paul@cantuar.UUCP (Paul Ashton) (01/20/89)
I have been asked if I can get some information on database benchmarks
that vendors often publish results for. The debit/credit and TP1
benchmarks are ones I am aware of, but have little information on.
If people can give me references to specifications on how to construct
and run these benchmarks I would appreciate it.
Please respond by email, I will post a summary.
Thanks, Paul Ashton
--
Internet(ish): paul@cantuar.{uucp,nz} JANET/SPEARNET: p.ashton@nz.ac.canty
UUCP: ...!{watmath,munnari,mcvax,...!uunet!vuwcomp}!cantuar!paul
NZ Telecom: Office: +64 3 667 001 x6350
NZ Post: University of Canterbury, Christchurch, New Zealandjon@altos86.UUCP (Jonathan Ma) (01/24/89)
In article <880@cantuar.UUCP>, paul@cantuar.UUCP (Paul Ashton) writes: > I have been asked if I can get some information on database benchmarks > that vendors often publish results for. The debit/credit and TP1 > benchmarks are ones I am aware of, but have little information on. Benchmark programs are generally available from the database vendors. I got mine from both Informix (Credit/Debt) and Unify (TP1). They are very useful and easy to use too. We are using TP1 to benchmark our hardware and application software. -Jon- Jonathan Ma, Altos Computer Systems UUCP: {pyramid,sun}!altos86!jon
davek@rtech.rtech.com (Dave Kellogg) (01/30/89)
In article <820@altos86.UUCP> jon@altos86.UUCP (Jonathan Ma) writes: >In article <880@cantuar.UUCP>, paul@cantuar.UUCP (Paul Ashton) writes: >> I have been asked if I can get some information on database benchmarks >> that vendors often publish results for. The debit/credit and TP1 >> benchmarks are ones I am aware of, but have little information on. > Benchmark programs are generally available from the database vendors. > I got mine from both Informix (Credit/Debt) and Unify (TP1). They > are very useful and easy to use too. We are using TP1 to benchmark > our hardware and application software. To overview this in a nutshell. * TP1 has no published specification. There are typically somewhat severe differences in the way different vendors implement TP1. Most vendors use the "DebitCredit" transaction as defined in the DebitCredit benchmark, but ignore many other parameters specific in DebitCredit (e.g. database size). In short, TP1 is usually some degenerate form of DebitCredit. How different people degenerate is not always the same. * DebitCredit has a published specification which was written by Jim Gray of Tandem along with 20+ "friends" (from many different firms and universites) as contributors and reviewers. The spec was published in two places. It, unless you have old copies of Datamation lying around, is probably most easily obtained by asking Tandem. 1. "A Measure of Transaction Processing Power", Tandem Technical Report TR85.2. February 1985, Tandem Computers 2. "A Measure of Transaction Processing Power", Datamation, Volume 31, Number 7, 1985 David Kellogg davek@rtech.rtech.com > > -Jon- Jonathan Ma, Altos Computer Systems > UUCP: {pyramid,sun}!altos86!jon
gupta@cullsj.UUCP (Yogesh Gupta) (02/08/89)
In article <2632@rtech.rtech.com>, davek@rtech.rtech.com (Dave Kellogg) writes: > > To overview this in a nutshell. > > [A succint comment about TP1] > > [A brief history of DebitCredit] > > 1. "A Measure of Transaction Processing Power", Tandem Technical > Report TR85.2. February 1985, Tandem Computers > > 2. "A Measure of Transaction Processing Power", Datamation, Volume > 31, Number 7, 1985 > > > David Kellogg > davek@rtech.rtech.com One thing that people may want to note is that NO commercial vendor today (including Tandem) does the DebitCredit benchmark as specified by Anon et. al. in their paper(s) in 1985. The places where DebitCredit is compromised is: 1) the size of the database; 2) the number of terminals simulated; 3) the response time limitation; 4) the think time interval; 5) the network component. Also, it is important to note that the paper did NOT emphasize the transaction rate alone, but the cost per transaction as well. Most vendors today calculate their costs differently from the way the paper specifies or the way the other vendors calculate, making the numbers VERY hard to compare. However, even with the above state of affair, there is hope in that there exists a council trying to standardize a benchmark which is similar to the DebitCredit to clear up some of this. And yes, I think almost all the DBMS vendors are participating! -- Yogesh Gupta | If you think my company will let me Cullinet Software, Inc. | speak for them, you must be joking.