Conference ulysse::rdb_vms_competition

840.0. "Rdb/VMS breaks 300 TPS on a 6000 cluster TPC-B Benchmark!!" by BYBLOS::TAMER () Mon Jan 14 1991 22:20

From:	TLE::SCHUTZMAN "A qui la Faute si le Beurre est Cher?  14-Jan-1991 1414" 14-JAN-1991 14:14:53.74
To:	@NAD$DISTRIB:NAD$DEVELOPERS
CC:	
Subj:	rdb getting faster

From:	NOVA::HORN "Database Systems Product Management  14-Jan-1991 1240" 14-JAN-1991 14:13:02.79
To:	RDBPT_INT
CC:	
Subj:	Rdb/VMS breaks 300 TPS on a 6000 cluster TPC-B Benchmark!!


Brian Duggleby
(603) 884-2423






                     DIGITAL'S Rdb/VMS DATABASE

          PERFORMS 300 TRANSACTIONS PER SECOND RUNNING THE

             TPC-B BENCHMARK ON VAX 6000 PRODUCT FAMILY


MAYNARD, Mass. -- January 14, 1991 --Digital Equipment Corporation 
today announced Transaction Processing Performance Council Benchmark 
B (TPC-B) performance results for Rdb/VMS, Version 4.0, its 
strategic relational database. The test ran on a VAXcluster system 
consisting of four VAX 6000 Model 540 systems. These results, in 
which Rdb attained a rate of 300.1 transactions per second (tpsB), 
were audited by the independent consulting firm of KPMG Peat Marwick 
and reported to the Transaction Processing Performance Council (TPC) 
on January 11, 1991.
     "The 300 transactions per second that Rdb sustained during 
testing is the highest TPC-B performance rate reported by any 
relational database vendor to date. High-performance, full support 
for the VAXcluster environment and high availability and integrity 
features make Rdb the clear database choice for production system 
applications. Rdb Version 4.0 enhancements mean even higher 
performance in a VAXcluster configuration," said Vickie Farrell, 
Marketing Manager for Digital's Database Systems Group. Rdb, Version 
4.0, began shipping in December, 1990.


     Unlike TP1, the TPC Benchmark B is a fully specified benchmark 
standard and is the result of Digital and other members of the TPC 
working together to create an industry-wide standard for evaluating 
the performance and price/performance of transaction processing and 
database systems. TPC-B was approved by the members of the TPC in 
August, 1990. The configuration and results of the Rdb test are given 
below. The VAXcluster system utilized during the Rdb performance test 
was configured with 72 disks storing approximately 35 gigabytes of 
database files. The complete report, including the auditors' 
attestation letter, is available from Digital upon request.

System                                              TPS      Price per TPS
Configuration   Nodes  CPU Model     Software      (tpsB)     (K$/tpsB)
--------------------------------------------------------------------------

VAXcluster       4     VAX 6000-540  Rdb/VMS V4.0   300.1     $18.0
                                     VMS V5.4


     In addition, Digital also announced version 1.1 of DECtrace for 
VMS software, a VMS-layered product that collects and reports event-
based data for use in performance analysis, capacity planning and 
database tuning in conjunction with DEC RdbExpert software. The 
latest version of DECtrace software features a Runtime license 
option, a lower cost alternative for systems which only need to 
collect DECtrace data for subsequent formatting and reporting on a 
node executing the Full Development DECtrace software. The DECtrace 
Runtime capability will be particularly useful for enterprises with 
multiple computer systems, allowing them to economically incorporate 
DECtrace technology into their applications and database 
environments. Clusterwide pricing for the DECtrace Full Development 
option is $4370; the Runtime option is priced at $1740. DECtrace, 
Version 1.1 will ship in June, 1991. 



     Digital also announced that DEC RdbExpert, an AI-based database 
design and tuning tool, developed specifically to assist Rdb 
Database Administrators and programmers, will ship on January 28, 
1991. RdbExpert analyzes database logical design, transaction 
workload environment, data volume and the system environment to 
generate SQL Data Definition Language and procedures to enhance 
database performance and efficiency. Clusterwide DEC RdbExpert 
pricing is $17,100. 
     Digital Equipment Corporation, headquartered in Maynard, 
Massachusetts, is the leading worldwide supplier of networked 
computer systems and services.  Digital offers a full range of 
computing solutions and systems integration for the entire 
enterprise -- from the desktop to the data center.
                                ####


Note to Editors:  DEC RdbExpert, DECtrace for VMS, VAX, VAXcluster, 
                  Rdb/VMS, and VAX/VMS are trademarks of Digital 
                  Equipment Corporation.

