iSite Blog

Today i implemented (or finished implementing) Memcached. To get that running, I implemented something I personally refer to as a side-by servlet, but I actually got the idea from OpenWebCMS over at OpenNTF.

Instead of implementing something myself, I went with the idea that somebody had already made what I need, and been a lot smarter about it than me. I tried implementing both EhCache and Memcached, but the "problem" with EhCache is that it runs inside the JVM, and if I dont cluster it (is that possible on the Domino JVM?), it will contain a copy of the cache on all members in a cluster. It also feels like Memcached is faster to me.

Memcached runs as one or more servers on a network, and it was created by the people at Livejournal so that it could run on "leftover" memory. It's no problem setting up a bunch of Memcached servers with 64MB cache on each server (or a few 2GB caches), and you can mix and match server sizes with ease. It all acts like one big cache.

Initial testing has not concluded yet, but number of responses seem to have tripled, and number of concurrent users have doubled. I will come back with more numbers once I have finished crunching them.

With a small modification of the @formula-directive, the formula can now assign a freemarker variable as a Vector. You can do this by using the the parameter name. By setting the name parameter, the formula will not output its result to the page, but assign it instead to a variable.

Example:

<@formula name="_myformula">
@DbColumn("Notes":"Cache";"";"viewDocumentTags";1)
</@formula>

<#list _myformula as v>
Value is ${v}
</#list>

The plan is to use this trick for the Archive on the right, but I have to wait for this change to wind its way trough a proper release before I can use it. It is scheduled for a 6.0.7 release.

When I set up my new development cluster a couple of weeks back, I selected the default server setup for drives on Ubuntu Server 9.10. The default is LVM on Ext4. Apparently Tivoli Storage Manager is unable to read from those drives, and I need to reinstall.

While that sucks a lot, how come Tivoli cannot just use my OS to read from the drives? The OS for sure knows how to read and write to them, so why the need to do something "magical"?

Let's just say I don't like Tivoli at the moment :)