home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!!CONSULTANT"
- X-Envelope-to: sas-l@vtvm2.bitnet
- X-VMS-To: BITNET::"sas-l@vtvm2.bitnet"
- X-VMS-Cc: RTIHCG,CLU,HPK
- MIME-version: 1.0
- Content-transfer-encoding: 7BIT
- Message-ID: <01GMZXR1CN1E00004U@RTI>
- Newsgroups: bit.listserv.sas-l
- Date: Thu, 30 Jul 1992 18:20:02 -0400
- Reply-To: "Harlene C. Gogan: RTI's PC-SAS Technical Consultant"
- <RTIHCG@RTI.BITNET>
- Sender: "SAS(r) Discussion" <SAS-L@UGA.BITNET>
- From: "Harlene C. Gogan: RTI's PC-SAS Technical Consultant"
- <RTIHCG@RTI.BITNET>
- Subject: Wanted: A machine to run SAS
- Lines: 75
-
- _____________________________________________________________________
- CONTENT: Question: opinions solicited
- SUMMARY: We need to purchase machine for running LARGE SAS jobs,
- i.e., LOTS of observations, CPU intensive models
- REL/PLTF: what do you suggest for task at hand
- E-ADDR: RTIHCG@RTI.BITNET
- NAME: Harlene C. Gogan
- ADDRESS: Research Triangle Institute, PO 12194,
- Research Triangle Park, NC 27709
- PHONE: (919) 541-6565
- ______________________________________________________________________
-
- The dilemma--we need to purchase a machine that will run LARGE SAS jobs.
- The definition of LARGE is 80,000 observations per year, 14 years of
- data, say 25 categorical numeric variables (average of 3-5 categories)
- and 5 continuous variables. These are birth records for North Carolina.
- We want to model infant mortality. Thus we must run polytomous logits
- with a 5 category outcome: death in 1st 24 hours
- death in 1st week (mutually exclusive)
- death in first month (categories)
- survival, low birth weight
- survival, normal birth weight
- (please note, though I said 25 categorical variables, these would probably
- be entered as independent dummy variables, i.e, 3 categories=2 dummies,
- 5 categories=4 dummies, etc.)
-
- We have been attempting to run these models on a mainframe. Since we
- are not for profit, we must buy time on a local university computer--
- thus we pay REAL money for computing. A single model using whites only
- costs $600-800. At that price we cannot afford to do the research that
- is required without resorting to some sampling strategy, or reducing
- categories, which is unappealing from a research perspective.
-
- We have a grant to purchase equipment and have approximately $14,500
- to spend. What machine would you suggest for datasets/models of this
- magnitude and why? We would like to be able to integrate this machine
- into our DECnet network (so we could at least FTP data in/out).
-
- We are aware that mainframe "costs" are a way of recooping the "people"
- costs of operating and maintaining a mainframe. The costs associated with
- the machine we buy (for installation, operation, etc.) would not be paid
- out in the same manner, i.e, we already get paid, just the scope of our
- job changes to include operations, etc. We may see "real" costs paid
- out if we hire a consultant to do installation/optimization for SAS.
-
- Our in-house computer operating system people do exclusively DEC VMS
- and DOS (which includes PATHWORKS). We'd be on our own for other
- platforms, i.e, if we needed help, we'd need to hire a consultant.
- In the future, I would like to see the machine we buy fully integrated
- into PATHWORKS, so we could have those capabilities in addition to the
- ones being integrated into DECnet gets us. (I know I'm using integrated
- loosely, it's not readily definable.)
-
- I asked SI people at last years SUGI (New Orleans) this question (before
- we had $ available for a machine), and they suggested a RISC platform.
- Given the politics of our computing situation (at that time we had
- some UNIX people on site, and the VMS people were toying with getting
- into UNIX), a DECstation 5000 was suggested (with 1Gb hard disk and
- 32Mb memory).
-
- Does this advice still seem solid, given what the world of SAS computing
- has learned since then? Are there specifics that I need to be aware of
- and be sure to order?
-
- Thank you for your advice.
-
- ________________
- Unafraid to open Pandora's box, she bravely queried SAS-L.
-
- Harlene C. Gogan
- RTI's PC-SAS Technical Consultant
- RTIHCG@RTI.BITNET
- RTIHCG%ZEUS@RTI.RTI.ORG
- RTI, PO12194, RTP,NC 27709
- (919) 541-6565
-