Hello COH-NUG
Many have asked me "will FlashCache help me in my environment". Well, the answer depends on the degree to which your workload exhibits small, random reads. This, of course, begs another question: "What's does my workload exhibit?".
So, now, to avoid a Catch 22 😉 , my task is to define and explain steps so that I can help you know:
- How to enable the FlashCache simulator for stats enablement.
- How to run perfstat for performance data gathering.
- How to send to me via FTP.
- Wrapping up after perfstat and FlashCache efforts
#1 Enabling FlashCache simulation aka PCS
TR-3801 will tell you how to use PAM II simulation. Please read : http://www.netapp.com/us/library/technical-reports/tr-3801.html
“NetApp® Predictive Cache Statistics (PCS) offer the ability to emulate large read cache sizes and to
measure their effect on system performance. In this way PCS provides a means to approximate the
performance gains of adding one or more Performance Acceleration Modules (PAM or PAM II) to a system.”
Since quick approach might be preferred with less reading, follow these steps:
- Issue all commands on BOTH controllers, which comprise the NetApp storage cluster
- These commands can be run via the console, telnet or ssh.
- Run “options flexscale.enable on”
- To ensure it is on run “options flexscale.enable”
- Does it say so?
- You can run “stats show -p flexscale-pcs” anytime you want to monitor. Look at “reads replaced per second”. This is an optional step.
#2 Running perfstat
- Assuming you have a NOW account, download perfstat from http://now.netapp.com/NOW/download/tools/perfstat/perfstat7.shtml
- Assumes you are using Data ONTAP 7 or 8.0 7-mode.
- If you wish to run the utility from Windows, be certain that the version of Windows you are using has rsh or ssh. Latter versions do not always have these.
- Note: the perfstat needs to be run for *each* controller, which comprises a controller
- There are many options with perfstat. In order to simplify, please follow these simple steps to monitor for 24 hour periods
- A common practice is to capture perfstat data 24 hours a day.
- The recommended method for doing this is to capture multiple small samples throughout the day, rather than one 24 hour sample.
- Many customers script this or place it in the cron job log.
- Many customers now use WINDOWS to run perfstat. NOTE: For Windows users the example below shows a nice way to format perfstat output file names with date and time information:
-- perfstat -f ___<hostname-of-a-controller>__ -t 30 -i 46 > perfstat-<hostname-of-a-controller>"Date_%date:~4,2%%date:~7,2%%date:~10,4%-Time_%time:~0,2%h%time:~3,2%m%time:~6,2%s"
-- Don't forget to run for each controller
-- Example will capture 46 samples of 30 minute intervals. This is a total of 23 hours.
-- Since perfstat uses rsh commands and the time to complete the rsh's can vary, it often runs for slightly longer than the specified time. To compensate for this (and not have 2 perfstat tools running concurrently), we recommend a 23 hour run to cover a 24 hour period.
#3 Sending via FTP
- You can email the results to me. However, since these files are so large, this is usually impossible.
- A better approach might be to upload the resultant files to http://upload.netapp.com/users / where there are simple instructions.
- At this site, it may be confusing, an email address is requested. You should enter *your System Engineer's* NetApp email address.
- Zip files are fine
#4 Wrapping Up
- When finished with all of this, you can disable the FlashCache simulation.
- Use “options flexscale.enable off” from the NetApp ONTAP command line
- Your local NetApp SE will be happy to review the data and let you know if FlasCache would be of benefit.
- Don't forget that FlashCache is able to be tuned to specific types of data, for instance Meta Data. Ask your local SE for more info.
Please don’t hesitate to ask any questions. Thank you!
Matt Kletzly
NetApp Enterprise SE
