Monday, June 09, 2008

coding to save the environment...

Each one of us has ways to help save the environment. Interestingly, there are even some options exclusively available to coders ;)

I was made aware by Michael Biebl some time ago that the Linux community is trying to save power by disabling unnecessary wakeups inside the softwares. For example, the Linux kernel has become mostly tickless, that is does not necessarily wake up frequently. This is important, because modern CPUs have power save modes: simply said, the less is computed, the less power is needed. An idle CPU uses least power. However, when an application wakes up, the CPU resumes into modes that need more power. Of course, the CPU does not know if what it needs to be executed is actually useful or could be done more smartly. You can find more details at the site

But why am I telling you this? The reason is that rsyslog, too, had (and unfrotunately has) some wakeups which are not strictly necessary. For example, the current imfile implementation needs to poll input files. This is the most portable way to do things, but it is not power-effcient. An update to a more power-efficient solution (on the platforms that support it) is scheduled. Next, the --MARK-- message generator runs periodically. This one can probably never be avoided, because the actual functionality is the periodic activity. Any user should think if the functionality is neeed and, if not, not enable this plugin.

Finally, and the core rsyslog awoke every 30 seconds to check if repeated messages need to be flushed. However, repeated message reduction is no longer the default and I assume that most people do not use it. The default is to have it turned off. But even when it was turned off, rsyslog still awoke to do the (unnecessary) check. I have changed this, rsyslog now only initiates the wakeups when they are actually needed. To many (if not most) environments this means rsyslog will never again wakeup the system if there is no need to. It's a small saving, but every little bit counts toward the ultimate goal. And given the potentially large number of systems running rsyslog, the saving may actually be larger than one initially assumes.

In the future, repeated message reduction will probably be implemented based on a different paradigm, which enables us to not run on a tick even when it is turned on.

In any case, I hope the new feature will contribute to - you name it - slightly longer notebook battery life and less environent pollution. And, ultimately, it's a interesting fact to see how a software developer can affect environment health these days ;)

The new functionality will be available starting with rsyslog 3.19.7.

1 comment: said...
This comment has been removed by a blog administrator.

Introducing new team member

Good news: we have some new folks working on the rsyslog project. In a small mini-series of two blog postings I'd like to introduce the...