Benchmarking ZFS On FreeBSD vs. EXT4 & Btrfs On Linux

[RahulSundaram]

That is cool. But, there is a long way from wanting BTRFS to provide complete data integration, and for BTRFS to really do it in practice. It is not as easy to just add some checksums here and there. It is much more difficult. Today, after all these years since the first hard drive was sold, only ZFS provides good data integrity (which has actually been proven by comp sci researchers in papers) in a common filesystem. No one has succeeded before.

Could you provide a reference to the paper you are talking about because I don't simplistic claims. I was just pointing that the design is intended to provide full data integrity rather than just metadata integrity because of the fundamental design. I assume you are not a filesystem developer and neither am I so I am not going to argue about implementation details. If you want a additional reference, you can look from someone who was a ZFS developer and now works on storage subsystem in Linux.

http://lwn.net/Articles/342892/

[kebabbert]

@kebabbert

We all agree that ZFS is better today, but you are claiming that this would never change. You are claiming facts like:

"Linux is really bad as a Large Enterprise server. It is not because of the bad filesystems, but because of limitations in the Linux kernel"

You don't now how linux will evolve in the future.

No, I dont claim that Linux will always have those limitations in the kernel that the senior storage expert talks about. Actually, I made it clear on several places I talk about now and near future: I say things as BTRFS will be better in the future.

You are not showing academic papers you are shoing us magazine articles.

I have showed academic papers. And other credible sources, such as official benchmarks, interviews to Linus T, etc. Those sources are credible.

Someone like that doesn't have PhD in math, as much you have PhD in narcissism.

I have never claimed I have a PhD in math. I do not. My level is Master's. Regarding narcissim, I do not have a PhD in that either. But I do react when people say wrong things about ZFS, then I correct them. And I do react when I read all the FUD here. I try to counter balance all FUD about Solaris. I do not prefer to be silent, and letting the FUD to continue "Solaris is dead", "ZFS is slow", etc. I am trying to counter FUD, by pointing out errors in people's posts and linking to credible sources. If you prefer let the FUD spread, it is fine with you, but not to me.

[kebabbert]

I want to see papers backing up your claims, otherwise you're a FUDer and a troll. Those journals aren't papers and those links you gave are FUD, because they don't show papers neither. Writing more things and repeating them you're FUDing and trolling more. Show me the papers.

No, you dont understand. Research papers must not be published in only the journals you link too. There are 1000s of journals. So, no. BTW, I have posted links to several researchers.




Kebabbert wrote
"There are several people that wants you banned. I wonder what the moderator say when I say that? What do you think, kratfman?"

It's a matter of believing. :> I think he can consider you're lying. I hope you didn't go away, because there are some things which can be fun.

You know what, Kraftman? I just lied and FUDed about you. I have got no PM or mail from people wanting to ban you. There are no PM nor mail about you! This was just an exercise to show you what true FUD and lies are. Now, do you finally understand what FUD is? Do you see that I do not FUD or lie about Linux, ZFS, etc? I FUDed about you, I lied, I said many false things - but I could never prove my claims about you if you had asked me. You should have asked "I dont believe you, prove there are lots of people wanting to ban me". And then I must be silent, and then you would understood I just lied and FUDed about you. So, you must always ask for credible sources when someone says something. If he can not give credible sources, then maybe he is lying.

So let me ask you again, do you still think that Linux critizim is FUD? The same thing? Dont you see that FUD is not crititizm? I do not make things up, I post to credible sources. You see that I do not FUD. I just strongly critizice, but that is another thing than lie and FUD. I do not lie.

(I dont want you banned, because you give me a reaon to post about Solaris and ZFS and Dtrace, etc! )

[kebabbert]

Kraftman wrote:
"Solaris is dead. Prove me wrong moron:"

I did not understand what you want me to prove. Could you be more clearer? "Solaris is dead"? I mean, it is sold today. People runs it today. Development continues today. etc. I dont understand what you want me to prove?

Again, could you specify what you want me to prove? If you do not specify then I can not prove it. How do mean "solaris is dead"?
1) Do you mean development of Solaris has stopped?
2) Do you mean Solaris is not sold anymore?
3) Do you mean no one is using Solaris?
4) Do you mean there is no market for Solaris?
5) Or do you mean everything 1-4?

What do you want me to prove, Kraftman? Please specify. I can prove everything above. But what do you ask? I do not understand. How many times earlier have I asked you to be more specific? 30 times?

[kebabbert]

http://vger.kernel.org/~davem/cgi-bin/blog.cgi/2007/04/10#bonwick_scalability

It says Solaris guys and Bonwick FUD.

I think we have discussed this link earlier.

You didn't backup your statements by research papers, white papers, official benchmarks and interviews you have posted (I'm not sure you posted all of these, but ok).

Kraftman, you are hilarious. First you deny I posted research papers, white papers, etc - and then you say something like "I am not sure you posted these, maybe you did". Jesus Kraftman. Why do you make up things? If you dont know if I posted this sources (which I have done) then dont deny it! Kraftman, dont make things up, because then you FUD. Please stop FUD, I told you.

This is the most credible link in my opinion. More credible then links from sun, solaris devs etc. So, prove me wrong.

In my opinion it is not credible at all. Research papers are more credible. I can also start to post links to Solaris guys explaining how bad Linux is, but I dont. Because Solaris guys are biased against Linux. And Linux guys are biased against Solaris.

It is much more credible if you post to Solaris guys pointing out flaws in Solaris. I post to Linux people pointing flaws in Linux. I do not post Solaris guys pointing out flaws in Linux - that would be pointless. Therefore, you link is not credible at all. Your link goes to a biased Linux guy.

