pavlov@hscfvax.UUCP (840033@G.Pavlov) (11/20/85)
First, thanks to all of you who replied to my request; some individuals went well out of their way to phone me, or simply to take the time to talk to me. Second, I received requests from people to forward or summarize the informa- tion and opinions I received. Since the bulk of these were by phone, I'll summarize, rather than printing short excerpts. Will try to keep reasonably short - tho DBMS's are a "long" subject. PERFORMANCE My request concerned multi-user perfomance of Ingres, particularly compared to Oracle, if possible. The responses fell into three categories: a. "Ingres is a heavy user of resources", "can slow a system to a crawl", etc. In virtually all cases, these statements came from people who have a heavy user load on their systems, aside from Ingres application users. Adding Ingres processes to this mix quickly pushes response to unacceptable levels; b. Most respondents were at VAX sites (most were managers of those sites). I asked some of them to guess what the saturation point of their system would be, if all the users on their system were running Ingres. 3 out of the 4 "guesses" fell into a very consistent pattern: VAX 750 - 10-12 users; VAX 780 - apx. 18 users; VAX 785 - 20 to 24 users. - no, not very scientific ............. c. One individual formally compared performance between Ingres and Oracle through (what sounded like) a well-thought benchmark set (no, she can't release them and I don't have them). Her basic conclusions: Oracle runs faster when "small" quantities of data are involved; Ingres when "larger" amounts are involved. Once the two curves cross, they continue to diverge. The results were the same, regardless of the number of users involved (range of no. of users: from single user to simulated 18 concurrent users). naturally, discussions went beyond this issue: FUNCTIONALITY Everyone who discussed features and tools were happy with the extent and functionality of them. Several people mentioned shortcomings in some of them. The major ones were: 1. the report writer is somewhat limited; some individuals generate several unique reports, which they call in succession from an HLI program; 2. retrievals in the HLI cannot be nested. SUPPORT What impressed me most from the comments I received on this issue is that the level of support claimed by people, overall, was better than from any other software vendor I have dealt with. All comments, save one, ranged from good to excellent to "beyond the call of duty". One individual stated that she at one time worked in a shop with a Callan system - and that RTI went from bad to worse in support - eventually refusing to deal with them further. A related issue: several individuals mentioned that the AT&T Ingres port was executed by AT&T, and is "supported" by it directly - and that neither the support or the port is very good. One person had much the same to say about Burrough's port to its OEM'd CT machine - but worse than the case with AT&T. It doesn't seem that RTI made a very good decision in this area, in the long run ............. Comments re Oracle's support, overall, fell into a more typical bell-shape, ranging from poor to very good. From comments on the net in the last two years, it appears the Oracle has improved in this area somewhat. ONE LAST COMMENT Several people, myself included, received a benchmark "summary" from Oracle, containing a comparison of version 3 Ingres, version 4.1.4 Oracle, and a beta release of version 5.0.7 Oracle. This was executed on VAX VMS machines by a company called Nokia Information Systems (Finland). There are a number of problems with this "summary" and the benchmark, which I won't go into since I can't tell, in all cases, which are benchmark and which are "summary". Naturally, the "summary"/benchmark claims "signifi- cant performance improvements (for beta Oracle) over Ingres" - all is fair in software vendor wars - HOWEVER, I was told by one person that Nokia decided to buy Ingres. Does anyone know if this is true - and why ? Thanks once again, greg pavlov, FSTRF, Amherst, N.Y.
pavlov@hscfvax.UUCP (840033@G.Pavlov) (11/20/85)
Sorry: two points were garbled in the last message: 1. Re the Ingres-Oracle benchmark run by a person I talked to: when I said that the results were the same, irrespective of number of users, I meant that the relative (Ingres:Oracle) results were the same; 2. The comments re functionality all pertained to Ingres. greg pavlov.