home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.dcom.cell-relay
- Path: sparky!uunet!think.com!ames!kronos.arc.nasa.gov!iscnvx!news
- From: myoung@nebula.ssd.lmsc.lockheed.com
- Subject: Re: fat cells (was Re: Computers dont like ATM?)
- Message-ID: <1992Nov23.185845.5156@iscnvx.lmsc.lockheed.com>
- Sender: news@iscnvx.lmsc.lockheed.com (News)
- Reply-To: myoung@nebula.ssd.lmsc.lockheed.com
- Organization: LMSC, Sunnyvale, California
- 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>
- Date: Mon, 23 Nov 92 18:58:45 GMT
- Lines: 40
-
- >
- >Mark.Hayter@cl.cam.ac.uk (Mark Hayter) writes:
- >>We have been using a system similar to this to allow interworking between
- >>our ATM networks and the Ethernet for several years. The ethernet packet
- >>contains the headers for all of the cells followed by the data - thus the data
- >>is contiguous at the host. This system works well, and the router forwarding
- >>code is efficient (since it is basically just doing cell forwarding).
-
-
- How far does the virtual path extend?
-
- In the data center, point to point optical links, with small windows,
- and variable length packets will be the norm. The model here is that high
- performance risc machines will be connected into a single loosely coupled
- system for data base operations mainly, but also for some finite element
- compute problems.
-
- If all the desktops do utilize ATM interfaces, then their interface
- to the data center may be serving hundreds of users, each having up to
- 10-40 circuits open. Each of the (up to 4000) messages will have to
- be constructed from cells, and then sent to into the data center as
- variable lenght packets.
-
- Extending the virtual path:
-
- If the virtual path is carried into the data center, then this
- interface does not have to be an address gateway. Each user would
- build connections directly to some end point. The data center would
- need to use the same virtual path structures for its internal
- communications, to provide uniform user access to distributed processes.
-
- Result:
-
- Then doesn't this argue for "fat cells" being the variable packet
- equivalent for the fixed cell? If we have sold the virtual path as a
- desktop interface; then the imperative is that it be continued into the
- data center.
-
- Also, the fat cell definition would simplify the mapping to
- traditional MAC layers.
-