[kebabbert]

Kebabbert has started half a dozen never-ending troll fests this week. Every time you counter one of his points, he responds with about 50 new ones in a series of responses that no one has the time to respond to,

Or, maybe you can not prove me wrong. Maybe that is the reason you dont even try. Come on. Go ahead, prove me wrong. In university, I have studied maths for many years, and all we did, was to find errors in proofs. I am highly trained in proofs.

and half of them just repeat exactly what you had just proved wrong.

Woah! Did you prove me wrong? Can you quote that post where I was wrong? I missed that. Or did you just make it up?

In fact, I showed that YOU lied.
http://www.phoronix.com/forums/showp...6&postcount=88

But you have not showed that I lie anywhere. You are maybe lying about this, too?

[kebabbert]

Could you provide a reference to the paper you are talking about because I don't simplistic claims. I was just pointing that the design is intended to provide full data integrity rather than just metadata integrity because of the fundamental design. I assume you are not a filesystem developer and neither am I so I am not going to argue about implementation details. If you want a additional reference, you can look from someone who was a ZFS developer and now works on storage subsystem in Linux.

http://lwn.net/Articles/342892/

I am not a filesystem developer, that is correct.

I am only saying that to offer data safety is extremely difficult. You can not just add checksums all over the place in BTRFS. It is not that simple. I have linked to research papers that show that filsystems like ext3, ReiserFS, XFS, etc does not give data safety. Neither does raid 5, raid 6. And all these technologies have lots of checksums. But still they does not offer adequate data protection. Why do you expect BTRFS to be different? There is one BTRFS developer that tries to do that.

Sun ZFS team has many years experience of data corruption. They can do it right.

In my opinion, if Oracle really wants to bet on BTRFS, Oracle should have assigned a whole team of developers to BTRFS. But guess what? There is only one single developer on BTRFS. Oracle is not interested in BTRFS, now that they got the best: ZFS. Come on, do you really expect BTRFS is a serious attempt by Oracle? There is only ONE full time developer! Jesus. Who are you trying to fool?

[Jimbo]

No, I dont claim that Linux will always have those limitations in the kernel that the senior storage expert talks about. Actually, I made it clear on several places I talk about now and near future: I say things as BTRFS will be better in the future.

Ok perfect! You didn't sound that way 10 pages ago.

In my opinion, if Oracle really wants to bet on BTRFS, Oracle should have assigned a whole team of developers to BTRFS. But guess what? There is only one single developer on BTRFS. Oracle is not interested in BTRFS, now that they got the best: ZFS. Come on, do you really expect BTRFS is a serious attempt by Oracle? There is only ONE full time developer! Jesus. Who are you trying to fool?

mmm I don't know how many devs are under oracle nominee. I suggest you to look at btrfs mailing list and kernel fs dev mailing list, you could see that btrfs is being actively developed by a bunch of devs, not only one. There is a big difference from other projects where only 1 person develops, you are painting the wrong picture. On linux fs dev list you could also see how FS improvements on multicore environment are being vastly improved for 2.6.36-37, there are bench too, done by devs. If you read more on developers mailing list and less from 2008 consultant / magazine articles you could see that your ideas about near future are basically wrong.

[jalyst]

Please, stop the madness!

Lets just discuss interesting facets of the filesystems in-question in their respective environments.
What's the point of all this "tit-for-tat"?!

There's no such thing as the l337est environment/filesytem.
All are worthy of informed, useful, constructive/helpful discussion.

[kebabbert]

Ok perfect! You didn't sound that way 10 pages ago.

Maybe I was not clear enough, or you misinterpreted me. Anyway, as of now, Linux has limitations. But in the future, Linux most probably will only be better. IBM basically says this too. IBM AIX is phased out, it will be killed. IBM will bet on Linux instead. This is outspoken and official. So, basically, IBM also says that Linux is not mature enough right now. IBM sales personal always tries to sell IBM AIX which is very expensive, and as a last resort, IBM tries to sell Linux. But IBM does not try to sell Linux the first thing IBM does.

mmm I don't know how many devs are under oracle nominee. I suggest you to look at btrfs mailing list and kernel fs dev mailing list, you could see that btrfs is being actively developed by a bunch of devs, not only one. There is a big difference from other projects where only 1 person develops, you are painting the wrong picture. On linux fs dev list you could also see how FS improvements on multicore environment are being vastly improved for 2.6.36-37, there are bench too, done by devs. If you read more on developers mailing list and less from 2008 consultant / magazine articles you could see that your ideas about near future are basically wrong

I have read on a link here, that there is only one full time developer on BTRFS. I know there are part time developers, and probably some hobby developers working on their spare time. But still, if Oracle is really seriously committed to BTRFS, they should dedicate much more resources, preferably some ZFS developers. You can not develop an Enterprise filesystem in the basement, by some part time developers. If Oracle wants to sell and earn money on BTRFS, then it could be wise to develop most of BTRFS in house? Now there is one full time dev, and other part time developers. I dont see how you can get an enterprise product to support and sell, from such a small effort?

But maybe I am wrong, as you say. Or maybe I am right: now that Oracle has ZFS, then Oracle quickly lost interest in developing another filesystem that does exactly the same thing as ZFS does, without adding something new. Why develop two identical products? Does not make sense from a business perspective. But time will tell. Maybe both of us are wrong!

From a personal point of view, I hope BTRFS succeeds, because that forces ZFS to become better. Which forces BTRFS to become better. etc. Upward spiral. Competition and choice is always good for us customers.