Slashdot Mirror
Novell vs. Microsoft - Benchmarks
I Just found this article in The Register which talks about KeyLabs comparing Novell's NDS eDirectory vs. MS's Windows 2000 Active Directory. Here is the KeyLabs benchmark report (requires PDF reader) and here is another link to Novell's eDirectory Benchmark. (also, here is what Novell says about it). I'm sure this is only round 1 and we will surely see some other reports...
On the Novell page, comparing reliability..
The scale is between 80 and 100 percent. The low 90 that MS scores just PALES in comparision to that full-height 100.
Of course, ~10% downtime is an extremely significant number if you're talking mission critital.. But still, it's not the 50% that it looks like.
Anyone who has not read it, should read How to Lie With Statistics. Highly recommended for becoming alert to a lot of tricks like this one.
Cheers,
Ben
My usual seat in the cluetrain is at A HREF="http://pub4.ezboard.com/biwethey.ht
Lies.
Damn lies.
Statistics.
Benchmarks.
"Destroy science and religion. Science would re-emerge exactly the same; but not religion." - Penn Jillette, paraphrased
Ummm, NDS and Active Directory are not just different implementations of a standard, they're completely different pieces of software. They serve a similar purpose, but I honestly can't see how they could be objectively compared...
MoNSTeR
The same company was used a few days earlier by Microsoft. Microsoft won those tests. OTOH those tests were chosen by Microsoft and did not include any "contains" searches, the NT box was optimized, etc.
As Jeremy Allison said, everybody knows ahead of time who will win a particular test, and nobody will participate if they don't honestly think they are going to win. He said that when Linux' IP stack gets optimized and Solaris' file-system gets improved that you will see him repeat the Mindcraft benchmarks - but not until...
Cheers,
Ben
My usual seat in the cluetrain is at A HREF="http://pub4.ezboard.com/biwethey.ht
As a longtime /. reader, however, I say "before we slam anyone (Microsoft, Novell, the test lab, etc.) let folks familiar with Win2k post information about the test parameters to show any biases toward Novell in how the test was set up -- a la the accusations against Mindcraft which IIRC were mostly proven to be false in the 3rd run of that particular benchmark.
What I like about this particular study as published is that they were very up front about the tuning parameters, such as they are. My hope would be that Microsoft will respond with their "ideal tuning parameters", and then the benchmark can be run again, with the results posted here.
...Open Source isn't the only answer -- but it's almost always a better value than the alternatives...
Nobody got fired for buying IBM^H^H^HMicrosoft.
Burris
The thing is, Novell may never become a great application platform (their Java-induced delusions notwithstanding) but you don't really need to use Netware for that. In fact, it's now downright easy to manage NT servers through NDS and use them for what their good for.. (uh, that's nothing!) well, application support. and then do file and print services with Netware.
I'm still not sure how they plan on making money off this (they have to deliver more directory apps) but Active Directory is NO compitition, it's really not even worth talking about unless you have a 100% MS shop. Now you can manage solaris, linux, etc. users with it as well (this includes file rights, group membership, password..). Then there's a product that let's you manage Exchange and Notes directories through NDS, hot damn.
While I'm babbling on about Novell.... ever notice their products just appear suddenly. No preanouncement, no nothing. Just BAM: now shipping. Unfortunately it's that kind of marketing that gets them screwed all the time. Anyway, don't mean to sound like a shill for Novell, but you don't hear much rah rah about them around here so.... here it is: rah rah Novell.
Keylabs was spun off from Novell years ago and initally funded by none other than the CEO of Novell(correct me if i'm wrong). The people that started Keylabs were former Novell internal test-lab employees. So yes, I think we could assume a little bias here.
Virtually every aspect of Netware can be tweaked and configured depending on your configuration. WindowsNT doesnt give you that level of modification. For example, you can set the Maximum Number of File Locks on a Netware server. On a WindowsNT server, you cannot set this (It is theoretically set to be infinite. If I'm wrong, tell me).
The reason why NT was not tweaked nearly as much as Novell is because that the things that were configured under Novell are simply not accessible under NT.
This underlies the fundamental difference btw NT and Novell (or NT and Linux for that matter). While Netware (or Linux) is more difficult to set up and optimize, it is far faster and far more stable because the server can be optimized far more than NT, with a lot less garbage code (Who needs a GUI on a server anyways?).
Jailbrekr.
Feed The Need[goatse.cx]
As a Nescape employee, I can't help but ask why Netscape Directory Server wasn't included in these tests? Since Netscape Directory Server has > than 70% market share in the LDAP market, I don't understand how it wouldn't included in a major benchmark.
I don't know the particulars of these benchmarks, but Netscape Directory Server has been benchmarked at speeds almost an order of magnitude higher. Of course, I don't know if that's comparing apples to apples. But I'd like to know!
