Mahoney isn't thinking clearly
|
| Author | Content |
|---|---|
| cjcox Oct 13, 2006 12:17 AM EDT |
Regardless of what you think of reiserfs.... to say that a major weakness of reiser4 is the fact that you can't migrate easily from reiserfs v3 is just silly when the alternative is to migrate from reiserfs v3 to ext3 (which would be equally as difficult). So... the new SUSE.... let's see: 1. Dump many parts of YaST and do administration is some sort of adhoc fashion with disparate tools (like Red Hat). 2. Dump reiserfs for (Red Hat's) ext3 because (I guess) nobody understands it except Hans (of course, like Red Hat). 3. Get rid of KDE, use Gnome... eventually (really) it will have all of the features of KDE (again, just like Red Hat). Conclusion.... use the new SUSE which is now close to a Red Hat clone... or... use Red Hat. ... sigh... And PLEASE, don't mention Novell's anti-Linux, anti-normal, anti-simple enterprise platforms stacks. Sheesh.... messy, ugly, difficult, unreliable, bugridden, etc, etc, etc,... Is it too late to fire Nat and Miguel? Couldn't Novell go back in time to a version of SUSE that was stable, reliable, dependable and wasn't trying to be a Red Hat clone (cloning ONLY the weakest parts of Red Hat btw)? I've never wanted slap somebody so hard in all of my life!! My recommendations: 1. Get rid of Zen (the new update mechanism). It would have been better and CHEAPER to extend YaST's update mechanism. Zen is VERY, VERY, VERY, VERY slow. Reminds me of how slow YOU was many releases ago. The reason for the switch to Zen is so SUSE can track your licensing better (and AGAIN, to be JUST like Red Hat). 2. Sell the pro-Red Hat, pro-Gnome Nat and Miguel to Red Hat. Shoot... just give them to Red Hat. IMHO, they already work for Red Hat. 3. Put some real effort into an enterprise level filesystem. If that's ocfs2 (or variant thereof, fine). Look at Sun's zfs for some neato features, and don't discount reiser4, even if you hate it... at least look at what it was trying to provide. And ext4 isn't an answer... ext4 will be as good as reiserfs v3 (well... probably not). 4. Stop work on Xgl/Compiz. I'm sorry, but the world is moving much faster and anybody wanting the eye candy has already moved on to things like xorg 7.1 and Beryl. 5. Got to figure out multipathing folks. SUSE could show up the big boyz by developing a reliable multipathing implementation. 6. Get Xen working well.... but beware of the beast in Redmond. 7. Stop focusing so much on the desktop .... SUSE works quite well on the desktop already. 8. Do Gnome... fine... BUT don't push everything into it at the expense of a good KDE environment. If they spent half the time they spent on Gnome on KDE, the KDE SUSE desktop would be light years ahead of Gnome today. Nothing is very broken with KDE, just like nothing was very broken with Gnome... it wouldn't take much to make KDE menus more "usable". 9. We need an enterprise level distributed filesystem. If that's NFSv4, fine. Regardless, a good distributed filesystem is essential for the enterprise (NFSv4 may be the best/only choice just for integration reasons alone). 10. Develop a reasonable distributed authentication system (the answer to this is NOT LDAP... though the solution may be LDAP based). 11. Make Autoyast2 as easy to use as Red Hat's kickstart. Right now the tools for Autoyast2 do a very, very, very poor job and nobody wants to edit a ton of XML (Autoyast2 was certainly implemented by an engineer...). 12. Packages. Don't restrict SLES (the expensive product) by pulling out the SLED elements, let me decide if I want my server to be workstation like. Also, SLES/SLED even combined has thousands fewer packages than openSUSE (hint for anyone wanting to "upgrade" to enterprise SUSE). Areas where SUSE still excels (vs Red Hat): 1. YaST (though Miguel and Nat are doing their best to destroy it) 2. hwinfo (makes kudzu look like a joke). 3. LVM (Red Hat only very recently figured that LVM was a good thing) 4. reiserfs (love it, hate it... it's growable on the fly, even the most recent version of ext2/3's online resizer is risky at best and has interesting limitations). 5. init scripts, network handling (Red Hat's implementations of startup and configuration make WAY too many assumptions and just aren't as flexible... SUSE appears to have put in a whole lot more enterprise thought into things). 6. Samba (as much as I hate it.... I suppose SUSE does the best job of dealing with it). 7. Installation (Kickstart is nice, but shoot, you can install SUSE a gazillion different ways.. you have much, much, much better package controls... it's just better thought out... with the notable exception of Autoyast2). 8. Apparmor (makes application firewall protection EASY). 9. Ability to run and administer WITHOUT a full GUI (Most all of Red Hat's admin tools require a GUI to run... SUSE is better designed to go headless via just a serial console, one of Red Hat's biggest enterprise weaknesses... and even the, the Red Hat GUI tools are not all that great). 10. SNMP, SLP, Openwbem (more work still needed, and hey... how about wbem in openSUSE and not just SLES??). |
| tuxchick2 Oct 13, 2006 1:49 PM EDT |
