“Seems the new version of Samba ups the ante in file/printer sharing options for Windows network, with a 2.5x performance gain over Windows Server 2003, according to tests by IT Week.” Read it here.
Samba 3 Benchmarked Against Windows 2003
About The Author
Adam Scheinberg
Technology Executive • Web Developer • Father • Foodie • Music Snob • OS enthusiast
Follow me on Mastodon @[email protected]
I like samba because it s so easy (if setup for workgroups, PDC is much harder) and fast but I would like to see the numbers.
This article alludes to testing done by IT Week yet it’s not on the IT Week servers. And it doesn’t have any numbers
This is horrible, with no proof of benchmark scores, it was probably funded by microsoft so they can refute these claims with some “real” false results.
Anyway this isnt cool, where are the numbers and proof? this is worse than a MS study.
Good to hear, but without numbers the result is useless.
http://www.itweek.co.uk/News/1144312
but still no hard numbers. Maybe they’ll be in a publication.
Here ( http://www.itweek.co.uk/News/1131114 ) is an older article about Samba knocking the stuffing out of Win2K and mentioning that Microsoft declined to comment on the results.
Not only does it not give us numbers, more importantly it did not tell us which platform Samba was run on. Solaris? Linux? BSD?
over on slashdot people who have a subscription to the paper version said there are numbers in the paper version.
Anyone remember the previous IT Week article (http://www.itweek.co.uk/News/1131114) comparing Samba 2.x to Windows 2000, making just as outrageous claims and still no numbers to back it up after over a year?
There is a scant amount of information on the actual tests performed in this article: http://www.itweek.co.uk/News/1144312
According to the guys at Slashdot, the print version of the article (with the numbers) is available in ebook format from:
http://www.itweek.co.uk/ebook
I’d download it, but their crap software only runs on Windows and Mac. Anybody wanna try and get those numbers? I think they’ve got a free trial issue offer.
There are a couple of things to keep in mind about this article:
1) ITWeek is hardly Linux-biased. They’re your average IT rag, and I doubt they have any interest either way. I suppose its not out of the question that IBM bribed them to do the article, but IBM’s Linux department doesn’t have a history of doing this, while Microsoft does.
2) They are testing the next-generation Samba (3.x). The old Win2k3 tests are done against the previous generation Samba (2.x). Thus, its entirely imaginable that the performance difference is real. I run a Linux server at home with exported samba shares, and I can tell you that its a whole lot faster than any of the WinXP machines at sharing ifles.
3) The SMB protocol sucks goats. I’m surprised that in this day and age, something so crappy is still the default shared files mechanism in Windows. I’ve seen SMB connections fail for absolutely no reason, even though both computers could ping each other. And “My Network Places” always freezes up for several minutes while discovering new connections in the workgroup. If only XP came with an FTP server by default! Hmm, kinda got off on a tangent there. I’ve got half a dozen machines sharing using SMB at home, and the protocol gives me nothing but trouble. Anyway, SMB sucks so bad, I wouldn’t be surprised if you saw a 2x increase every product generation. God knows that the protocol doesn’t have a hope of saturating the network link…
I hear nobody arguing about ACL issues. Defaulkt W2K3 comes with NTFS, ACL’s enabled. Samba 3.0 comes with share level permissions but no file-system ACL’s. My guess is this should be at the very least one of the causes for a performance difference.
But, as previous posters said, there is nothing but marketing speak in this article and no figures. I would be surprised to read they even considered the ACL issue…
By Rayiner Hashem (IP: —.res.gatech.edu):
“And “My Network Places” always freezes up for several minutes while discovering new connections in the workgroup.” – That can be easily corrected. There is ms publication about it.
By Serge van Ginderachter (IP: —.kabel.telenet.be): I compiled samba with –with-acl-support. About publication: That is the whole point. There is not much about OS/software config. Unless they provide all data this is nothing but noise
I don’t care if Linux is faster… I care about TCO. My enterprise relies on B2B and CRM technology to achieve maximum ROI, and Microsoft allows for this by helping us meet our strategic goals and expand our reach with e-bussiness. .net technology is just what America needs to get out of its recession.
Windows XP: non-free as in supporting the economy
Linux: free as in communism
Thats a wonderful statement full of meaningless buzzwords, but how about you restate it with actual useful things? Like “I can do X with Windows Server, but cannot do X with samba”.
All you have managed to do is come off like your stuck in the sales department.
Isn’t is fitting to see how “Slammer” is an acronym for “MS Lamer”…?! *LOL*
It’s not an acronym. It’s an anagram.
