Hey Paul - I should add that "iostat -xnz 1" is a great method for determine how well the SAN is performance. The asvc_t times are disk IO service times in milliseconds. I usually start there to sanity check disk IO times...
Thanks, /jim Paul Clayton wrote: > Hello.. > > Due to growing performance problems on numerous systems, I have been reading > through a lot of information about DTrace and what it can find out about a > system. It is a great tool, and while there is a learning curve to using it > successfully, getting useful information quickly is helped by such books as > the ‘Solaris Performance & Tools’ book. > > In light of us also ramping up new SANs and ever larger SAN fabrics, I have > long been wondering what the times are for getting data in/out of the SANs. > If we talk with the SAN vendor they do the say the SAN not a problem and we > should look elsewhere. Talk with the fabric folks and they say no problems, > look elsewhere. > > So, it was with high interest that I have been reading Chapter 4 in the book > about disk activity multiple times and trying various commands out on systems > here. > > One thing I have been puzzled with a lot this weekend is the information and > plot in Figure 4.7. This section if I understand it correctly, offers the > means to track the actual times from when an IO starts in the kernel to when > it completes, implying the time to either read or write from disk or memory > cache. > > I have been using a data file for an Oracle database as the test subject for > this work. > > I have several confusion points with this section. > > • The graph mentions ‘strategy’ and ‘biodone’ which seem to imply TNF > based data, not output from the DTrace script above Figure 4.7. > > • In looking at the data gotten from the DTrace script I see no way to > generate the graph of Figure 4.7. Specifically the time difference between > ‘0’ and the points for ‘strategy’. With the DTrace script we have the start > time of the IO. I see no way to determine some amount of time between ‘start’ > and something earlier. The time values on the ‘X’ axis also don’t fall out of > the data generated by the DTrace script. > > • How can it be determined if the IO completed from a memory cache or > required an I/O to a physical disk? I have a lot of times less than 0.5 ms > but also have a fair number that are in the range of 1 ms to 300 ms. > > I modified the script to dump out the size of the I/O being done and that was > interesting to see some unexpected lengths. Also added ‘start’ and ‘end’ to > the appropriate lines as a sanity check to make it easier to pair up the > entries. Should always have one start/end pair for a block address. > > I have attached an Excel spreadsheet with what I was able to create based on > the data collected. > > My thanks for any clarifications to these confusions. > > pdc > > ------------------------------------------------------------------------ > > _______________________________________________ > dtrace-discuss mailing list > dtrace-discuss@opensolaris.org _______________________________________________ dtrace-discuss mailing list dtrace-discuss@opensolaris.org
