Talk:High Performance File System
What's "IBM" about it?
If this was, as indicated, developed primarily by Microsoft, why does the title say "IBM"? Or does "IBM" modify "OS/2"? At the time it was developed, OS/2 was a joint IBM/Microsoft product. Guy Harris 20:15, 19 November 2006 (UTC)
IBM rewrote the commonly used hpfs.ifs. The story I heard was that IBM and Microsoft decided to write a next generation hard disk driver and what they would do is each develop one interdependently and then compare the results. Anyways when the comparison was made Microsoft's version was faster and so was excepted. Afterwards when IBM examined the code they discovered that MS had cheated by using i386 assembler when the agreement called for the driver to be built in C and compiled for i286. Upshot was that IBM rewrote hpfs.ifs leaving out the ACL code and limiting the cache to 2MB. The OS/2 client came with IBM's hpfs and the server had Microsoft's hpfs386 as an expensive option. MS charged about $1000 for hpfs386. Eventually IBM rewrote JFS as a replacement for hpfs386 and also forked the new JFS for Linux Dryeo 01:39, 6 March 2007 (UTC)
OS/2 performance issues
One other thing that should be pointed out is that the reason DOS and Windows programs ran faster under OS/2 than DOS was due to the superior performance of HPFS compared to FAT Dryeo 01:42, 6 March 2007 (UTC)
Move
I moved this page back to "High Performance File System". If you search IBM's website for the exact phrase "OS/2 High Performance File System" you get three (3) results. If you search for the exact phrase "High Performance File System" you get hundreds. Same result using either Google Search or IBM's own website search. AlistairMcMillan 20:50, 6 March 2007 (UTC)
HPFS Today
Which Operating Systems still support HPFS today, if any? are there also drivers or utilities that allow operating systems like say, 2000, ME, XP and Server 2003 and Vista to read and write to it? RingtailedFox • Talk • Stalk 19:47, 16 June 2007 (UTC)
- eComstation supports HPFS today. Windows 2000, XP and Server 2003 provide support via a modified pinball.sys (look for hpfsw2k), with both read/write. BartPE (Windows XP boot cdrom) can support it in the same way. Some forms of Linux support it too, as a data disk. Wendy.krieger 08:07, 4 September 2007 (UTC)
HPFS under Windows NT
Windows NT 3.1 contains full support for creating and installing in HPFS partitions. The ability to create HPFS partitions was removed by Windows 3.51, and the driver disappeared in NT 4.0. None the less, if pinball.sys (the driver) is installed in NT 3.51, and upgraded to NT 4.0, then it would still use it.
NT 5.0 and later use a different model, and the driver needs a further patch (eg search for hpfsw2k), such drivers can be used under Windows 2K and Windows XP (there is even a BartPE plugin for it). Wendy.krieger 08:04, 4 September 2007 (UTC)
pinball.sys
is the driver really called pinball.sys? this is kinda LOL. --grawity talk / PGP 18:04, 21 December 2007 (UTC)
- Yes, it is. 195.4.207.39 (talk) 10:19, 22 February 2008 (UTC)
- HPFS codename is, indeed "Pinball" —Claunia (talk) 10:20, 9 April 2008 (UTC)
HPFS386 Clarifications
I've made a clarification for HPFS386's memory usage ability (for caching). HPFS386 is not limited by available memory on OS/2. It is limited by avail memory in the SMA, which is DRASTICALLY different. Inotherwords, if a machine has 4GB of RAM, one cannot use 4GB of RAM (or 3GB or 2GB even) for cache. The limitation will be set by how the various OS/2 memory arenas are defined. Increasing the cache to something retarded, regardless of how much memory OS/2 is recognizing (and it can and does recognize the entire 32bit addressable range), will cause resource issues for other things that use the same memory arena (like thread handling and such). I'd provide citations, but they'd simply point to myself or my colleagues at OS/2 World and elsewhere. ;-) ROBERTMFROMLI | TK/CN 23:16, 2 December 2010 (UTC)
- The cache limitations are also affected by which kernel is in use. The v2-v4 kernels allocate memory different than the Aurora/v4.5x kernels, which also affects how much of the actual RAM is allocatable as an HPFS386 cache. ROBERTMFROMLI | TK/CN 00:05, 3 December 2010 (UTC)