It was funny to read your post… In a meeting the other day, one of the director’s in my company said something alongs the lines of this…
“I don’t want to hear ROI and TCO garbage, we live in a real world…”
They look pretty on paper, and make the execs think they are getting some great deal on something, but usually, in the real world, the ROI and TCO savings are usually not there…
Windows *: Capitalist Greed
Linux : New choice of freedom from Capitalist Greed
Uhh.. you are right. Still though…
… As someone pointed out w/ ACLs … but also
I’d like to know what the FS is under the Linux numbers, after going through nastyness to recover my ext2 FS recently, I’m grateful for the Journaling of NTFS file meta-data (though not the data, beating everyone to the NTFS isn’t a real Journaling FS punch
. They should make sure a FS w/ approximately the same feature set (Ext3/Reiser/What’sThatOtherJournalingFS?) is used to store the data.
“I don’t want to hear ROI and TCO garbage, we live in a real world…”
And in the real world TCO is very much important. Making your computing systems or software purchase decisions without evaluating the total costs of the product is short sighted and foolish. Failure of a mission critical system or software package can result in downtime which costs far more than the investment in the system or software.
…by looking at all the MS apologists attacking the study. I’m sure the numbers will come out at some point – but then the apologists will surely deny them, saying the study is biased. Seems the only thing that convinces them is when the study is paid for by MS…
I don’t want to troll, but seriously, why so much suspicions towards the results of an independant study, published on a non-aligned web site? If the study had been commissioned by IBM and had been published on Linux.com, I might understand the outcry, but here it simply sounds like a lot of bad faith on the pro-MS side. The possibility that Samba might beat Win2K3 isn’t even considered, and indeed is almost instantly brushed off…
Sometimes I do wonder if there aren’t some MS marketing people posting on this site…
I’m grateful for the Journaling of NTFS file meta-data (though not the data, beating everyone to the NTFS isn’t a real Journaling FS punch
Journaling filesystems typically do nothing to ensure protection of data in the event of a power loss. In fact, XFS contained a bug in which metadata operations would be journaled before the write buffer cache was committed, so if a power loss occured during the window between those two events any files with data in the buffer cache during the power outage would be filled with zeroes after the journal was replayed… nasty. Journaling isn’t all it’s cracked up to be… it has many negative consequences.
I don’t want to troll, but seriously, why so much suspicions towards the results of an independant study, published on a non-aligned web site?
The study gives somewhat outrageous figures (Samba performing 150% better than Windows Server 2003) for operations which are primarily I/O bound by the network and disks of the server and then does not give the actual figures of the benchmark itself, only analysis of the results.
In practice, the journaling filesystems in Linux are usually competitive with ext2 in terms of performance. Interestingly, Reiser4, which offers full data integrity through atomic transactions, is also one of the fastest filesystems on Linux.
>>Isn’t is fitting to see how “Slammer” is an acronym for “MS Lamer”…?! *LOL*<<
Good one! But, I think you meant: anagram. I can hardly believe anybody follows the msft load of BS as closely as the this “ms lamer.”
The study gives somewhat outrageous figures […] and then does not give the actual figures of the benchmark itself, only analysis of the results.
Actually, I think what you mean to say is that the article doesn’t give the actual figures, but an analysis of the result. Unless you’ve actually seen the study…
This is big news. The study itself is bound to become available (I think it already is for paying customers?).
Even if it had the same level of performance, it would still be advantageous to use samba servers instead of Win2K3 due to the price advantage.
I care about TCO. My enterprise relies on B2B and CRM technology to achieve maximum ROI…
Are you quoting a Windows Server 2003 advertisment or something? Extreme overuse of acronyms IMO.
Regardless, while you are babbling away I recently purchased Samba 3.0 for less than Windows 2003. Actually it wasn’t much of a purchase as the product is free. As for TCO? Well, you do need a computer to run it on.
You want the maximum return-of-investment, and yet you don’t care about speed? And when did “Linux” come into this? The article is about Samba 3.0. Do you have any idea of what you are talking about, or do you just enjoy using financial buzzwords?
Ok, this is one o the best comments I’ve read in a while. Too bad others don’t seem to notice the sarcasm.
Do more with less!
these are definately network io bound tests. i have benchmarked 17MB/s between xp machines over firewire (using modest hardware). now unless they are using gigabit, i see no reason to believe that EITHER os is incapable of completely saturating a 100Mbps link.
i would be greatly curious as to exact test methodology. i dont think anyone on either side of the preference fence can take these claims seriously until we see that information.