A lot of people experience problems with Reiser, because it is flawed. I used to use it a lot on mailservers with maildirs, because it handled all those many bitty files well. But it's just plain not reliable. What good is all that speed and glitz when it falls apart under load? Ext3 is rock-solid, and when Ext4 is ready for prime-time the upgrade path looks like it will be painless, just like Ext2 to Ext3. In a production environment I need reliabilty, future-proofing, and stability way more than glitz. I've also done extensive testing on XFS and JFS, and I keep coming back to Ext3. It just works. I'm not too worried about speed- that will come with time as EXT3/4 continues to improve. As it is, in real-world use it performs just fine. Other factors affect performance a lot more, such as adequately provisioning hardware, tuning applications, and using apps that are well-written in the first place. Additionally, Hans Reiser does not help by being obnoxious and uncooperative. Flaming kernel devs and everyone else who says things he doesn't want to hear doesn't get the job done. |
| jimf Oct 13, 2006 2:03 PM EDT |
> it's just plain not reliable Sorry, but I can't agree with that. While I wouldn't use reiserfs for a major server, for the average desktop user reiserfs3 may very well be the best choice. Note that Suse's reason for dropping reiserfs is not unreliability, but rather, scalability, a feature which will become more important as we approach terabyte size HDs. |
| tuxchick2 Oct 13, 2006 2:26 PM EDT |
There you go impinging my freedom again. Dang, I slay me. You're right about the scalability reason, I missed that. |
| 1c3d0g Oct 13, 2006 4:18 PM EDT |
But ReiserFS is hardly the pinnacle of stability...so I wouldn't rule that out either (they may acknowledge that behind the curtains, so to speak). |
| jimf Oct 13, 2006 4:38 PM EDT |
1c3d0g, That seems to depend on who's talking and what the're using the system for. What I've seen in the last 3-4 of years doing support for Mepis and Debian with many many desktop users is that reiserfs3 is entirely stable for the desktop user. I can't remember anyone reporting a problem with it. I've used Reiserfs3 for that whole period, and I've never lost a bit of data. When first introduced, reiserfs3 did have problems, but we certaintly don't see that now. I suspect that some are still thinking of reisers early problems when they talk about it's 'instability'. It may be that some got burned by that earlier version, have not tried it recently, and, are still holding a grudge. |
| dinotrac Oct 13, 2006 5:37 PM EDT |
I've also been very pleased with Reiserfs on my desktop systems and am concerned for the future of Reiser4. I would bet that my desktop is not all that unusual, with bazillions of little files and enough big ones (video editing, music recording) to notice. Reiserfs swallows it up like a champ and I've never had trouble. |
| jimf Oct 13, 2006 10:04 PM EDT |
> concerned for the future of Reiser4 the concept of Reiser4 is absolutely brilliant. No other FS I've seen offers the same potential for an advanced desktop file system. Apparently, it is working reasonably well, but, reasonably well doesn't make it with a major file system. It will be some time before it's debugged and bulletproofed. It would have been nice if Reiser had been able to add 4 support to the kernel before his current personal fiasco. That would have given the project a lot more credibility and possibility of continuance. Reiser's personality flaws were certainly a negative in that endeavor. On the other hand Reiser isn't the only developer involved, and, we'll hope that the others involved in the project will be able and willing to debug and finalize Reiser4. I think that everyone hopes that Mr. Reiser's problems are not of his making, and, many of us would also like Reiser4 to become solid and reliable. Right now, I'm afraid that's all in the wind. |
| herzeleid Oct 13, 2006 11:14 PM EDT |
Quoting: jimf: While I wouldn't use reiserfs for a major server, for the average desktop user reiserfs3 may very well be the best choice.Well, we *do* use reiserfs for major servers. All of the shops I do business with are using suse linux servers, and every last one is running on pure reiserfs. Of the production linux servers at my primary employer, none has ever crashed, and they all had around 400 days uptime until we recently did kernel upgrades.. We recently installed the OS on a new server (Quad 64-bit Xeons, 16G RAM) and did some quick and dirty comparisons to see what the relative performance was with different filesystems. As it turns out, the old ext2 filesystem was the fastest, but that's a lean mean sort of filesystem without journalling. We also looked at ext3, jfs, xfs and reiser, which consistently gave the best performance results over a wide range of load. Regardless of the political situation, reiserfs is a good solid fs, and we haven't seen anything whatsoever until now, that would warrant a change. If distros move away from reiser as a result of the politics involved in distancing themselves from bad publicity, we'll naturally have to pick an alternate. Problem is, none of the current choices are very palatable in terms of performance. Possibly veritas, we'll have to check out the performance on that. |
| Sander_Marechal Oct 16, 2006 2:48 PM EDT |
Quoting:We also looked at ext3, jfs, xfs and reiser, which consistently gave the best performance results over a wide range of load.I'd be interested in those numbers. AFAIK there's only one other decent comparison of different filesystems. A potential LXer feature maybe? |
| herzeleid Oct 16, 2006 4:46 PM EDT |
Quoting: sander: I'd be interested in those numbers. AFAIK there's only one other decent comparison of different filesystems. A potential LXer feature maybe? Yeah, maybe I'll whip up some graphs and put it out there on that interweb thingie, when I get some time, if there's interest. |
| dcparris Oct 16, 2006 6:35 PM EDT |
Yeah, herzeleid! You could write it up and post it on LXer as a feature story. Hint! Hint! |
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!