home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!pipex!warwick!uknet!pavo.csi.cam.ac.uk!mdh
- From: mdh@cl.cam.ac.uk (Mark Hayter)
- Newsgroups: comp.dcom.cell-relay
- Subject: Re: fat cells (was Re: Computers dont like ATM?)
- Message-ID: <1992Nov23.180730.22608@infodev.cam.ac.uk>
- Date: 23 Nov 92 18:07:30 GMT
- References: <sjdveq8@sgi.sgi.com>,<1992Nov20.122227.6002@infodev.cam.ac.uk> <1992Nov20.150333.17701@iscnvx.lmsc.lockheed.com> <1992Nov20.211632.18945@infodev.cam.ac.uk> <1992Nov23.135258.29925@merlin.dev.cdx.mot.com>
- Sender: news@infodev.cam.ac.uk (USENET news)
- Organization: U of Cambridge Computer Lab, UK
- Lines: 27
- Nntp-Posting-Host: forties.cl.cam.ac.uk
-
- peterd@pjd.dev.cdx.mot.com (Peter Desnoyers) writes:
- > I'm curious as to why you chose to aggregate all the cell headers at
- > the front of the packet, rather than just the first - did this design
- > predate AAL5?
-
- Yes, our internetworking protocol and its datalink layers were in use
- well before AAL-5. The original version was designed by Derek McAuley
- as part of his thesis work in 1988-9, and this is based on an earlier
- SAR protocol. In this the SAR information is contained in the header, so
- we wish all headers to be carried to the end points (remember we are just
- doing cell forwarding at the ethernet-ATM boundary, this does not know
- anything about SAR). Clearly the end system on the ethernet is happier if
- all the data is contiguous (in the common case of no cell loss all that
- needs to be done is scan the SAR to see it has the expected values and
- DMA the data to its IPC buffer). The SAR information contains the
- equivalent of the "user indication" bit. In fact we experimented with
- two bits (essentially start and end of block) but only one is actually
- required.
-
- Mark
-
- --
- ------------------------------------
- Mark Hayter
- Cambridge University Computer Lab
- Mark.Hayter@cl.cam.ac.uk
- Tel +44 223 334645
-