…on my small home LAN, I did find that transferring a 117 MB file from my Samba2 server to a Win98 workstation took about 55 seconds, while with Samba3 it took only about 38 seconds. So there seems to be a definite improvement between Samba 2 and 3, at the very least.
This is most effective when you are using SCSI controllers with battery-backed cache modules. This way the data in the cache buffers is backed until power is restored and the data can be flushed to the disk. In this situation the journal will not cause file corruption.
I think journaling on systems without critical data is not only pointless but causes performance problems. If Linux distributions would at least default to ext2 for normal installations and ext3 (or reiserfs, xfs, jfs, etc) on servers, I think it would be a good thing.
“I think journaling on systems without critical data is not only pointless but causes performance problems. If Linux distributions would at least default to ext2 for normal installations and ext3 (or reiserfs, xfs, jfs, etc) on servers, I think it would be a good thing.”
What a load of crap. The use of a non-journalling FS is a great way to lose your system… and isn’t that ‘critical data’? I’ve had years of experience with ext2, and it was almost enough to make me give up on linux. It simple trashed disks if they weren’t cleanly unmounted. Ext3 and ReiserFS are godsends. Further, the performance difference is not that great (have a browse through recent Slashdot articles for details).
“Re:Nice advertising (Score:4, Informative)
by AstroDrabb (534369) on Monday October 13, @01:15PM (#7200063)
Sadly the MS Empire does not allow you to release benchmark stats for their products. You agree to this when you use their products through their EULA. I am sure IT Labs doesn’t want to get crap over it. Then agian, maybe they will find a way to post the numbers.”
http://slashdot.org/comments.pl?sid=82083&cid=7200063
Ouch.
“I think journaling on systems without critical data is not only pointless but causes performance problems. If Linux distributions would at least default to ext2 for normal installations and ext3 (or reiserfs, xfs, jfs, etc) on servers, I think it would be a good thing.”
You have delt with ext3 too much. ReiserFs is as fast as ext2. And no it would be insane to ship a non journaled FS as the default for desktops systems. Right now people file into linux help channels with systems currupted because they hit the reset switch. They were the unlucky ones who created ext2 partitions. Tell them that their files were “uncritical”.
On the other hand, having used ReiserFS for years of unstable systems.. I have never lost one file in a power out.
1) People consider calling something “communist” as an insult. I know in the good old US of A, where people scream like a teenager at a Aerosmoth concert when president goes past and that anyone who dares speaks up against the government must be un-American, un-Patriotic and obviously a “communist”. In the rest of the world, however, Communist Parties still exist, and unlike in America, people don’t see it as in insult.
2) SMB sucks but what is the replacement? the replacement is NFS, however, no one yet has been willing to write an NFS driver for Windows. If someone was willing to make a nice opensource and free NFS driver for Windows, the reasons for sticking with SMB would be pretty much mute.
The Slammer troll is on the rampage today lol
I think the test was with multiple clients and how well the server deals with them. Just exporting a filesystem to a client over a network isn’t anything special. However, when you are juggling many clients, it becomes a bit more complicated than simply I/O.
Not defending any OS here, just observing that the article implies this was the sort of thing it benchmarked (at least one of the things).
“1) People consider calling something “communist” as an insult. I know in the good old US of A, where people scream like a teenager at a Aerosmoth concert when president goes past and that anyone who dares speaks up against the government must be un-American, un-Patriotic and obviously a “communist”. In the rest of the world, however, Communist Parties still exist, and unlike in America, people don’t see it as in insult.”
Ah well. That’s because ”communism” is generally seen as ”what happened in the cold war” and ”dictatorship-alike” in current Western society. Ofcourse such people also believe that propaganda only exists in such countries. Such people are scared when Linux is called communism. I mean, who doesn’t want to be Free? Well Linux brings us to communism, it is communism! So stay away from it! It’s FUD. That’s what it is.
IMO Linux is more a form of consensus and anarchism, but because it protects itself with GPL, it’s less anarchistic than BSD. However, anyone who wants can see throught it and copy it over under the same license which is a power which seems to work and give freedom.
Communism still means there’s an authority, but for the workers class there’s a forced equality. The ‘communism’ of say Stalin, was more like a dictatorship then about the original more anarchistic ideas. Now, can Linux ”take over the world and make it Linux only”? No. Why not? There’s power to fork and look throught it. Also, because GPL tools can be used to create, yes, to create non-GPL tools. Example: BSD. BSD uses GCC. So? They may release compiled source under non-GPL. Another example: Emacs. GPL. But one may code a commercial site wih it, or whatever.
