Arch Alternatives

sjmorgan · 2007-02-06 20:42:14

Hi,

Can anybody recommend any alternatives to Arch? I've been using it a while now and really like it; the only problem is pacman. I'm really sick of having to wait literally 2 minutes to perform a search or install a package when the "database" isn't cached in memory.

I feel a bit weird posting this on these forums (I hope nobody thinks I'm trolling) but I figure other Arch users will have a good idea of distros that are similair in concept. The main things that appeal to me are the up-to-date packages and KISS philosophy so anything that fits that bill I'd be interested in hearing about.

Thanks.

noriko · 2007-02-06 20:44:42

well there is slackware. but it's not exactly similar, it's just almost as fast.

have u actually made any attempt at improving that speed ?
http://aur.archlinux.org/packages.php?d … hans=&SeB=
pt-pacman-cage from pactools ?

cactus · 2007-02-06 20:49:27

Just for my own curiosity, what file system do you use? Reiserfs perhaps?

As to your question, my distribution usage depends largely on the need that I am trying to meet.
If arch is out of the question, then my choices would probably be...

Server: debian, slackware, or maybe a BSD.
Workstation/Desktop: ubuntu or debian

sjmorgan · 2007-02-06 20:52:56

noriko wrote:

well there is slackware. but it's not exactly similar, it's just almost as fast.
have u actually made any attempt at improving that speed ?
http://aur.archlinux.org/packages.php?d … hans=&SeB=
pt-pacman-cage from pactools ?

Slackware doesn't have up-to-date packages though, as far as I'm aware. I'm also assuming it still has that really crappy package manager.

As for pacman-cage: I've read a bit about it before; I'd just rather use something that isn't inherently flawed than have to resort to workarounds.

sjmorgan · 2007-02-06 20:55:50

cactus wrote:

Just for my own curiosity, what file system do you use? Reiserfs perhaps?
As to your question, my distribution usage depends largely on the need that I am trying to meet.
If arch is out of the question, then my choices would probably be...
Server: debian, slackware, or maybe a BSD.
Workstation/Desktop: ubuntu or debian

I use ext3 for everything.

I switched to Arch to get away from Ubuntu/Debian . I don't like "release cycles" (again, I'm looking for up-to-date packages), dpkg is way too complicated and bugs had a tendency to sit around in their bugzilla for months without any action.

baze · 2007-02-06 21:18:23

ext3 for /var makes pacman extremely slow. either use sth like pacman-cage or use reiserfs for /var (which i do) and pacman will be fast again.

phrakture · 2007-02-06 22:13:09

pacman 3 is faster in this regard. Still there is the whole filesystem issue, with having to read hundreds of little files (this is why reiser helps).

I've said hundreds of time to similar issues: if you think pacman is "inherently flawed", please provide a solution or patch and we can work on it. For me, a search takes maybe 15 seconds the first time I run it (i.e. the files are not cached) and then 2-3 seconds thereafter.

sjmorgan · 2007-02-06 23:57:58

phrakture wrote:

pacman 3 is faster in this regard. Still there is the whole filesystem issue, with having to read hundreds of little files (this is why reiser helps).

How much faster, may I ask?

I've said hundreds of time to similar issues: if you think pacman is "inherently flawed", please provide a solution or patch and we can work on it. For me, a search takes maybe 15 seconds the first time I run it (i.e. the files are not cached) and then 2-3 seconds thereafter.

Maybe if you've heard it stated a hundred times, there may actually be a problem? Or maybe I'm just imagining it...

$ time pacman -sS pacman
<snip>

real    1m43.501s
user    0m0.712s
sys    0m0.748s

Nope.

As for solutions, how about not using hundreds of little files? Maybe some kind of database like sqlite if you don't feel like coding your own?

phrakture · 2007-02-07 00:22:20

pacman 2:

$ time pacman -Ss pacman
...snip...
real    0m1.206s
user    0m0.760s
sys     0m0.444s

Sounds like you have a filesystem problem. This is on an ext3 partition, with nothing fancy done to it (no dir_index or any of that junk). That is cached, however, the first run after a period of inactivity would probably take 5-10 seconds on this box.

