in terms of 512-byte sectors. But in the virtual memory scheme used by the VAX firmware, memory pages are in blocks of 8K each. Since the disk page will be read into [virtual] memory, in makes more sense to zap in the disk I/O in 8K blocks. Amazing that I still remember those details from a VMS internals class 30+ years go. Now, what did I have for lunch yesterday? <grin> +----- Jacob Salomon --------------------------------------------------+ | The first lesson of economics is scarcity: there is never enough of | | anything to fully satisfy all those who want it. The first lesson of | | politics is to disregard the first lesson of economics. | +--------------- Thomas Sowell (Is Reality Optional?::Student Loans) --+