perm filename MILLER[S77,JMC] blob
sn#281416 filedate 1977-05-10 generic text, type T, neo UTF8
TYPE <GORIN>LETTER.MILLER1
00100 OFFICE MEMORANDUM * STANFORD UNIVERSITY * OFFICE MEMORANDUM
00200
00300 LOTS COMPUTER FACILITY
00400
00500
00600 May 9, 1977
00700
00800 To: Professor William Miller, Provost
00900
01000 From: Ralph E. Gorin, Manager
01100
01200 Subject: Upgrade opportunities for LOTS
01300
01400 Since my trip to Boston in mid-April, Digital Equipment Corporation
01500 has offered Stanford a new hardware upgrade package for LOTS:
01600
01700 256 K words of main memory $126,000
01800 1 RTP06 200 megabyte disk and channel 47,000
01900 1 RP06 200 megabyte disk 35,000
02000 1 Cache Memory 100,000
02100 Total list price: $308,000
02200
02300 DEC's price to Stanford for this package is $215,000. The previous
π
02400 offer from Digital that we discussed with you was the same hardware
02500 as above, less one RP06 disk, for $230,000. DEC's current offer
02600 seems to be rather more attractive.
02700
02800 This upgrade will double our memory capacity, more than double the
02900 speed of the processor, double our I/O throughput and triple our disk
03000 storage capacity.
03100
03200 To amortize these upgrades the LOTS budget needs to be increased by
03300 about $32,000/year.
03400
03500 The LOTS system is already loaded to an extent where we doubt the
03600 system's ability to deal with the expected load increase in the fall
03700 unless we either upgrade the hardware facilities, decrease the load,
03800 or both. We are investigating the means by which the software can be
03900 made more efficient. However, we do not expect that software changes
04000 alone will substantially increase throughput.
04100
04200 One large component of the anticipated load increase is for the
04300 introductory Computer Science courses. Last fall, these courses
04400 included 250 students. We estimate that␈this usage requires us to
04500 support an additional 14 terminals for this purpose alone. The
04600 reasons to move the introductory courses to LOTS are compelling. If
04700 LOTS is to fulfill its purpose in the University, the introductory
π
04800 courses must be taught using LOTS so that students, after only one
04900 introductory course, can continue to use LOTS as a tool throughout
05000 the remainder of their studies here at Stanford. But LOTS, as
05100 presently configured, can not handle that load.
05200
05300 It would be possible, but undesirable, to ␈shed some load by
05400 restricting the use of LOTS to directed coursework and formally
05500 directed projects and research. One of the accomplishments that was
05600 sought from LOTS was the provision of general access computing for
05700 students. We believe that LOTS has been sucessful in this respect,
05800 and that general access computing is an educationally valid purpose.
05900 Undirected projects do, however, use a significant fraction of the
06000 LOTS resource (10 to 15%). Moreover, under the current funding
06100 arrangement, this use of LOTS is not paid for by anyone, in the sense
06200 that there is no fund nor allocator of money for undirected
06300 computing.
06400
06500 As you requested, John McCarthy and I have talked to Dean Royden and
06600 Dean Kays to see if they would be willing to allocate a larger share
06700 of their academic computing budgets to LOTS. Also, I have met with
06800 Paul Johnson, Assistant Dean of the Graduate School of Business, for
06900 the same purpose. We feel that we are not really in the right
07000 position to request the reallocation of funds. The last time that
07100 was done, it was done by the Associate Provost for computing. So
π
07200 far, we have had only limited success. Dean Royden has tentatively
07300 agreed to allocate an additional $10,000/year to LOTS. Dean Kays
07400 would allocate an additional $5,000 if we provided more terminals
07500 than originally planned for the Terman Center. However, the cost of
07600 the additional terminals there will probably cost more than the
07700 $5,000. (You will recall that the original LOTS budget did not
07800 provide any money for communicating with any remote terminals.)
07900 Finally, GSB, far fro␈ being willing to allocate additional money to
08000 LOTS, is now trying to spend their entire allotment at SCIP. The
08100 attitude at GSB regarding paying for LOTS depends to some degree on
08200 their plans regarding accquiring thei␈ own DECsystem 20. GSB is
08300 expecting to decide about these things by the end of this month.
08400
08500 LOTS is an effective and economical educational facility. We are
08600 presently serving about 1200 active student and aculty users. With
08700 the hardware upgrades proposed here LOTS can double its user
08800 community and provide better service. We propose to double the
08900 available service with less than a 15% budget increase.
09000
09100 We may be criticized because we have yet to firmly confront the
09200 resource allocation problem at LOTS. Although it would be a good
09300 thing to address this problem, the lack of resources exists as a real
09400 problem independant of a solution (or lack of a solution)␈of the
09500 allocation problem. The best possible allocation of the existing
π
09600 resource will leave many users without sufficient computing in the
09700 fall. Only an expansion of resources will satisfy the needs of the
09800 of the user community.
09900
10000 Finally, there is some need for haste. Other DEC customers are
10100 presently being quoted 5 to 7 month (or longer) delivery times on
10200 some of the items we need. We may already be too late to obtain
10300 these upgrades for the beginning of fall quarter, but we should try.
@