Path: utzoo!attcan!uunet!tank!ncar!unmvax!aplcen!haven!vrdxhq!bms-at!stuart From: stuart@bms-at.UUCP (Stuart Gathman) Newsgroups: comp.arch Subject: Re: What is a Mainframe? Summary: Crays are not mainframes Message-ID: <164@bms-at.UUCP> Date: 29 Jun 89 18:11:41 GMT References: <125@ssp1.idca.tds.philips.nl> <20752@winchester.mips.COM> <27637@lll-winken.LLNL.GOV> Distribution: comp.arch Organization: Business Management Systems, Inc., Fairfax, VA Lines: 53 In article <27637@lll-winken.LLNL.GOV>, brooks@maddog.llnl.gov (Eugene Brooks) writes: > In article <33942@bu-cs.BU.EDU> bzs@bu-cs.BU.EDU (Barry Shein) writes: > >Mainframes are generally at the top of the line in overall I/O > >(particularly disk) performance, that's the critical measure. > Okay, I will bite on this! Here are elapsed times in seconds for > the unpacking of a source archive on some representative machines. > MACHINE and OS TIME in SEC > Cray YMP running NLTSS 40 . . . > SUN3 WITH DISK SUNOS 4.0.1 2.7 A cray is a "supercomputer", not a mainframe. It is not optimized for I/O, but for vector computation. Vaxen are often used as cray front ends. What you need is figures for a 3090, but there're hard to come by since most mainframes run dinosaur OS's that have never heard of a source archive. (Programmers tend not to use macros with their assembler, let alone a high level language.) Most of the massive I/O bandwidth on mainframes is wasted due to stupid software. They can copy files at 3 Mbytes/sec, but even with only one user, editing is faster on a PC. A mini can load indexed files faster than DB2 on a mainframe. I just read an article in Computerworld about how many data centers are automating their operation. The most popular "automation"? A program to ignore the copious messages appearing on the console -- i.e. the operator doesn't need to keep pressing the acknowledge key. I think that even in I/O bandwidth, traditional (three letter) mainframes are about to be overtaken. A typical system has 12 x 3 Mbyte/sec independent (and intelligent) channels. This gives an aggregate throughput of 288 Mbits/sec. The peripheral controller Intel offers with the 486 claims an aggregate throughput of 150 Mbits/sec with 8 independent channels. This is getting close, and it is certainly not hard to beat a 6 Mips CPU these days. I hear that the NeXT has a high throughput peripheral chip also. The micro solution to high throughput disks is not fast (and expensive) disks, but disk arrays. By running lots of cheap disks in parallel you get the throughput of the big jobbers at a fraction of the cost. The fact remains that software, not hardware, is what limits a computer system at any price level. (With a few special purpose exceptions.) -- Stuart D. Gathman <..!{vrdxhq|daitc}!bms-at!stuart>