Mailing List ArchiveSupport open source code!
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]Re: tlug: Interesting new server
- To: tlug@example.com
- Subject: Re: tlug: Interesting new server
- From: "Manuel M. T. Chakravarty" <chak@example.com>
- Date: Sat, 16 May 1998 19:00:17 +0900
- CC: tlug@example.com
- In-reply-to: <19980515120831B.craigoda@example.com>
- Reply-To: tlug@example.com
- Sender: owner-tlug@example.com
Hi! I enjoyed the nomikai very much! > RESEARCH TRIANGLE PARK, N.C., May 12 /PRNewswire/ -- Red Hat Software, Inc., > publishers of the award-winning Red Hat Linux Operating System, today announced > the $29.95 high performance computer. > > Called Extreme Linux, and also known as The Beowulf Project, this collaboration > between Red Hat Software, NASA Goddard Space Flight > Center, and over two dozen leading research centers will > bring the speed and power of multiple computers--parallel > processing as one computer--to students, researchers, and > end-users worldwide. This is definitely an interesting project and one of the members of the Beowulf project (might even be the leader) wrote and maintains the 3c59x.c ethernet driver, which I use myself, but nevertheless, I think this announcement is a bit too optimistic. This is for a hardware and a software reason. Hardware: Even a 100MBit Ethernet is a joke compared to the networks used in `real' parallel machines. Just consider that, e.g., the Cray T3E provides each processing node with three bidirectional 600MByte links (plus an I/O connection, which in itself is comparable to FastEthernet); the latency in such a network is similarly more advanced (i.e., smaller) than in an Ethernet connection. Software: Unfortunately, here Linux is the problem. Although, the networking code in Linux is quite good for a conventional operating system, the administrative overhead incurred by a user program that wants to inject some data into the network is rather high. A current trend in OSs for parallel machines is to allow user programs direct access to the networking hardware, i.e., the expensive kernel entry, buffer copying etc. is avoided. Of course, it is not easy to allow direct hardware access and still guarantee the integrity of other process on the machine (one possibility is too use special hardware to support this). Extreme Linux is surely interesting for some applications, but `high performance computing' is still something quite different. Cheers, Manuel -------------------------------------------------------------- Next Nomikai: 15 May Fri, 19:30 Tengu TokyoEkiMae 03-3275-3691 Next TLUG Meeting: 13 June Sat, Tokyo Station Yaesu gate 12:30 Featuring Stone and Turnbull on .rpm and .deb packages -------------------------------------------------------------- Sponsor: PHT, makers of TurboLinux http://www.pht.co.jp
- Follow-Ups:
- Re: tlug: Interesting new server
- From: Karl-Max Wagner <karlmax@example.com>
- References:
- Re: tlug: Interesting new server
- From: craigoda@example.com
Home | Main Index | Thread Index
- Prev by Date: Re: tlug: RedHat 5.0 and CDE
- Next by Date: Re: tlug: RedHat 5.0 and CDE
- Prev by thread: Re: tlug: Interesting new server
- Next by thread: Re: tlug: Interesting new server
- Index(es):
Home Page Mailing List Linux and Japan TLUG Members Links