I've stated this hundreds of times because (now this is important) I don't have a problem. I did nothing fancy, and I don't even have the "proper" filesystem, and this takes 1-2 seconds.

phrakture · 2007-02-07 00:28:06

sjmorgan wrote:

I've said hundreds of time to similar issues: if you think pacman is "inherently flawed", please provide a solution or patch and we can work on it. For me, a search takes maybe 15 seconds the first time I run it (i.e. the files are not cached) and then 2-3 seconds thereafter.
Maybe if you've heard it stated a hundred times, there may actually be a problem? Or maybe I'm just imagining it...

Now here, you've missed my point. My point is this: regardless of what application you're complaining about, it helps no one involved when one complains. It helps to do. There is nothing negative that come from that course of action. It seems to me I'm saying this quite a bit recently: there is a plethora of ideas and opinions in the Open Source world. Opinions do not make software (sadly). Effort does. This is not about you, this is nothing personal. It just saddens me when I see so many people point fingers from the sidelines.

sjmorgan · 2007-02-07 00:40:41

phrakture wrote:

I've stated this hundreds of times because (now this is important) I don't have a problem. I did nothing fancy, and I don't even have the "proper" filesystem, and this takes 1-2 seconds.

If you don't have a problem, then why do you feel the need to keep stating it? Do you just like to insert it randomly into conversations? No, I'm assuming you have to keep repeating yourself because people are telling you there's a genuine problem but you'd rather ignore it than fix it. I've never understood the assertion that, because something works for one person then it must follow that there's not a problem. The logic defies me. If you go to a mechanic to get your car fixed, do they tell you "mine works fine" and send you on your way? If you go to the doctor with a pain in your leg do they simply state "works for me", and dismiss you?

If you can't or won't fix it, that's fine - it's your time and energy - but please stop acting like there isn't a problem.

bangkok_manouel · 2007-02-07 00:56:11

Did you run pacman-optimize to check if it helps ? Furthermore, using ReiserFS for /var should help a lot.

HTH

sjmorgan · 2007-02-07 00:58:54

bangkok_manouel wrote:

Did you run pacman-optimize to check if it helps ?

Yeah, it doesn't unfortunately.

phrakture · 2007-02-07 01:42:07

Again, you're missing the point. Let me explain, but first:

sjmorgan wrote:

If you go to a mechanic to get your car fixed, do they tell you "mine works fine" and send you on your way? If you go to the doctor with a pain in your leg do they simply state "works for me", and dismiss you?

Those are trained, paid jobs. We are talking about something that most people here deem a hobby. If I went to a hobbiest mechanic or doctor, and spent 0 dollars, I would be completely fine with whatever they told me. That, however, is all besides the point.

sjmorgan wrote:

If you can't or won't fix it, that's fine - it's your time and energy - but please stop acting like there isn't a problem.

Re-reading the thread, I did not indicate there is not a problem. In fact, I even suggested (as many others have) that this is a filesystem problem. What I DID suggest, however, is that you put forth some effort into fixing the problem. I do everything I can, but I don't have your setup or your hardware. I already stated that this is about effort, not opinions. Show me some work. Show me how to reproduce this. Show me what I can do to fix the slowdown on your machine. This is NOT about you. You seem to retort as if I mortally wounded you. That is not so. I never once said you were wrong - I just want you to help out. No one here is paid. Most of the developers don't really even get the credit they deserve ("Hey Xentac, nice work on srcpac and namcap!", "Hey tpowa, nice work on these ISOs", etc).

Everyone here is exactly as human as you are. Tell me, in a real-life circumstance, which is a better act:
* Laughing at a fat kid.
* Coaching the fat kid on how to eat better.

Curiously enough, it's the one that requires effort.