It’s totallt chicken shit to compare Linux with communism while communism is compared to the horrible happenings in the cold war. It’s used to keep conservative people conservative (with Windows)
I’d also recommend to look at the following statement. It’s the same scare tactic as i described above: “If you pirate MP3 you’re downloading communism! A message from your friends at the RIAA”. Couldn’t find the picture, it’s at the net *somewhere*. I thought at lame.sourceforge.net, but i was wrong.
Try ModernHumorist:
http://www.modernhumorist.com/mh/0004/propaganda/mp3.cfm
Peder
“Communism still means there’s an authority”
I meant the communism like Stalinism or Leninism, communism as in ‘state left-wing authorian’. Not all ‘left wing’ people want such a state, or are even pro state. Not all socialists are pro state-controlled left-wing semi-authorian. There’s also non-state socialism. Among others, based on love, sollidarity and consensus. This more near anarchism.
For anarchism, see http://www.infoshop.org
contains an extended FAQ about anarchism.
Thanks to Google i found the pic Here it is:
http://www.boners.com/content/789342.1.jpg
Ah well. That’s because ”communism” is generally seen as ”what happened in the cold war” and ”dictatorship-alike” in current Western society. Ofcourse such people also believe that propaganda only exists in such countries. Such people are scared when Linux is called communism. I mean, who doesn’t want to be Free? Well Linux brings us to communism, it is communism! So stay away from it! It’s FUD. That’s what it is.
Well, at the most fundamental level, the ideal communist society is where money doesn’t exist and everyone works for the good off each other. People contribute to society what they can and only take what they need, in other words a utopia.
This is the same utopia that is shared by Christianity, Judaism and Islam.
Lets back to Linux. What is so bad about people working for the good of each other? people taking a bit of code and give a bit of code back. Wouldn’t you say that his is an environment that fosters co-operation rather than dog-eat-dog environment which commercial development promotes.
But if you wish to hear opinion from people from ex-communist block:
MS and corporate practice at all seems more like REAL communism. Which was about stealing property (hiding it then behind either iron curtain or closed-source proprietarity) and cinism about laws and freedom.
While GPL looks like extreme libertarian capitalism, with its extreme care about property, laws and freedom.
And this is why Big Brothers dislike GPL so much – it requires fair play.
Actually, the “share and share like” theme has a direct relationship to communism. As was mentioned above, this is the idea of an ideal communist utopia. That said, I don’t think there is anything wrong with communism as how it is described in Karl Marx’s “Communist Manifesto” (note the similarity in title to GNU Manifesto- http://www.gnu.org/gnu/manifesto.html ).
Also, there seems to be a lot of propaganda involved in Open Source. Some folks intentionally lie about competitors (ie Microsoft Windows) in order to make their own operating system look better. “My mom couldn’t figure out how to use Windows 95, but she upgraded Mandrake Linux to run a custom Linux-2.6.0-test3 kernel.” Of course I don’t speak of all open source users, but a very vocal minority. And in doing so, they are eroding their own credibility, as well as casting a negative image on all of Linux. FUD only causes more FUD from the other side, and leads to a situation where no one believes anything.
If these members of the community where really about helping each other, they would describe positive AND negative experiences with Linux. Problems don’t get solved by denial. And people could mention the faults of Linux without being flamed. Perhaps someday when the marketshare of Linxu is composed of less angry 15 year olds who think they are rebels by installing Linux, and more of intelligent end users and developers, this problem will start to be less of an issue.
“Well, at the most fundamental level, the ideal communist society is where money doesn’t exist and everyone works for the good off each other. People contribute to society what they can and only take what they need, in other words a utopia.”
Society? You mean state-communism? Is there a state in this ”utopia” you describe? If so, it’s like state-communism or state-socialism. This is not a utopia to me, because there’s still authority.
What you describe, sounds more like a form of anarchism to me. Opensource would be a role in such a form of happiness, and i don’t think it’s an ”utopia” because that means it’s impossible. If enought people want it to be possible, it’s possible, be it currently in small collectives (yes it exists!). If people keep seeing and saying it’s an utopia and state it’s impossible, it will ofcourse never happen in vast majorities. Self-fullfilling prophecy, thus a fallacy.
Currently, people fight for liberty, and it’s possible to live very liberal without beeing attached to ‘the system’ of capitalism even in capitalistic countries. There are always ways to achieve an autonomous liberty together, based on consensus.
Opensource is one of such…