95
0
2. Set IHS to compress (gzip) content so that it travels the network efficiently
3. Use mod_disk_cache to cache statics and off-load the Portal servers of this work.
4. Disable the ESI cache in the plugin.xml. While this does not necessarily improve performance, there is a suspected bug in the ESI handler that affects Portal.
So, I've attached a copy of my httpd.conf that I use in my cluster to do all three. If you use my copy, you still need to adapt it to your specific setup, but it should get your started.
There is one line like this: "CacheEnable disk /". That simply means "cache everything". Normally, you would not want to cache anything that starts with "my" such as /wps/myportal or /wps/myContentHandler". But Portal places a "pragma no-cache" stanza along with a "cache-control: no-cache" in content that is not allowed to be cached.
Another common question is "how do I know content is caching in IHS? A simple trick is to place the following stanza in the LogFormat option: "Age in Cache: %{Age}o". This will print the age of the content in the cache (in the access_log). If the content is served from the Portal, there is no "Age" header and this you will see "Age in Cache: -". When the content is server from the IHS cache, you will see the "-" replaced by the age of the content in the IHS cache (in seconds).
Note also that when using the reverse proxy cache (like I have done here), that you would normally use htcacheclean to insure that your cache doesn't get too large or that stale entries are periodically removed from the disk. I have put the following in a cron job to do just that:
0 1 * * * /opt/IBM/HTTPServer/bin/htcacheclean -n -t -p/opt/IBM/HTTPServer/cache -l10M >/dev/null 2>&1 # Clean up the IHS cache, to a max of 10Meg
This runs every night to force the cache to occupy no more than 10Meg of space.
There was a bug in mod_cache whereby s-maxage was not honored when making caching decisions. However, IHS has a post-2.2.8 s-maxage fix (PK98225) that resolve this.
With respect to disabling the ESI cache, if the IHS server is controlled by a Deployment Manager, the check box to control this is "Web servers > {your web server name} > Plug-in properties > Caching". If done here, a regen of the plugin will disable the ESI cache without manual intervention. Otherwise, you need to edit the top near the top of plugin-cfg.xml which looks like this: "<Property Name="ESIEnable" Value="true"/>" and change "true" to "false".
Performance: I recently read a blog article that shows that when using Linux (ext3/4) as your IHS mod_disk_cache OS, you can see performance improvements by using "noatime and data=writeback" as fstab options. Linked here is the blog article.