So please, let's do something about this, complaining (laughing at the proverbial fat kid) gets us no where. What are your system specs? Hard drive? hdparm -tT output? File system. Have you tried dir_index-ing the filesystem? What have and haven't you tried. Do you have anything disk related in your dmesg output?

stonecrest · 2007-02-07 02:08:42

I just booted up and ran pacman -Ss test, this is on a ext3 filesystem (no dir_index or anything special here either). It took 5.3 seconds. It seems like your problem is more than just pacman.

xabbott · 2007-02-07 02:29:36

Just for comparison I have /var on a rfs partiton.
/dev/sda4 /var reiserfs noatime,notail 0 0
"cold" search

[~]$ time pacman -sS pacman
real    0m21.033s
user    0m0.534s
sys     0m1.320s

The same search about five minutes later.

[~]$ time pacman -sS pacman
real    0m0.875s
user    0m0.469s
sys     0m0.400s

mrunion · 2007-02-07 02:36:53

I ran pacman after a reboot and timed it on ext3 filesystem:

real     0m18.361s
user    0m0.444s
sys      0m1.052s

iBertus · 2007-02-07 03:13:12

sjmorgan wrote:

$ time pacman -sS pacman
<snip>

real    1m43.501s
user    0m0.712s
sys    0m0.748s

Mind sharing some information about you setup? Seems pointless for anybody to try and help with such a small amount of information. What kind of hardware are you running? How old is the install in question? Is /var mounted with any strange options? Please try and change the filesystem configuration and see if it improves. If so, odds are that it's not a pacman bug. But, if it is a pacman bug we have to figure out how to reproduce it and fixing it manually on your system seems like a logical first step.

deficite · 2007-02-07 03:18:03

Wow, I just ran a time on pacman and I got a weird result:

real    0m42.758s
user    0m0.827s
sys     0m1.192s

I think that's just because my computer has been sitting idle for quite some time and maybe things have been swapped over. I use ext3 and usually the uncached time is about 20-30 seconds and after that is < 2-3 seconds.

Trust me, you are not the first person to complain about this issue. Search the boards and you'll find several band-aids for this. A while ago I remember people talking about moving to a database system such as SQLite. The problem is the fact that your hard drive has to seek around to find a bunch of small files on your hard drive. It's not really pacman's fault, it's the filesystem's fault. I agree with what phrak has to say, but most people didn't want to go to SQLite or something to the like. That would have the possibility to somewhat alleviate this issue (perhaps, I've never tried it).

I'm really optimistic about libified pacman though. I think you'll see some frontends come out that will provide some nice features. I haven't looked at any of its code but hopefully it's designed as to where one could use feature x and y of pacman and not z (I would think it would be designed this way). If so, one could make a frontend that stores the package database in SQLite or the like. Are there any doxygen's on the web for pacman3 yet? I've been interested in writing a frontend for quite some time but I've been holding off on pacman3. Or do I have the wrong idea of what libified pacman is? I'd love for it to handle everything but searching its database (I would make my own implementation of that). Phrak says the speed is going to improve. If enough so, I'd just stick with the officialness.

Anywho, thanks a ton dev team! I'm glad to know that pacman3 is coming along.

Last edited by deficite (2007-02-07 03:20:03)

iBertus · 2007-02-07 05:50:05

How often do people really update? I'd rather spend 1 minute waiting for pacman to sync then 45 minutes waiting for firefox to build from sources. Type pacman -Syu and go get a cup of coffee and come back. Done.

mucknert · 2007-02-07 07:05:58

Here is a little brain-teaser for sjmorgan:

if pt-cage would be integrated into pacman and be the standard-method of storing the package-information, does that mean that the inherent flaw is no more? For me, and I believe everyone else, it would be. Problem solved! What do you think app like SQLite would do? Packing all the stuff in one huge file that can be searched with SQL-Syntax (in the case of SQLite a rather mediocre one) and be done with it.

