home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!mcsun!dxcern!sunsci!Roberto
- From: Roberto@sunsci.cern.ch (Roberto Divia CERN-ECP/DS +41 22 767 4994)
- Subject: Re: Cache Coherence
- Message-ID: <1992Jul27.065844.15681@dxcern.cern.ch>
- Sender: Roberto@sunsci (Roberto Divia CERN-ECP/DS +41 22 767 4994)
- Organization: CERN - European Laboratory for Particle Physics
- References: <1992Jul23.140423@riese.informatik.uni-frankfurt.de>
- Date: Mon, 27 Jul 1992 06:58:44 GMT
- Lines: 24
-
- I tried to reply directly but the e-mail bounced back...
- > From MAILER-DAEMON@thi.informatik.uni-frankfurt.de Mon Jul 27 08:50:08 1992
- > Subject: Returned mail: Service unavailable
- > <<< 553 riese.thi.informatik.uni-frankfurt.de host name configuration error
- > 554 <feivel@riese.informatik.uni-frankfurt.de>... Service unavailable
-
- Here is my original message in reply to Christine Gross request for
- information about caches and methods to get cache coherence.
-
- Christine,
- are you interested in the way SCI (the Scalable Coherent Interface) keeps
- cache coherency accross ringlets? Snooping is impossible since there is no
- exchange of signals accross separate ringlets. So, it must be done via
- messages and directories.
-
- SCI keeps a distributed coherency directory scheme (opposed to centralised
- directories very much used today but very delicate and non scalable) which is
- source of many discussions but proposes a very elegant, scalable and robust
- solution.
- --
- | Roberto Divia` | Love at first sight is one of the greatest |
- | ============= | labor-saving devices the world has ever seen |
- | CERN : European Laboratory for Particle Physics, 1211 Geneva 23 |
- | Switzerland (CH) |
-