Not very useful
|
| Author | Content |
|---|---|
| richo123 Mar 07, 2007 5:40 PM EDT |
Benchmarking a small set of applications in a pretty uncontrolled fashion without proper in depth analysis and getting results which are not that different is really a waste of time imho. |
| Aladdin_Sane Mar 07, 2007 9:09 PM EDT |
I gave it a thumbs up as a "good first try." My biggest complaint is that the author used Ubuntu instead of Debian. RH Fedora is in no way derivative as distributions go. Ubuntu is a derivative of Debian. For "apples v apples" comparison I think that derivative should be against derivative, and non-derivative against non-derivative. For example, Ubuntu v Knoppix, or RH Fedora v Debian. |
| swbrown Mar 07, 2007 9:19 PM EDT |
The danger of this kinda thing is that it can be like benchmarking MySQL in non-strict mode with the non-ACID DB backend vs PostgreSQL - you could easily declare the speed winner to be the one that skips saftey. In a distribution context, that means the one that doesn't compile everything with -fstack-protector and use selinux. Debian currently doesn't require everything be built with -fstack-protector, it's left up to the maintainer to decide if it's worth it. Personally, I think this is a mistake, as a lot of these bugs crop up in applications that aren't traditionally considered security sensitive, like mp3 players, but it's the way things currently are. Also, Debian's selinux support is lacking compared to RedHat's - Debian's reference policy rules had gone somewhat stale and are just recently being brought back up to date. With both, you're probably looking at around a 3% performance loss. Other than that, you're basically going to have the same speed. There's not going to be significant difference when everyone's using the same kernel, same compiler, and often the same packages. |
| Aladdin_Sane Mar 07, 2007 9:31 PM EDT |
>There's not going to be significant difference when everyone's using the same kernel... I see your point. But major non-derivative dists (incl. both Debian and RH Fedora) custom compile features in or out of their "official" kernels. Come to think of it, it really is hard to get apples v apples. For example, does one compare Current RH Fedora (FC6) v Current Debian Unstable, or Stable? Since these performance differences are mostly about game performance, do you use closed-source manufacturer drivers, or the open-source drivers? Really, I like it better when selinux and other security settings are optional, and off by default: I simply do not have the same security requirements as the NSA. |
| dcparris Mar 07, 2007 9:39 PM EDT |
I don't know about benchmarks, but in my experience, Debian Etch blows Ubuntu out of the water, speed-wise. Well, maybe not quite out of the water, but definitely up on the beach. ;-) |
| dinotrac Mar 08, 2007 3:05 AM EDT |
>Come to think of it, it really is hard to get apples v apples. For example, does one compare Current RH Fedora (FC6) v Current Debian Unstable, or Stable? You're going a bit overboard in defining your apples. Comparing A + A from another vendor is not the point of benchmarking. Comparing potential solutions is. In that vein, you might compare a mainframe solution with a Unix solution, or just about anything else that is being considered as a solution to your problem. Fedora v. Ubuntu is fine. |
| tuxtom Mar 08, 2007 9:57 AM EDT |
I have to agree that this is pretty useless. Comparing the two distros with the "default" installations? What would contitue the "default". There are plenty of options that differ with both. Both will have some different proceses running, etc., and with a given set of tests, either could be tuned to out perform the other. |
Posting in this forum is limited to members of the group: [ForumMods, SITEADMINS, MEMBERS.]
Becoming a member of LXer is easy and free. Join Us!