The Problem (I agree that it is one) lies in the way a lot of files are handled by the filesystem. It is because the package-list of pacman grew so big so fast so that it gets cluttered all over the filesystem and the read/write-head has to move around a lot. Putting all into one file with caging is not only a workaround: it is a nice fix until someone (Like you, maybe? Oh noes, you prefer to whine about it!) creates a patch that gives pacman the ability to use a Database. It works for me, it works for anyone else. Granted, it is a design-issue in pacman but it is easily resolved. All I see is that you are anal about pacman being "inherently flawed" (which is a true assertion) and that you just refuse to aknowledge known fixes (which is stupid). To give a real-word exapmle: the concept of the automobile is "inherently flawed" bacause you can have an accident with it. Still you don't see people going around and say they should be scrapped completely. Nope, they just use seatbelts and airbags.

End of rant.

bones · 2007-02-07 08:05:40

No problem here, quite happy with the results

time pacman -Ss pacman

real    0m33.830s
user    0m6.640s
sys     0m9.949s

I use ext3 on all my partitions

iphitus · 2007-02-07 08:35:21

There's simpler methods of dealing with this than pacman-cage.

A quick way to force a cache on boot, is to put

pacman -Ss &> /dev/null &

into your rc.local.

I hear a lot of people recommending reiserfs, others recommend ext3. Regardless of what filesystem you use, the same problem exists. The fault is not at filesystem level, but disc level. -- the files will be written across the disc. One may start off quicker, but over time both and all will degrade.

My workaround is to put /var, or even just the pacman db, on a seperate smaller partition. Like pacman cage, but just a tad more reliable. If you don't have the hdd space, chop 100mb off your swap and make it a new partition.

As for distro's like arch but without pacman? I like *bsd. They're tidy, clean and excellently documented. Though for NetBSD and OpenBSD, their package management is weaker, albeit simpler. FreeBSD I havnt had as much experience.

James

Mikko777 · 2007-02-07 08:46:36

I have an old laptop pacman searches still take only couple of seconds...

But ofcourse phrakture and pacman sucks if he cannot code core2 quad and T1 line performance outta P2 and 14.4k modem

zeus · 2007-02-07 08:51:14

pacman-optimize IMHO enough

[zeus@gefest ~]$ time pacman -Ss pacman
real    0m0.944s
user    0m0.188s
sys     0m0.261s

Arch Linux

#1 2007-02-06 20:42:14

Arch Alternatives

#2 2007-02-06 20:44:42

Re: Arch Alternatives

#3 2007-02-06 20:49:27

Re: Arch Alternatives

#4 2007-02-06 20:52:56

Re: Arch Alternatives

#5 2007-02-06 20:55:50

Re: Arch Alternatives

#6 2007-02-06 21:18:23

Re: Arch Alternatives

#7 2007-02-06 22:13:09

Re: Arch Alternatives

#8 2007-02-06 23:57:58

Re: Arch Alternatives

#9 2007-02-07 00:22:20

Re: Arch Alternatives

#10 2007-02-07 00:28:06

Re: Arch Alternatives

#11 2007-02-07 00:40:41

Re: Arch Alternatives

#12 2007-02-07 00:56:11

Re: Arch Alternatives

#13 2007-02-07 00:58:54

Re: Arch Alternatives

#14 2007-02-07 01:42:07

Re: Arch Alternatives

#15 2007-02-07 02:08:42

Re: Arch Alternatives

#16 2007-02-07 02:29:36

Re: Arch Alternatives

#17 2007-02-07 02:36:53

Re: Arch Alternatives

#18 2007-02-07 03:13:12

Re: Arch Alternatives

#19 2007-02-07 03:18:03

Re: Arch Alternatives

#20 2007-02-07 05:50:05

Re: Arch Alternatives

#21 2007-02-07 07:05:58

Re: Arch Alternatives

#22 2007-02-07 08:05:40

Re: Arch Alternatives

#23 2007-02-07 08:35:21

Re: Arch Alternatives

#24 2007-02-07 08:46:36

Re: Arch Alternatives

#25 2007-02-07 08:51:14

Re: Arch Alternatives

Board footer