Re: Sybase 11 TPCC on DEC (your comments please)

From: Mark David Butler <butlerm_at_xmission.xmission.com>
Date: 1996/01/06
Message-ID: <4clv72$4ql_at_xmission.xmission.com>#1/1


A 300 Mhz Alpha is not equivalent to a 300 Mhz Pentium. The ratio of CPU performance to clock speed depends strongly on processor architecture ( e.g. bus size, cache size, register size, number of registers, instruction   set, pipelining, branch taken statistics, clocks per instruction, and   much more ). Among other things the Alpha processor is a true 64 bit chip which is greate for implementing relational databases because a 64 bit address space is large enough to map nearly every disk on the planet into virtual memory, potentially removing a layer of I/O architecture from the database code.

It is not important that the TPC-C benchmark is not described. It is only important that the same benchmark was used for both the Oracle and the Sybase tests.

TPC-C is an industry standard relational database benchmark developed by an organization with the same name (The TPC Group?).

  • Mark Butler

travis_at_richmond.infi.net (Glenn Travis) posted with deletions:

 Sybase Leapfrogs Competition; Achieves 'Best On Platform' With SQL Server11

 Source: PR Newswire  

 EMERYVILLE, Calif., Dec. 22 /PRNewswire/ via NewsPage -- Sybase, Inc.
(Nasdaq: SYBS), The Enterprise Client/Server Company(TM), today announced
 that Sybase(R) SQL Server(TM) 11 achieved the industry's best TPC-C  benchmark results ever recorded on the 300 megahertz Digital AlphaServer  8400. The Sybase/Digital benchmark achieved 11,014 transactions per minute
(tpmC) at a price/performance of $222 per tpmC with Sybase SQL Server 11 --
 17 percent faster and 32 percent less expensive than Oracle's best  performance result utilizing the same Digital Alpha chip.

In article <4cf0l7$ik_at_nova.dimensional.com>, The Right Reverend Colin James III <cjames_at_melchizedek.cec-services.com> wrote:

The result seems specious as the TPC-C benchmark is not described.

By contrast, my "primitive" load tests on hardly an optimized object database in compiled BASIC, not in the target development language of Eiffel, produced this result on the following test on a 586-133, 32 MB, on one IDE hard disk, no buffer, no cache (hardly a $40,000 + Alpha box).

One transaction consists of 9-reads from 3-memory-tables, 4-reads from 4-disk-tables, and 10-writes to 6-disk-tables: 3,360 tpm (I don't know what "tpmC" is).

If a 300 MHz Alpha is linearly equivalent to a 133 MHz Pentium (which it may not be), then 300/133 or about 2.25 * 3360 tpm = 7580 tpm which is about 45% less than the Sybase mark. (It should be an order of magnitude less (10-times less or 1000% less), because my test is in a high-order language BASIC, not an intermediate-order language C.)

The Sybase results do not impress this writer.

-- 
Mark David Butler         ( butlerm _at_ xmission.com )
Received on Sat Jan 06 1996 - 00:00:00 CET

Original text of this message