     		  TPC-Benchmark B is a trademark of the Transaction 
                  Processing Performance Council.

CORP/91/056

	Isn't this more significant since it users 4 processors? In
	the previous TPS reports, we were limited to 2 processors as
	the lock overhead after this point caused exponential 
	diminishing returns. 

	Is the full configuration available for this? Is it on-line?

re .1

Why don't you contact 
Brian Duggleby at (603) 884-2423 or the author of mail message at NOVA::HORN.

PS: Digital News in the January 21, 1991 issue, page 56 has an article about it.

    [ I forwarded .0 to an on-site Digital consultant, who replied as
      follows; can anyone help? 
    
      Thanks, 
      Larry.]
    

	Larry,

	Thanks for the benchmark data.  I was wondering if there is a
	benchmark comparison between a VAXstation 4000 Model 60 and a VAX4000
	Model 200.  The customer is comparing these 2 machines.  We are
	doing a benchmark between these 2 machines internally but it would
	be great if we can get one from the real 'benchmark' people.


	Thanks again.

	Teck

�         <<< Note 840.3 by SWAM2::MCCARTHY_LA "Lie to exit pollers" >>>
�                    -< VS4000 M60 vs. VAX4000/200 TPC No.? >-

Larry,

I checked with the TNSG performance group (who brings us the TPC benchmark 
figures).

Typically they don't benchmark on the workstation class machines (of which 
the 4000-60 is a workstation) so I wouldn't expect to see any TPC figures 
for it.

Currently, there are only TPC-A figures for others of the 4000 family 
(there are no plans at this time for -B figures for the 4000s).

VAX 4000-200, Rdb/VMS V4.1, VMS V5.5		17.80 tpsA
VAX 4000-300, Rdb/VMS V4.1, VMS V5.5		31.90 tpsA
VAX 4000-500, Rdb/VMS V4.1, VMS V5.5		62.40 tpsA

	Re. .4

	> (there are no plans at this time for -B figures for the 4000s).

	Well, at least not with Rdb, but there may be plans for a TPC-B test
	with other database vendors. Check with the 4000 development group.

	-amitabh.

    re: .4,
    
    Thanks, I suspected as much. Would anyone care to hazard a rule of
    thumb, SWAG, or anything else, as to how the 4000-60 might compare to
    it's related 4000 Model x00 (which ever that is)? 
    
    Thanks again,
    
    Larry.

The VAXstation 4000-60 is based on the MARIAH chipset and has a SPECmark
rating of 10.6 (12 VUP), so its faster than the SOC/C-based VAX 4000-200
(5 VUP) or RIGEL-based VAX 4000-300 (8 VUP).  The NVAX-based VAX 4000-500
(24 VUP) is twice as fast as the VAXstation 4000-60.

The key factor in using the 4000-60 is likely to be I/O.  The 4000-x00
products use DSSI and offer up to 4 DSSI busses (2 on the CPU board and 2
via KFQSA on the Q-Bus).  The 4000-60 uses SCSI I/O and appears to have
only 1 SCSI adapter.  It also has a single Turbochannel slot, which
*might* be able to take a SCSI (or 3rd party storage) adapter but that
isn't currently shown as an available option.  Between the relative
merits of DSSI over SCSI, and the performance and connectivity limits of
a single SCSI bus, the 4000-300 should outperform the -60 in disk I/O
intensive applications, despite the -60s CPU advantage.

Beyond that I don't know what to say.  I won't speculate on actual TPC-B
or TPC-A numbers.  The 4000-60 is designed to have fast graphics,
reasonable expandability, and fast I/O within the limits expected of a
workstation.  It is not designed as a high disk activity server.  The
4000-x00 family is specifically designed to be good servers.

Hal

The nearest thing to a VAXstation 4000-60 that has been tested is the 
MicroVAX 3100 Model 80.

The results (using Rdb 4.1 and ACMS 3.2) were 27.9 on the TPC benchmark.

You can get the full report, and lots of others besides, from 
TPSDOC::SYS$PUBLIC:[TP_PERFORMANCE]

    Thanks to everybody who pitched in on this. I forwarded your replies to
    the on-site consultant, who, as you'll recall, has been helping the
    customer benchmark the 4000-200 and the VS4000-60 with the customer's
    application. His reply:
    
	Larry,

	Thanks a lot on the benchmark info.  It was very helpful.  Just 
	FYI, our internal [ie, done with the customer] benchmark shows the
    	VAX 4000 Model 200 was	about 40% faster than the VAX 4000 Model 60
    	on a very i-o intensive application.

	Once again, thanks.

	Teck