It hardly matters. Netware may well be superior but Novell's share of the market is declining and will continue to do so, IMNSHO. Their turnaround came too late to save them. If Operating System choice was normally decided on logic alone, Microsoft would have gone bust just after the second time they released a buggy OS. The problem is *not* a technical one - Microsoft has won the hearts of the suits, Apple the hearts of the artists, and *nix hearts of the geeks. There's just not sufficient room left to Novell for them to survive in the long term.
Consciousness is not what it thinks it is
Thought exists only as an abstraction
What evidence do you have for that?
Mindcraft?
The problems were not in Samba, they were in the Linux TCP-IP stack being a bottle-neck with 4 network cards. And Solaris has problems with slow file-systems. Samba itself performed just fine.
Regards,
Ben
My usual seat in the cluetrain is at A HREF="http://pub4.ezboard.com/biwethey.ht
SMP support in Windows is way ahead as compared with Linux ...
.... ..
If you don't know about that then obviously you know nothing about Windows
Huh,, just anonther vi expert
Possibly cause there's more NT servers out there than there are novell servers?
:P
You should tell your friends not to play quake on their NT servers
This is pretty funny... I just finished reading a "report" from Key Labs comparing various Fibre Channel switches. They grade each switch (A, B, C, etc) overall and on specific tests. One particular swich, from Brocade, scored straight As - on every test. The other switches barely passed (according to Key Labs). Shortly after I read this eye-opening report, I found an article that shed some light on the situation (although I can't find the link now...). Low and behold, Brocade funded the Key Labs report.
Free suggestion to Microsoft - get these guys to run your next Linux vs. Win2k FUD benchmarks. Be sure to write them a big check.
---- I made the Kessel Run in under 11 parsecs.
Windows 2000 Professioal supports 2 processors. Windows 2000 Server supports 4 processors and 4 GB of RAM. Windows 2000 Advanced Server supports 8 processors and 8 GB of RAM. Windows 2000 Datacenter Server supports up to 32 processors and 64 GB of RAM. All Windows 2000 versions are only available for the x86 architecture.
-Entropy [think outside the system]
But that's what this forum is for... ;-P
Windows NT has a larger percent of servers than Netware because of specific applications that require it. If you were to compare installations of Windows and Netware in a scrictly File/Print/Directory services role, I think those numbers would be completely different.
The fact is, stuff like Fax Servers, Database Servers, etc are primarily written to the MS platform . Netware doesn't compete in this segment, so including these servers in the comparison skews the results.
Also, when considering File/Print/Directory Servers, a single Netware server can out-perform NT when serving large numbers of users. Windows NT is basically a workgroup platform, so more NT servers are requred to serve the same number of users. Again, this fact skews the raw numbers when comparing number of units shipped or whatever.
Finally, people - especially the media - have been thumping the Netware is dead drum for years! Yet the platform continues to improve, and yes... sell! You might be surprised at how many Netware servers have been in production for more than five years in small offices which don't necessarily upgrade or update their systems often due to budget concerns..
I AM, therefore I THINK!
Microsoft and several independent authors have warned against using Netscape LDAP with AD; it's tuned and optimized (of course) for Microsoft's LDAP3 standard. Netscape is optimized for NetWare, because that was the most prevelant when it was released. The NetWare test is on a shipped, finished, tuned product; the AD tests were on a pre-release build (2194 is the one just before the RTM build, which is NOT identical to the retail.) Finally, the tests were designed to test functions AD is not optimized for; AD is intended to fill the distributed IntraNet environment.
My point is that when you read a report, consider the facts before you slam Microsoft. There are good tools made by Microsoft; Windows 2000 is one of them. Oh, and by the way; a GUI is useful on a server for a number of reasons. To name three:
-- The meek shall inherit the Earth. In very small plots, about 6 feet by 3.
I was under the impression that the Novell-sponsored test used the same Win2k setup as the MS-sponsored test.
It's not enough to bash in heads, you've got to bash in minds. - Captain Hammer
- It's better to monitor performance by watching graphs than by waiting for error messages. Graphs show trends, which error messages never do.
- In a GUI, you can have multiple control panels open at once. A command line forces you to work one at a time.
- GUIs are more impressive and "boss friendly." Don't underestimate the importance of this one; anyone who doesn't think it's important has never had to fight for a budget.
I can do all of these with my servers, none of which have a GUI (or a monitor, for that matter). What they do have is X libraries, that let me display to any X server on my network. As for being boss friendly... do you really take your boss down to the machine room to show them a GUI? I'd say it's far friendlier to show them the same information on my personal workstation. Or even at my boss' own workstation without them having to leave their desk."The invisible and the non-existent look very much alike." -- Delos B. McKown
One million two hundred thousand percent difference is equal?
Novell v. Microsoft isn't even a competition. Novell kicks Microsoft's ass handily in every area it tries to. Okay, so you can't play Solitaire on Novell, aside from that, it's great.
As for stability, it's an admin's dream. Power, extendibility, insert buzzword here, Novell is just plain great.
As for why you don't need those directory services, for your home machine, you don't. Now run a business. A large business. Now you need them.
----------------------------