aglew@ccvaxa.UUCP (09/26/86)
>This is another problem with virtual memory - central memory is >starting to get cheaper and bigger than the disk memory. The biggest >disk drive available with a Cray these days is a DD-49 (holds about >151 MW). One memory image of the 4GW Cray 3 would fill 26 DD-49s to >capacity. What are you going to operate virtual memory out of? > >J. Giles >Los Alamos Main memory will be as much high speed RAM as you can afford. Virtual memory will be slightly slower speed but less expensive RAM. You'll still be able to page - but paging will be done so fast that you won't have time for all the complicated paging algorithms that are used today. Andy "Krazy" Glew. Gould CSD-Urbana. USEnet: ihnp4!uiucdcs!ccvaxa!aglew 1101 E. University, Urbana, IL 61801 ARPAnet: aglew@gswd-vms
phil@amdcad.UUCP (Phil Ngai) (10/02/86)
In article <5100141@ccvaxa> aglew@ccvaxa.UUCP writes: > >>This is another problem with virtual memory - central memory is >>starting to get cheaper and bigger than the disk memory. The biggest >>disk drive available with a Cray these days is a DD-49 (holds about >>151 MW). One memory image of the 4GW Cray 3 would fill 26 DD-49s to >>capacity. What are you going to operate virtual memory out of? >> >>J. Giles >>Los Alamos > >Main memory will be as much high speed RAM as you can afford. >Virtual memory will be slightly slower speed but less expensive RAM. In fact, the Cray 2 uses ordinary dynamic RAMs for main memory. These are the same type of devices your IBM PC uses. Of course, Cray interleaves them many ways to get the data bandwidth needed. But the point is, main memory is already made out of the most cost effective memory devices available. There is no "slower speed but less expensive RAM". Note that I am talking about main memory, not cache, which is made of expensive high speed devices. As cache is handled by hardware, it is not relevant to a virtual memory discussion. There was a claim that RAM is getting cheaper than disk. Assume 474Mb eagle at $10,000. This yields $2E-5 per byte. Assume 256Kbit DRAM at $2.56. (see 10/1/86 San Jose Mercury News Fry's Electronics ad) This yields $8E-5 per byte. Rotating machinery is still cheaper than silicon. -- In Arizona they brag about how much water it takes to maintain their lawns and golf courses. Can you say "aquifer overdraft"? Phil Ngai +1 408 749 5720 UUCP: {ucbvax,decwrl,ihnp4,allegra}!amdcad!phil ARPA: amdcad!phil@decwrl.dec.com
peters@cubsvax.UUCP (Peter S. Shenkin) (10/05/86)
In article <amdcad.13214> phil@amdcad.UUCP (Phil Ngai) writes: >There was a claim that RAM is getting cheaper than disk. Assume 474Mb >eagle at $10,000. This yields $2E-5 per byte. Assume 256Kbit DRAM at >$2.56. (see 10/1/86 San Jose Mercury News Fry's Electronics ad) This >yields $8E-5 per byte. > >Rotating machinery is still cheaper than silicon. And the advent of R/W optical disks will make this truer; even though you can only write once, and access is basically sequential, a large part of their use will be to replace disk rather than tape. For instance, I'm now sitting on about 10Mb of data coughed up by molecular mechanics programs; I have to look at this stuff again and again, reanalysing it in the light of what I never thought of before, in order to make sense of it. My continually changing analysis programs will continue to reside on our Eagle, but as soon as we get our optical disk (hopefully this year) all this stuff can be moved onto it. Our facility also does optical image processing, and the situation is similar; we have a 36Mb image library on the Eagle. We also have groups of people doing pattern-matching on protein and nucleic acid sequences; the library for this info is also huge, and is on the Eagle. In fact, the entire Brookhave National Lab's Protein Data Bank is something we desperately need to have on line all the time, but can't because we don't have room. We spend large amounts of time moving files in from tape and deciding wwhich ones to delete. Thus we're going to have the distinction between static and updatable on-line storage, and the static storage is going to be cheaper than the dynamic. Also, since the cost of the medium itself is negligible (as opposed to the cost of the machine that reads and writes on it), static doesn't even have to be THAT static; all the manual pages can go on it, for instance. Peter S. Shenkin Columbia Univ. Biology Dept., NY, NY 10027 {philabs,rna}!cubsvax!peters cubsvax!peters@columbia.ARPA
joel@peora.UUCP (Joel Upchurch) (10/23/86)
>Peter S. Shenkin {philabs,rna}!cubsvax!peters cubsvax!peters@columbia.ARPA >> >>Rotating machinery is still cheaper than silicon. > >And the advent of R/W optical disks will make this truer; even though you >can only write once, and access is basically sequential, a large part of >their use will be to replace disk rather than tape. For instance, I'm This discussion started on the virtues of virtual versus non-virtual memory systems. The optical devices I've heard of so far would make terrible paging devices. The disadvantages of using a WORM optical disk as paging device are obvious. Even proposed devices that could rewrite blocks, would not make good paging devices because of the awful seek times and rather anemic transfer rates associated with these devices. Has anyone seen any optical disks with numbers in these categories comparable to good magnetic disks? That is not to say that a optical disk dosn't beat the Hell out of having the data sitting in the tape library. -- Joel Upchurch @ CONCURRENT Computer Corporation (A Perkin-Elmer Company) Southern Development Center 2486 Sand Lake Road/ Orlando, Florida 32809/ (305)850-1031 {decvax!ucf-cs, ihnp4!pesnta, vax135!petsd, akgua!codas}!peora!joel