Friday, May 21, 2010


So as I'm sitting here at the table, drinking my light-roast coffee and feeling pretty good about life, it's time to let my readers know (all 1 of them, me) about some cool techy stuff that my mind is on.

The above link is to the changelog for kernel 2.6.32 (which I currently use) and its improvements. Notably, some great improvements have been done in virtualization as with KSM (Kernel Shared Memory), a daemon process goes through page files determining which files are the same and merges them to reduce memory usage. That in itself is great but it doesn't put anything in perspective. Until you read the following:

"The result is a dramatic decrease in memory usage in virtualization environments. In a virtualization server, Red Hat found that thanks to KSM, KVM can run as many as 52 Windows XP VMs with 1 GB of RAM each on a server with just 16 GB of RAM."

If an XP machine was to be run under minimum requirements, using 128 MB of Ram, you'd get a barely functional OS and 52 of them would use a total of 6.656 GB of Ram. I've used XP in a VirtualBox VM using 512 MB of Ram and got some decent performance out of it. (Ran SAP and Borland Together in it; not at the same time) So assuming that each user is limited to a 512 MB Windows XP Pro OS, you'd be looking at a total of 26.624 GB of Ram required if these were dedicated machines. By sharing memory and having 52 XP machines running on a bare-bones Hardware Virtualization using Red Hat, each OS would require 307 MB of Ram if it were dedicated, but as it's sharing memory, the end result would mean more Ram to use. Assuming having shared memory resulted in each OS feeling as though it had 512 MB of Ram, that's a savings of ~10 GB of Ram, or an efficiency of 166.4% (2/3rds greater) over dedicated Ram resources.

And that's just one kernel improvement. I wonder how Windows kernel is doing in comparison.

The 2.6.32 kernel also has scheduler improvements in its CFQ low-latency mode.

Some people have had issues with KDE's Akonadi, getting a "Not Registered At DBUS" message. From my limited experience in dealing with this error, it generally only happens when trying to use an external SQL database. I believe that this is an upstream issue, as Akonadi is constantly being worked on during this time and big things are expected of it. (This has been tested on a KDE 4.3.5 and 4.4.2 system) The internal database doesn't seem to have this issue; it works at least, though sometimes produces a warning message. It may take a few more releases before this bug is dealt with,

In one Fedora forum, it seems that using an external sql database works under the root user. I wouldn't recommend this route.

Others say that it's an AppArmour issue, that it is blocking akonadi. It seems odd, as I still got this issue using Gentoo which doesn't have that firewall installed. The developer team which worked on it has been layed off as well.

There are some issues with Gentoo's KDE Unstable 4.4.2 release, though that's to be expected.

Sometimes closing Kontact does not completely close it and attempting to run it again will have no effect. The process must be killed before it can be sucessfully launched again.

Kile (2.1 beta 2 & 3) seems to have an issue where closing the program results in an unexpected error minutes later or if the program is left open, unattended for too long. It doesn't seem to impact my work (autosave it on) and is more of an annoyance. Still, it's a great program for working in LaTeX.

Umbrello is a nice UML modeller program, but I get segfaults semi-regularly. I'm not sure why that is and I still use it. This doesn't seem restricted to the unstable KDE release, but I thought I'd point it out anyway.

Some final thoughts on KDE 4.4.2: It's faster, looks better, and has more functionality. I'm getting used to tabbed file browsing and I like where the KDE team is headed.

A special thanks to Aaron J. Seigo, a KDE guru and expert. This guy is real inspiration and I hope to do some work on KDE in the near future.

Powered by Blogilo

No comments: