OSS Web Stacks Outperformed by .Net?

Gimble writes "eWeek has an article up that looks at the performance of portals using open source stacks and comparing them to their MS equivalents. The article's conclusion is that .Net outperforms the open source stacks, mainly because of its tighter integration, but also notes that running the open source stacks on Windows (WAMP) delivered strong performance." From the article: "Based on our forays into user forums for many top open-source enterprise applications, there are many IT managers attempting to run open-source products on Windows servers--attracted, no doubt, to the benefits and efficiencies of using open source without having to become Linux administrators. The results of our WAMP stack tests indicate that these folks might be on to something."

Left out?

From the article: "The criticism we expect to hear most is of the stacks we left out--including commercial J2EE platforms, such as those available from BEA Systems, IBM, Oracle and Sun Microsystems, as well as the many other database and server platform permutations." I can't believe they came to this conclusion on such little data. They did, however, create a blog to disparate results can be shared.

Re:Short memories

Even if that was true I'd still rather have a proper UNIX environment anyways. Not to mention a real multi-user OS without shelling out huge dollars. Oh and the development tools and ...

Sorry, but with modern windows, what exactly is missing that dis-qualifies it from being a real multi-user OS? And FWIW, I've found VS2005 to be much better than any development studio I've tried for linux.

If you buy Windows to run AMP servers ... you're wasting your money.

So you think its better to have servers which don't integrate with your corporate network? $400 for Web edition isn't a whole lot to anyone running web farms.

Linux and BSD are more efficient to work with in server contexts

Your opinion, there's no fact to that. I replace my linux server at home with SBS 2003 because its easier to manage the network using SBS2003 than it was with Linux + SMB.

work better with 64-bit processors [Win64 is a huge compatibility joke atm]

I've heard quite differently; indeed, some people claim XP64 is the best desktop OS, even better than 32bit XP. I don't think the 64 bit servers are suffering huge problems either.

don't require you to call India each time your HD breaks and you need a re-install.

A bunch of FUD here; mearly replacing an HD doesn't require calling anyone, and alot of shops image the drives every night so that if there is a failure, they slap on the saved image and are up and running again in no time.

Heck, in BSD/Linux doing a ghost of your system is as simple as burning a tarball to a DVD.

You can easily save an image of a Windows installtion as well.

No need for 3rd party ghosting tools and praying that Windows lets you "get away with" using your OS...

Go ahead and spend hours and hours tarballing your server; the fact is that it would be faster to just create an image of the HD and burn that to a DVD. There's no praying involved; my former employer had great success doing such restores. And putting down a new image is faster than un-tarring a file to disk again.

Re:Short memories

Sorry, but with modern windows, what exactly is missing that dis-qualifies it from being a real multi-user OS? And FWIW, I've found VS2005 to be much better than any development studio I've tried for linux.

I think the grandparents point is that though it is technically a multi-user OS, it isn't a very good one. The kernel level schedulers on Windows give a really poor response under multiple sources of medium to heavy load.

I'm not aware of that many "development studios" for Unix. There's a couple, like eclipse and such, but many Unix writers tend to work a different way. In line with the Unix philosophy (rules 1, 3, 4, 5, 7 and 8 if you're interested) there's a lot of people who keep the editor, compiler, linker etc all seperate. If you're looking for a "development studio" for Unix, I think you're looking for the wrong thing. We don't have a Start Menu either. :)

So you think its better to have servers which don't integrate with your corporate network? $400 for Web edition isn't a whole lot to anyone running web farms.

Well, I'm sure anyone could have pointed out that large companies with large budgets can afford the $400 dollar cost. But that $400 dollars is a cost, and to a lot of companies, it's a big cost. It's an even bigger cost when many difference licences have to be purchased. (Obviously, Unixes have costs of their own, but the costs tend to increase in line with the use being made, or the complexity of the "solution" etc, and up-front licences payments (which are immoral, in my opinion) normally end up being a far less weildy solution than Free Unix is) And in my, and the grandparents opinion, it's an unjustified cost.

Your opinion, there's no fact to that. I replace my linux server at home with SBS 2003 because its easier to manage the network using SBS2003 than it was with Linux + SMB.

Of course it's opinion! The whole damn subject is opinion! As if a (very very poorly run) test on the speed of a server is the last word on the subject! There's little universal fact to your personal experience either. ;)

I've heard quite differently; indeed, some people claim XP64 is the best desktop OS, even better than 32bit XP. I don't think the 64 bit servers are suffering huge problems either.

Have you actually used it? WinXP 64bit really doesn't work very well at all. I wouldn't know about 64 bit servers, but 64 bit WinXP is really not something that works very well.

Go ahead and spend hours and hours tarballing your server; the fact is that it would be faster to just create an image of the HD and burn that to a DVD. There's no praying involved; my former employer had great success doing such restores. And putting down a new image is faster than un-tarring a file to disk again.

It's possible to use either image or tar with Unixes. Tar is useful because it's easy to use diffs and delta compression on backups then. Tar's also useful when you only want to backup a certain area; such as /etc or /var. It's not like you're stuck with tar either - disk imaging still works fine. When you can tarball up only the parts of the machine you need to keep, the idea of disk imaging seems redundant (for most cases).

Re:Left out?

They seem to have a generalized poverty of data. Their charts seem absurd to the point of being straw men. I mean, come on - I don't think there's anything seriously wrong enough with Linux that WAMP would have a score of 12 transactions/sec, competing with Windows, whereas LAMP would have a performance of 2. My experience with Windows vs. Linux has always been that they are similar in terms of speed from pure processing tasks to 3d games. Sometimes Windows does a little better, sometimes Linux is better. But they're usually in the same ballpark. The numbers are just too neat. It's like they put up a chart saying that Republicans, Germans, Koreans and Canadians have sex once a month, whereas Democrats, Brazilians, and the British have sex five million times per second.

Moreover, the whole rest of the article is morass of poetic circumlocution. My gut feeling as somebody who works with words a lot is that they're trying to obfuscate something with a giant wall of banal text. I don't know exactly what that is, because I don't feel like reading all of it, but if I had to guess I'd say that the real thing to take away from this article is that anybody can set up .NET and a Windows box, but that it requires a little bit of patience and research to make Linux work properly - research that these people were not willing to do.

WAMP vs LAMP

I'm no system administrator, but I have a home box running WAMP (XAMPP on 2003) and it's good enough for my needs. Recently I tried out Ubuntu Server to see what it's about, and I'm tempted to buy a new pc just to run that. When I tried to run mod_python under WAMP it took a whole lot of debugging and configuration (apparently it didn't like the already installed python 2.4), but with Ubuntu it was as simple as apt-getting it.

I would very much like it if I could continue using Windows (because I run other programs that are not available on Linux) but it can't match the simplicity of Ubuntu.

I'm going to have to use the /. rule of thumb

If Linux wins, its a fact.

If M$ wins, its fud and was paid for.

If apple wins, its because of Steve Jobs.

If OS/2 wins, we're trapped in a parallel universe.

Re:I'm going to have to use the /. rule of thumb

It isn't as if they haven't been caught buying studies before. So the distrust is well justified.

Justified perhaps, but automatically accurate... not necessarily.

Plain and simple fact is if Microsoft could compete with the usefulness of a solid Linux distro their product would speak for itself. In some cases this is true but in essentially all technical senses Microsoft is just a plain loser.

Ok, fanboy.

Most of Microsofts problems is that they don't listen to the customers. I mean sure they listen to Dell, RIAA, MPAA, maybe even IBM and other big wigs. But what about us users? What does WGA give me in terms of a useful feature? What does the bloat that is WMP give me over a simpler mplayer?

So when MS doesn't add new features they are slammed for not innovating enough, and when they do add new features they are slammed for contributing to bloat that you don't want. People bitched about IE6 not having tabs, etc. Firefox came out and MS finally realized it had to update IE so it added a lot of features people were asking for and the most-heard comment on Slashdot after IE7b2 was released was "it's ugly". Face it: Microsoft just can't win.

Why must they invent their own file formats [e.g. Office files, WMV, WMA] that are proprietary instead of using or establishing more open standards? etc, etc, etc. Everything MSFT does is to benefit the stock holders through locking the "customers" into their system. Use our OS, use our office suite, use our media tools, use our development tools. All the while they ignore any sense of established standards [ISO C99 anyone?] which make interoperability a bitch for Windows users. There is simply no reason why MSFT uses these awful platform dependent libraries.

You sorta' answered yourself there.

Take DirectX for instance. On any OTHER platform you combine Allegro with OpenGL and have essentially the same thing [just 1/10th the size and in C]. But no, we must use the DX "experience" because somehow the hype makes it shinier!

Not every product is a winner. MS historically doesn't release every single product as a beta and quietly stop promoting the ones that suck. Instead they release final versions and some fall on their face. No company has a perfect record.

I know what I'm saying is "no duh", but you seemed to be hinting that MSFT hatred is not warranted. Us "OSS" users don't hate MSFT because it's better. We hate it because it lulls people into a sense of superiority when all it does is move to separate them from their money. It creates nightmares for us who chose to chose.

The problem is not that criticism isn't warranted, it's that MS can't win no matter what. If they release a weak or buggy product they get slammed, but if they take too long to release they get slammed. If they don't add new features they get slammed, but if they add new features it's called bloat. If an MS product gets bad reviews the reviewers are being honest, but if they get good reviews the reviewers are obviously being paid. For years MS got slammed for security issues, and they beefed up SP2 and suddenly there were waves of "but it broke my application" complaints. The list goes on.

Microsoft has gotten so big that they are in the impossible position of trying to keep everyone happy. I'm not particularly a Microsoft "fan", but I hate this wanton "Micro$oft is teh suxors!1!" b.s. OSS fanboys need to grow up and realize that Microsoft can't go back in time and correct the sins of the past, and since it is a monopoly it can't just genuinely screw its customers and break every file/application by releasing a new version of Windows that corrects all the problems of the old versions but offers no legacy support. They have a tough balancing act to do and, while they're not perfect, they're getting better.

Nice thing about OSS

This'll now be a high priority - beating .Net speed-wise - in the next few releases, such that by Christmas, we'll see *AMP performance picking up, whereas we have to wait on MS for .Net

Hits per second ~20 ??

I am amazed that they got Windows 2003 to run on a wrist watch!

Running a web server over an RF port from the wrist watch to a phone scewed the results a bit, but its the only communication mode they had.

The smartphone was the only client they had handy to test with, since the test was carried out on a long flight.

Amazing stuff!

Worst... Benchmark.... Ever...

I'm still not exactly sure what they tested. They have vague terms like "Request per Second" and "Throughput", yet they don't actually say what each page that is being requested is actually doing.

For the .NET tests they say they used "Sharepoint". Huh? For what? Considering that Sharepoint is *extremely* complicated and has incredibly rich functionality they should be very clear as to what they used it for.

Not to mention the fact that using a portal application in your tests means that there is really very little way to isolate if it was a poorly written portal application or a crappy framework that the portal application was built on that's causing perf issues.

It is very difficult to test framework vs framework, but this is just about the worst way one could even attempt it.

At absolute best, this compares portal frameworks on various platforms. Even if they were trying to do that, they did a piss poor job.

MindCraft would be the "worst" so far.

But this is a very close second place.

#1. NO tuning was done on the LAMP stuff. None at all. They ran the stuff "out of the box".

#2. They didn't write their own app. That means they didn't test the SAME processes on each system.

#3. They didn't bother to find WHERE the differences were. Is it in the IP stack? Is it in the OS? Is it in the scripting language? Is it in the app?

How bad can "research" be and still be published in "eWeek"? There wasn't any research done for that article.

Microsoft has, in the past, taken various short-cuts when IIS was the server and IE was the browser. Is that the case in this "study"? Are the other "stacks" "slower" because they follow the protocols?

You won't know because they'd didn't LOOK for the REASON behind their "results".

At least MindCraft was paid to do poor research.

If I may expand upon your post ...

You raise some excellent points. If I may be allowed to expand upon them ...

#1. Set a price limit. You can set multiple limits ($1,000 vs $5,000 vs $25,000 vs $100,000 vs $1,000,000+). The key concept here is that you get different characteristics as your budget increases/decreases. What characteristics does each "stack" offer in each price range? Yep, this does give the advantage to Free stacks (Free like speech, free like beer). Deal with it. In the Real World the bottom line is the bottom line. Each team gets to spend the money however they want to.

#2. Get the "experts" to tune each stack. BUT they must document each modification they make, including WHY they made that modification (what testing did they run and how did those test results tell them what mod's to make) AND they are only allowed to make mod's that can be found via public websites (no secret tuning parameters that are only known to the organization writing that software) AND they aren't allowed to touch any source code. They get what everyone else gets.

#3. The fun part. Each team gets to pull apart the work of the other teams. Even if your solution is faster for the specs given, how much wiggle room do you have? Is faster and fragile better than slower and stable? How much "slower" is acceptable for how much more "stable"? Can the other team defeat your security (network access only)?

#4. Freeze those systems. Then, over the next year, patch them and re-test them. Do the patches break the "tuning" that was done?

Now that would be an informative test process (and would result in lots of articles and interviews for the magazine publishing it).

Yeah, you can run Linux / Apache / MySQL / perl on a single drive workstation and get damn good performance for less than $300.

But that will be completely different from Oracle / Java on a cluster of Suns costing $10,000,000. And not just in the number of boxes you'd be running.

Test components too variable

How can you test the performance of a stack and compare it to others when the back end database servers, portal software and web server software is different?

How is the statement that .NET stacks are faster true when it could be the implementation of SQL being faster than MySQL? This test just doesn't make sense to me.

Retarded

The .NET CLR runs compiled bytecode and IIS runs in kernelspace, the only httpd I know running in kernel space on nix is tux (redhat content accelerator) and nobody in their right mind is going to serve dynamic content with that. Did they test PHP with an opcode cache and the CLR running a dynamic language? This isn't even apples and oranges, it's apples and teapots. If I wanted performance, I wouldn't be running either Apache or PHP!

Girly man coder

If you're getting your ass kicked by .net, you are one girly man coder.

ASP.Net is pretty nice...

Not to start a war here, but ASP.Net is a pretty damned nice environment to work under... I've used a lot of PHP, Cold Fusion, some JSP, and Classic ASP in the past. ASP.Net is my favorite.. I've been peeking in with Ruby/Rails but just haven't had the time to dive in much. of the above Ruby/Rails is probably the closest competitor on an ease of development/functionality level.

Linux still wins

Because I don't feel like paying $1500 per machine for Windows 2003 server on every server in my web farm. Shit, that's twice as much as the servers I'd run it on! Grid computing and server farms are very poorly suited to a commercial operating system.

Re:Linux still wins

Ya, I wouldn't want to pay that either. Luckily, Windows doesn't cost that much money.

Windows Server 2003 Web Edition, 32-bit version - $399 Open NL
Windows Server 2003 R2 Standard Edition - $999 (5 CALS)
Windows Server 2003 R2 Standard Edition - $1,199 (10 CALS)
Windows Server 2003 R2 Enterprise Edition - $3,999 (25 CALS)

http://www.microsoft.com/windowsserver2003/howtobu y/licensing/pricing.mspx [microsoft.com]

You can also get licenses for a lot less than retail on eBay, and it's perfectly legal. I've purchased Web Edition for as little as $200, and Enterprise for $1200. There are lots of companies who buy these things in bulk and end up not using them.

In addition, if you're not hosting an external site (customer facing) you can get an Action Pack subscription for about $300 that gives you access to up to 5 licenses for each of these OS's.

See: https://partner.microsoft.com/40016470 [microsoft.com]

Re:Linux still wins

It's actually much worse than $1500 a box. And I'm a windows guy, I admit it, however licensing for windows products REALLY sucks for a small business that wants to run legit.

Here's the price breakdown for a SINGLE webserver that allows external connections to authenticate (non-domain, say a e-commerce site with user accounts) against a SINGLE SQL 2005 Database. Sql Express is free, however it's not licensed for unlimited users in a production environment.

Web Server (Prices from CDW.com):

• 1 Copy Windows 2003 R2 (5 CALs): $959.99
• 1 External Connector License for Windows: $1,969.99
• Total: $2,929.98

SQL Server:

• 1 Copy Windows 2003 R2 (5 CALs): $959.99
• 1 Copy SQL Server 2005 Workgroup Edition 1 Processor License: $3,819.99
• Total: $4,779.98

Grand Total for a single web and sql box: $7,709.96.

And don't forget that you'll need SOME hardware to run that OS. Even barebones boxes with no data protection will run you $500 a box.

So, to start a basic e-commerce site on the legit, you're talking roughly $9,000 for windows and $1,000 for Linux/OSS.

TOUGH sell for Microsoft for the little guy.

Integration vs. Cost effectiveness

I'm actually wondering how the wonderful non-biased folk here at /. are going to interpret these results.

I don't know a damn thing about any of this but it says to me from a layman's point of view that invidually maintained and installed components are just not as efficient as a completely integrated suite of applications, and this is exactly how the ignorant bosses of knowledgeable admins will see it. Though I was interested to see the rise in the use of OSS in the workplace.

I could have gone down the whole "OSS SUX" route but that's a flamewar I'm not starting.

(Today.)

"performance"?

What was the set of measures? For me, "performance" has more to do with uptime, reliability and security. Those are the performance standards I care about.

Holy Throughput!

I'm wondering if the high throughput numbers for the .Net stack were caused by it deliving huge binary files to the client. Ya know, 17MB Active X controls. Anyway, I didn't randomly come up with this conclusion, the article didn't mention the transactions per second for .Net. So, I conclude from ALL of the data that it did one transaction of 17MB*.

*No math was done to come up with the 17MB figure.
Also, no animals were harmed during the writing of this comment.

The article is NOT that conclusive

I've read the article before it hit /. and their conclusion is that there is no clear winner. .Net outperforms OSS solutions on some tests and vice-versa. The surprising(*) results are how good WAMP performed in some of the tests (if you really want specifics RTFA). Here is a direct link to the tests [eweek.com].

* - I've seen similar results in benchmarks of Mono & .Net, i.e. Mono apps with .Net framework vs pure .Net and pure Mono, so although there is no connection between JIT compilers and web servers performance, the trend is there.
Too bad the article haven't touched Mono.

I don't really understand what they are testing?

Can someone explain it to me?

I don't think it is FUD, but I do get the impression that they are trying to invent a benchmark that really doesn't make any sense. Different PHP projects can have vastly different performance; and I'm not sure that Plone compares to Sharepoint server. I wouldn't know, though, because I don't use Sharepoint, and I have little/no idea what they did in the test.

Anyone have a closer hunch?

Lies, dammned lies and benchmarks

Different day, same old story [slashdot.org] Makes you wonder when people will finally stop taking these kind of 'studies' serious...

Linux administrators

From TFBlurb: "...without having to become a Linux administrator"

Thank $DEITY I did try getting Linux desktops on my home network and shortly after settled for apt-based distros. Linux administration is a breeze compared to windows. Desktop users' life is also good if peripherals are recognized, especially if by OSS drivers. Your mileage may vary cause most of you were familiar with windows in the first place, I came from good old macos. Anyway I don't care to try and convince you with examples. Those who are curious don't need my opinion, and the lazy ones are better of wherever they are.

Back to topic: .net is faster? maybe. That is a reason to revert from open source, not memory hungry, nice to code with stacks to Microsoftland? haha, no. Apart from the main reason (Freedom), the secondary reason (smalltalk on net is not like ruby on rails), if people reverted back to Microsoft Microsoft would revert back to itself in the 90's. Are you sure you want that? [N/N]

I only skimmed the article

...but which 'P' did they use? Did they use mod_perl or mod_python, or just call things as straight CGI scripts? That would certainly kill performance. Did they preload often-used subroutines into the embedded apache stuff?

Don't forget that, besides the software vendor

...your .Net solution also locks you into x86 chip architectures.

Performance so often comes at the expense of flexibility.

Given a requirement to work nicely across arbitrary hardware platforms with 'Doze, how will you do this? Emulation? Sorry about that performance...

Certainly, if you're starting from scratch, homogeneous is the way to go, but sometimes you're no' so lucky.

stacks?

Am I the only person who has never heard of the word "stack" in this context?

Wikipedia: stacks [wikipedia.org] - Nope
Google definition of stack [google.com] - Nope.
Urban Dictionary: stack [urbandictionary.com] - Nope.
Dictionary.com - stack [reference.com] - Nope
Google search "IT stack" [google.com] - Only hit is the eweek article.

I think they made up this term.

s/stack/platform/g
or
s/stack/framework/g

No wonder Linux sucked!

For our tests, we ran what is essentially a pure Zope/Plone implementation, with Plone running on a SUSE Enterprise Linux system.

In some benchmarks, Plone was an average performer, sticking close to the middle. This is actually better than we expected, given that the Plone documentation is very upfront about the fact that Plone shouldn't be used alone in a production environment and should be run behind other servers to improve performance.

So, they ran an outward-facing Zope server (after being explicitly told not to) and the performace was lackluster? Go figure. In the real world, they'd run Zope behind an Apache or Squid proxy (as per every installation recommendation I've ever seen) which would immediately boost throughput by an order of magnitude. In short, using Zope to dynamically generate static content instead of caching the results whenever possible is insane, and pretty much no one does it. They also apparently forgot about ZEO, although I'm not sure how you can be savvy enough to get Zope up and populated without knowing about it's built-in clustering.

Apparently they had no interest in any tuning whatsoever, to the point of de-tuning it by installing it in an explicitly unrecommended configuring. And then it lost. Go figure.

Trains and planes

1. They say no word about the problem and the implementation of the solutions. Results may vary depending on the problem.
2. Comparing J2EE/.NET to PHP/Plone is bollocks. Problems that are solved with J2EE/.NET today are so complex that choosing PHP/Plone instead is no option. It's like comparing trains to airplanes.
3. Where are tables, figures and graphs?

Very accurate comparison . . .

Who knew that a stateless, event-driven constantly running MVC-like framework would outperform scripting languages that had to be reset for each request? It's a good thing they didn't compare stuff such as RoR or Django with FastCGI and page cache, or else ASP.NET wouldn't look as great as it did in this article and eWeek would feel ashamed for still using the obsoleted ASP . . . well, that last one should be valid either way, especially since they WTFA.

All well and good but...

As recently as last month, one of our customers corrected their very peculiar connection issues by replacing their windows server with a linux server. For some reason the windows server was changing the acknoledgement part of the TCP header- for the same client- at the same computer- for every transmission.

Windows Servers/App languages doesn't seem to scale well. It's a *great* way to get your business up and running asap. But you run into growth problems and need to switch to an enterprise solution (oracle, as/400, java, linux, etc.) once you reach a certain size. I still prefer windows as a desktop OS for now. I still slightly prefer windows office to openoffice. I think part of that is years of using office makes me comfortable but openoffice gets closer every day to replacing it for my home and personal use. I will probably not buy another version of office unless it is super cheap ($50/included free on the PC I buy).

Not so unlikely

I'm not surprised by this conclusion. Think about it for a minute; the .Net system was only ever truly designed to run on Windows which means the run-time's roots can go deep into the inner workings of the OS - kernel-level HTTP servicing for instance (a win2003 feature I recall).

In fact, so tightly integrated are .NET and Windows Server 2003 that I believe Microsoft were going to call it Windows Server 2003.NET at one point until they dropped the idea and opted for plain "Windows Server 2003" instead.

That by design would give the extra performance gain, but of course lose flexibility too - something the article didn't explore much.

Essentially this is just another "flexibility vs. performance" argument here.

Wrong side of the coin

While performance is important, it's not even remotely the most important factor when deciding upon an API stack. It's much more important to have an environment that will help/let you code quickly. And other factors like flexibility, correctness, security, and reliability generally come before performance.

What a completely useless set of results

In fact, what a completely useless concept for article.

Hardware's cheap. People aren't, and the business certainly isn't

For any platform, chances are you're going to look at a whole bunch of variables:

• What your developers are used to if you're developing your own application (or commissioning someone else to).
  
• If you're getting your application(s) from elsewhere (either COTS or Free), what platform they need.
  
• What current systems you have and (if necessary) how easy it will be to integrate the application with those systems.
  
• What systems you have experience in running.
  
• If the application will be crucial to the business, how confident you feel in betting the business on it. This is where Windows often wins - in the minds of the many CTOs who don't consider "I've posted a query to a mailing list" an appropriate action to take when hitting a problem, or who don't want to discover 18 months down the line that a crucial security fix requires upgrading part of the platform which in so doing will break the application (I'm talking about you guys, PHP).
  

Performance is going to come way down near the bottom of the list unless the difference is absolutely huge and cannot be made up by throwing some extra money at either a bigger server or a number of servers, which is likely to be significantly cheaper than the cost of retraining all your staff or hiring a team to write to a specific platform.

It depends on configuration and how you measure it

It depends on a lot more than just "is it IIS/.Net or is it LAMP?"

On equivalent hardware, with equivalent RAM, if you're running MySQL with the MyISAM engine it will blow away SQL Server performance for most queries, but at a cost: You do not have stored procedures or transactions. When you switch to InnoDB you gain those but take a performance hit and that advantage over SQL Server disappears. On the plus side, MySQL is free/free (unless you need the commercial license - which in turn depends on whether you bundle it with your application AND how you interface with MySQL AND the "license" of your software).

As far as .Net itself vs. php or .jsp - it depends largely on several things:

- What is the architecture and how efficient is the code? If you use, say, DotNetNuke as an example, it's a good argument AGAINST using .Net because performance (of at least 1.x.x versions, I haven't worked with it since) absolutely sucks to almost any PHP application.

- How much RAM can you throw at it, and is caching appropriate for your application? With fully dynamic web sites where pages may display random content (rotating images, randomized quote of the day, etc.) unless you override the caching mechanism, the cache can effectively break your web site. However if caching IS appropriate and/or you override it when necessary, .Net can provide blazing performance. Your code has to be very efficient and you often need to throw a real large amount of RAM at it. Also, when you start a .Net application, it can take a while for it to serve up the first page - a concern for Tuesday updates when you were forced to reboot for patches to finish installing.

- (related to above) are you comparing a poorly-coded, inefficient, beastly PHP application to a lightweight highly-optimized .Net application? Say, OS Commerce vs. a custom single-purpose .Net eCommerce application? I'd darn well expect .Net to blow away the LAMP solution, because OS Commerce is an inefficient beast - the only thing going for it is is that it's a "swiss army knife" open source eCommerce application. It does EVERYTHING, but sacrifices efficiency due to its hacked-together design. (yes, I'm using "hacked" as in "that guy is a hack"). On the converse, if you compare, say, DotNetNuke to mambo or Drupal and set out to provide the same or similar functionality, chances are that the LAMP solution will blow away .Net by any load test metrics you can come up with.

What is my point? Unless you are comparing apples to apples, it's FUD or at best an amateur comparison. The way to test it is to implement the same task with a similar (as possible) architecture on each platform, with the application on each fully optimized, with both IIS and apache/tomcat/whatever fully tuned and streamlined to pull out every bit of performance from it, then load test both of those using the same sorts of tests. Comparing a highly-optimized single-purpose application to a general-purpose portal platform is not a fair test.

Also: Even though the article says "Even the most ardent PHP fans will admit that PHP is not designed with performance in mind," PHP performs damn well considering it's heritage, the fact that its primary platform is "a patchy server" and that it is FREE.

Why would one choose LAMP over .Net? Licensing. Hardware costs. Have you ever set up a Windows cluster? It's not cheap, and even if you can come up with a cheap way to do it, you don't WANT to because it will be unreliable. SQL Server licensing is EXPENSIVE (or if the db is real advantage of Microsoft's .Net platform? The development tools, NO one has an IDE which is better than Microsoft's. Zend's PHP Studio is darn good, but the level of integration with documentation, the debugging environment, the code completion, and keyboard navigation does not match Visual

Pure and utter Bullsh*t.

Comparing Sharepoint to Zope is beyond silly.
Zope is an object relational application server, making it slower than anything else running standard DB's. Technologically wise Zope is ten years ahead of Sharepoint - this is payed for with performance hoging and heavy-weight memory usage. 2Gigs is not enough for running Zope/Plone in a serious production enviroment.
Sharepoint is a monolithic built-to-fit solution that was grown over the course of almost a decade and finally has turned into something that doesn't crash every odd hour and - at last - performs the way it was supposed to back in 2001.
Keeping in mind that Zope was allready working back in 2001 and actually hasn't changed all that much since then. The entire redo - Zop 3.0 - still is in developement.
Sharepoint is usually used for CMS purposes, while Zope is usually used for highly abstracted business application developement.
Nobody in his right mind would get the idea to build an ERP system with Sharepoint.

Bottom line:
These guys didn't know what they where testing.

Wish they had tested open source .NET apps

I wish they had tested open source .NET apps against open source non .NET apps. You really can't run sharepoint on public sites very easily or cheaply. There are a number of open source .NET portals they could have tested including
DotNetNuke http://www.dotnetnuke.com/ [dotnetnuke.com] VB.NET
Rainbow Portal http://www.rainbowportal.net/ [rainbowportal.net] C#
or my favorite
mojoPortal http://www.mojoportal.com/ [mojoportal.com] C# and also can run on linux with Mono

Cost

To do a fair comparison I would like to see the Cost of the systems as set up.

To test the .Net stack, we ran Windows Server 2003 R2, SQL Server 2005 and SharePoint Portal Server 2003. Across the board, this configuration performed very well, with the top overall average throughput (by far) at 4.59M bps.

Quick check.....
$2,792.00 (Froogle Directron) Microsoft Windows Server 2003 R2 Enterprise, 25 Clients
$5,489.18 (Froogle Non Academic) SQL Server 2005 Complete
$5,619.00 (MS Website Retail) SharePoint Portal Server 2003 Server License with 5 CALs

$1,124.00 (Dell) Suse Enterprise Linux 9 With Server Hardware
http://configure.us.dell.com/dellstore/config.aspx ?c=us&cs=555&l=en&oc=MLB1580&s=biz [dell.com] Couldn't find Suse Enterprise 10 Integrated LAMP Stackhttp://www.novell.com/products/linuxenterpris eserver/lamp.html [novell.com]

Hmmm, Could train a couple of Windows Admins with $11,000. Better yet just Hire a good Linux Admin.

To a large degree, we credit this strong showing to the high level of integration that exists among the components of this stack. While most of the open-source and Java systems are developed independently of each other, each of the .Net components is designed specifically to integrate and perform well together. Even if the .Net stack had bombed convincingly in these tests, it would probably still maintain popularity in many companies.

Some people (PHBs) will never come around.

But its strong showing should give companies confidence that the .Net stack will handle most high-level enterprise needs.

For more than $12Grand it better blow away the Free Alternatives and configure itself and require zero admin.

I know I will get slammed for not using TCO but I don't believe those numbers at all. In my experience it takes the same amount of time for day to day maintenance. And when there is a problem (and there will be, no matter which one you choose) It costs me less time and therefore money to bring back up the Linux box.

Cost is not the only factor in a buying decision but is a factor, and if performance is arguably equal than it is a huge factor.

Con Kolivas

http://members.optusnet.com.au/ckolivas/kernel/ [optusnet.com.au] I wonder how Con's patch would do in these tests. There are two patches, one for servers, and one for desktops. They make the box uber faster for its purpose, but it sucks ass for the oposite. Get the patch for the server, and the server is snappy. The patch isn't perfect, however. I use it for my desky. It's swift.

MS ASP.NET VS Mono+Apache ASP.NET

Something I've wondered about and have started playing around with on my dev boxes is the performance/capabilities difference between running ASP.NET in an MS environment (win2k server + IIS, it's what I got) VS running ASP.NET on Mono (Linux (Ubuntu/RedHat) + Apache) if the performance is similer MS may soon find that they have totaly lost control of ASP.NET and it is now in the OSS wild.

a curious mix of flawed logic

A curious mix of flawed logic, marketing waffle and technical language.

"Probably most surprising was the solid performance that came from the .. mix of a Windows server and open-source components .. businesses should seriously consider the combo for their enterprise applications."

fud.alert: LAMP runs better on Windows.

Why would anyone move to Windows to use Open Source? Don't you still have to pay per simultaneous connection.

"Microsoft's .Net stack performed very well in our tests, clearly showing the benefits of the tight integration among each of the stack components"

How does 'tight integration', which is a function of how easy the sysop maintains the system, affect the efficiency of a running 'stack'. Does the stack know it is better 'integrated' and therefore runs like a happy bunny?

"JBoss Portal is relatively immature .. JBoss Portal on Windows performed considerably better than JBoss on .. CentOS"

fud.injection: JBoss on CentOS is immature. JBoss on Windows is better.

"we credit this strong showing to the high level of integration that exists among the components of this stack. While most of the open-source and Java systems are developed independently of each other, each of the .Net components is designed specifically to integrate and perform well together"

fud.injection: open-source and Java don't perform well together. Open Source runs better under Windows. Oh please Mr. Manager don't move off that Windows boxen.

"Neither the open-source nor the Windows communities seem to be able to accept a marriage of open-source server components and Windows operating systems"

What licensing restriction do the current ms.EULA put on Open Source projects developed with\and for Windows? Name any possible benefit that would be obtained by running Open Source on Windows? And don't mention the ease of use GUI. A proper sysop writes scripts to maintain the system.

"there are many IT managers attempting to run open-source .. on Windows .., no doubt, to the benefits and efficiencies of using open source without having to become Linux administrators"

How by any logic is it easier on Windows? This totally fails the logic test. Apache on Windows requires the same kind of config as Linux. Name any Open Source app that is easier to maintain under Windows. Provide concrete examples not opinion.

"JBoss on Windows far outpacing its Linux brethren"

I'm sure Marc Fleury would be interested in how Microsoft managed to get JBoss running faster on Windows.

"Enterprise IT managers shouldn't hesitate to look into the option of deploying open-source stacks on a Windows Server platform."

Yea, remember you still have your yearly tithe to pay Redmond. That's five seperate times in that article that you advised people to stick to 'open-source' on Windows. I do believe we have now all fully gotten the sub.text.

it's not "stacks", it's portals

What they tested were "portals":

We used portals we consider popular--Microsoft SharePoint Portal Server 2003 (built on ASP), XOOPS (PHP), Plone (Python), and Liferay and JBoss Portal (JSP).

Now, I know that Plone is a dog, and XOOPS may be popular on Sourceforge, but I don't think it's the most obvious choice for building a high performance portal using PHP. So, using these two as the basis for testing is silly.

The fact that JBoss Portal on Windows outpaces JBoss Portal on Linux has a simple reason: JBoss isn't fully open source; crucial parts of it (namely the Java runtime itself) are under Sun's control, and hell will freeze over before Sun bothers to do a good job implementing Java for their competitors' Linux systems.

As for things generally running faster on Windows, that's implausible. Differences between raw Windows and Linux system performance are at most in the single digit percentages, so if they saw any significant differences between the same applications running on top of the two platforms, either the application vendor spends more time tuning for Windows (as in Sun Java), or the testing labs screwed up.

In fact, the whole test is really ill conceived: none of the "portals" they compared provide the same functionality; it just doesn't make sense to test them against each other. Overall, this test mostly seems to test the competency of eWeek, and they aren't doing too well.

Where's the tests

I've now read the 'article' 4 times, and can find no description of the tests anywhere. Sure, there's a blog mentiones, but there's nothing anywhere! HOW ON EARTH are readers supposed to add to these results.

Shall I now go off and write a fictional article about how wonderful LAMP is and desctibe the results of the tests I ran on my (purely finctional) quad opterons that I have racked up at home???

There is absolutely no credibility to this article. How did it ever get published?

Big Deal

Apache bloated : Film at 11!

I thought Portals were dead years ago

The concept of "Portals" was abandoned several years ago despite Oracle and Microsoft arriving after the party ended. Why are we having a technical discussion about portals now?

The 2003 article Is the Portal Dead? [dmreview.com] discusses Gartner's announcement that "the portal is dead, long live the portal." And more recently, Portals are Dead [typepad.com] reiterates that. But of course, there's plenty more about the death of the portal [google.com].

Actually RTFA

Well, most of it, anyways. It appears that the /. article linked to page 2, and the graphs [eweek.com] are linked from page 1.

I've played with Plone a little bit, and it is resource intensive, to say the least. However, when you look at their graphs, eweek ran plone under both Windows Server 2003 and Suse Enterprise Linux. Given that they used the built-in Zope application server as the web server for Plone under both Windows and Linux, I would expect the performance to be equivalent.

When you look at the graphs, Plone on Windows appeared to outperform Plone on Linux by an order of magnitude. Something smelled funny. Like debugging.

While I'm not sure how Suse configures their Plone packages, by default, the Zope packages come with debugging turned on, which cripples performance. If you look at Chapter 2 of the Plone Book [neuroinf.de] by Andy McKay, it states:

By default in Zope 2.7 debug mode is enabled. Note that Plone runs significantly slower in debug mode, approximately 10-20 times slower. To turn this off, add the following line to the configuration file:

debug-mode off

To make the out-of-the-box experience more impressive for Windows users (debug mode slows Plone down on Windows even more than on Linux), it ships with debug mode off already. If you have a Plone site running and want to know if debug mode is running, go to *portal_migration* in the ZMI and look at the variables listed there; this will tell you if debug mode is enabled.

If I were running an enterprise which needed to use something with the features and robustness of Plone, and was about to devote the hundreds (or thousands) of hours required to fill it with content, and tweak it to my heart's content, I'd read the [expletive deleted] documentation, and notice that I might need to turn off debug mode. Sure, eweek said that they wanted to keep everything untuned:

But the point was to test the stacks, not their ideal performance points, which is also why we didn't tune or optimize any of the systems but ran them as close to default as possible.

Too bad that they didn't turn Zope debugging on in Windows, just to be consistent.

This is not a complex tuning or advanced configuration issue. You don't need to use eye of newt, or sacrifice small animals on the night of a full moon to make this simple change. If debug was left on in Linux, it not only invalidates their results, it also shows their conclusions to be utter garbage. A big part of their conclusion that open source software worked better on Windows was based on the Plone example (the best "apples to apples" comparison in their entire test). Eweek said:

Probably most surprising was the solid performance that came from the stacks that contained a mix of a Windows server and open-source components.

Probably most surprising was the solid incompetence that came from the testers, and the failure to configure anything other than a Windows server in spite of readily accessible documentation on setting up these complex systems. The sad part is that some IT managers will rely on these flawed results.

Hard to take the article seriously

I'm finding it kind of hard to take the article seriously.

What is an IT Stack? Is that where you go on a team building exercise and have to make a pyramid (with sys admins at the bottom)?

or quotes like this:
A few years ago, microsoft threw around the .Net moniker so aggressively in so many areas that it became difficult to figure out exactly what the term meant. But, as all the irrational exuberance that comes with a failed marketing blitz finally pulled back, .Net went back to being what it was originally intended to be: the name of Microsoft's server and service stack.

First sentence is fine. Second isn't. .Net is not "the name of Microsoft's server and service stack". It is an application framework. What is a "server and service stack"?

In fact, why has eWeek decided to call everything a "stack"?

Equally (and they do acknowledge this), it seems to be much more a test of particular applications.

I understand the need to dumb things down a little, but seriously.

Re:What a farce

I think any test can be criticised; main reasons are usually tainted environments or lack of tuning. What is interesting here is the conclusion, and since the article made it to the slashdot/* crowd, it might probably someday get to IT managers / decision makers

Re:And if...

And if the title read "OSS Web Stacks Outperforms .Net" I would bet it would not say anything like this The article's conclusion is that .Net outperforms the open source stacks, mainly because of its tighter integration, but also notes that running the open source stacks on Windows (WAMP) delivered strong performance.

It is important to note that strong is not stronger and, in fact, could mean weaker. Couldn't it?

Also, in our imaginary world, the article might say: The conclusion is that OSS outperforms the .Net stack, mainly because of a stronger development model, but also notes that running the .Net stacks on OSS is impossible.

Re:Not so fast

The article mentions that the Linux test system was "untuned." If this means "out of the box, running a kernel compiled for i386 and without any network tuning" then these results are hardly informative.

Well, they don't tell us the overall performance, but they do tell us what a basic install will do. In theory, enterprise linux distributions should be detecting the CPU and installing the proper kernel image/modules/etc. NT manages to configure itself properly no matter what [supported] CPU is in the box.

In fact, although I don't believe it either one of Microsoft's bullet points for selling NT is that it's "self-tuning".

Re:Like saying 'A Ferrari outperforms a Mini'

That's a poor analogy. Microsoft software is a Ferrari, and by comparison Open Source is a Mini? I don't think so.

A better analogy is this, because they refused to do any tuning on the OSS technologies: They bought a Mustang GT and a Nissan 350z and put them in a race. Then they told the driver of the Nissan that he could only use first and second gear. It would be nice if the Nissan was so dramatically superior that it could win even with that handicap, but since even a Ferrari would lose to a Mustang GT with just two forward gears at its disposal, it didn't happen.