home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!paladin.american.edu!howland.reston.ans.net!spool.mu.edu!uwm.edu!ogicse!das-news.harvard.edu!endor!adam
- From: adam@endor.uucp (Adam Shostack)
- Newsgroups: comp.unix.aix
- Subject: Re: Prestoserve
- Keywords: I/O, disk
- Message-ID: <1993Jan12.150405.26425@das.harvard.edu>
- Date: 12 Jan 93 15:04:05 GMT
- Article-I.D.: das.1993Jan12.150405.26425
- References: <1irvnlINNhlk@corax.udac.uu.se>
- Sender: usenet@das.harvard.edu (Network News)
- Organization: Aiken Computation Lab, Harvard University
- Lines: 30
-
- In article <1irvnlINNhlk@corax.udac.uu.se> larsowe@csd.uu.se (Lars-Owe Ivarsson) writes:
-
- > Rumor has it that Legato is working on a Prestoserver for the RS/6000.
- >I thougt a Prestoserver, as used by Sun and Dec, was a kind of disk-cache.
- >As many operating systems--e.g. AIX and IRIX, unlike SunOS and
- >Ultrix--use memory-mapped files, I assumed that the operating system
- >would take care of that (given enough memory at least), making a
- >prestoserver unnecessary on IBM and SGI systems, or at least not cost
- >justified. Obviously I was wrong on that, so if anybody would please
- >explain the Prestoserve concept for me, especially in combination with
- >the RS/6000, I would be very grateful.
-
- An NFS client waits for a response from the server saying "ok, I wrote that
- packet to disk." Since disk is slow, and can be heavily loaded down,
- the prestoserve has (battery backed up?) ram, and sends the ACK as
- soon as it recieves the packet.
-
- That way, NFS writes are not slowed down by a heavily loaded server,
- waiting to write the packet.
-
- Does aix map a file into ram and then not guarantee it is written to
- disk until it is close()ed?!?
-
- Adam
-
-
- Adam Shostack adam@das.harvard.edu
-
- What a terrible thing to have lost one's .sig. Or not to have a .sig
- at all. How true that is.
-