Phil, I understand your concerns, that's why we ran it by Brendan, who ran it by SAE. As you point out there are issues with iozone in general, and with Solaris. filebench is clearly a much better benchmarking framework.
But, it is still a commonly used open source benchmark, and if we don't port it someone else will and we may be in a better position to promote some change upstream. Cheers, Jim Phil Harman wrote: > I have a number of concerns: > > 1) it should be called cachezone, not I/O zone > > Most of the examples in the short Iozone paper do not actually show I/O > performance, but get side-tracked on the various caches in the system. > > 2) it produces hard to understand data, which leads to meaningless > conclusions > > It does nothing useful to explore the size of the filesystem cache, and > can only draw a line between cached and not-cached performance, with it > being especially hard to produce data for non-cached filesystem I/O. > > 3) it is a support call generator > > I had one prestigious customer that was claiming 1GB/sec through a 4Mbps > HBA. I've seen a number of other examples where I have had to make up > for a poor undertanding of that the benchmark does, and why its data > isn't that helpful. And what about the cases where decisions are made on > unchallenged data? > > 3) it is not very Solaris savvy > > Most databases use O_DSYNC (not O_SYNC), but there is not O_DSYNC option > > 4) is it pro VxFS to the exclusion of UFS and others > > It hasVX_DIRECT support, but no notion of directio(3C) > > 5) the options are seemingly random, but not easily extensible > > 6) there is very little control over the workload mix > > 7) we already have filebench > > and so on. > > Phil > > > > James Walker wrote: >> I'm sponsoring this familiarity case for Ivan Shi. The requested >> release binding is minor. The man page has been posted in the >> materials directory. >> >> Template Version: @(#)sac_nextcase 1.68 02/23/09 SMI >> This information is Copyright 2009 Sun Microsystems >> 1. Introduction >> 1.1. Project/Component Working Name: >> iozone >> 1.2. Name of Document Author/Supplier: >> Author: Ivan Shi >> 1.3 Date of This Document: >> 27 April, 2009 >> 4. Technical Description >> Summary >> ======= >> Iozone[1] is a filesystem benchmark tool. The benchmark generates >> and measures a variety of file operations. It is useful for >> performing a >> broad filesystem analysis of a computer platform. >> >> Iozone3_321 will be integrated into the SFW consolidation as part of >> this proposal, and will be installed as SUNWiozone. >> >> This project requests a minor release binding. >> >> >> Dependencies >> ============ >> >> None >> >> >> Interfaces >> ========== >> >> Exported Interfaces Classification Comment >> ------------------- -------------- ------------------ >> SUNWiozone Uncommitted Package >> /usr/benchmarks/iozone Uncommitted Command >> >> Imported Interfaces >> ------------------- None >> Reference Documents >> =================== >> [1] http://www.iozone.org/ >> >> RFE ID# 6831877 >> >> >> 6. Resources and Schedule >> 6.4. Steering Committee requested information >> 6.4.1. Consolidation C-team Name: >> SFW >> 6.5. ARC review type: FastTrack >> 6.6. ARC Exposure: open >> >> >
