I'm just going to jump in here quickly and flat out call this the lamest, most ignorant, and most ridiculous attempt at trying to make it seem like Cherokee is not the web server it claims to be.
Firstly, do you have a fucking clue how to even run a benchmark. Static files??? STATIC FILES!!! What is this, 1996? Secondly, as others have pointed out, do you honestly understand the fact that running a benchmark on localhost will result in nothing but completely useless data? You're not even touching the network card as the requests are being sent and received via a loopback device, and due to the heavy caching of files on the file system by the kernel, you're likely not even hitting the hard disk after the first request. Honestly, do you even know what goes into the reading, processing, rendering, and transferring of a request before, during, and after a request/response process has completed? What you're doing, in essence, is removing ALL of the bottlenecks that web servers have to deal with during each transaction such as network latency, bandwidth restrictions, disk I/O latency, compression/decompression, SSL termination, reverse proxies, reading/writing/rewriting headers that go beyond the typical headers found when requesting HTML, CSS, Javscript, and/or other static files such as images, directory and file handlers, security, etc. etc. etc. (and that list of etc.'s goes on for MILES!) The bottom line: Go read a fucking book about how hardware systems, operating systems, network systems, and each and every one of the application stacks that sit on top of them actually work. Then go and write a benchmark based on a real world use cases that go beyond static files which contain: <html> <head> <title>This DOES NOT REPRESENT A REAL FILE SERVED BY A REAL WEB SERVER</title> </head> <body> <h1>Apache and nginx are a trillion times faster than Cherokee according to my tests because I have no clue what I'm even doing.</h1> </body> </html> ... which tells us what? That if we all would stop being so wasteful with how much markup and content we use then Cherokee would regain it's "rightful" spot as the /FASTEST/ completely useless static file server on the planet? Do you seriously believe this type of a benchmark says anything other than it's time for you to pick another profession? To everyone else who's been posting all of these worthless benchmarks with absolutely no basis on anything even remotely based on the real world of serving up web content to the masses, please do me one of two favors: Develop a benchmark that represents a real world, production web site which with each and every connecting client is serving up static CSS, HTML, Javascript, and image files, proxying requests for dynamically generated content to external processes, often times running on completely separate servers than the one directly handling the request and the response from the client., load balancing the requests for dynamic content using all sorts of sophisticated load balancing algorithms which require communicating with other nodes in the system to determine which node the given request should be sent to for processing, reading, writing, and rewriting headers based on all sorts of optimization techniques, terminating secure requests between the client and server, compressing dynamic content on the fly to speed up the transmission of that content between the server and client, caching often requested files using sophisticated caching mechanism that incorporate ETags and other HTTP Request headers to determine if a file on the file system has changed or of the requesting client as the most up to date version, and a list that goes on for miles and miles and miles beyond all of this, all of which play a significant role in just how efficient any given web server is or is not at a given task. So, with all of this, I guess the above is my challenge to anyone else who decides to make a post proclaiming just how much faster their favorite web server of choice is over Cherokee: Take at very least the above list of real world operations that each and every web server must handle which each and every requesting client, develop a benchmark that incorporates all of them based on content served up from a real world web site that receives a minimum of 100,000 unique visitors a day, and then come and tell me just how much faster your web server of choice is. Until then, shut the fuck up. Seriously, you're doing nothing but wasting the good Cherokee project developers time who instead of building and extending the feature set, fixing bugs, and in other forms making the Cherokee web server that much better, have to instead defend themselves from people who assume that because they can turn on a computer and configure a web server, they are somehow and in someway the premier expert on whatever subject they decide to make such claims about. Please stop. You're not helping. You're hurting. So please. Stop. Oh, you want to know the secret to knowing if someone is full of shit: Google their name in quotes > http://www.google.com/search?q="Eric+Lapouyade"; < and see not only how many total results you get (236), but how many of those results are related to the same exact topic (9 of the first 20 are related to Apache is faster than Cherokee, and by the time you reach page four [http://www.google.com/search?q=%22Eric+Lapouyade%22&hl=en&start=30&sa=N] the results are reduced to a grand total of 37 hits, the remaining 199 considered duplicates to the preceding entries and therefore not worth your time looking through them for fresh content. I don't mean to slam you, man, but you should really consider building your street cred ^Up before attempting to slam the hard, life long work and dedication of others with FAR MORE street cred that you may ever gain (especially after crap like this) vDown. I won't have it. I can assure you of that. On Tue, Jul 28, 2009 at 3:39 AM, Eric Lapouyade<[email protected]> wrote: > Hi all, > > I am still benchmarking Cherokee 0.99.20 on Ubuntu 9.04 on an Intel i7 940 > (4 cores) > > One month ago, I found that Apache2 was faster than Cherokee (14 000 Req/s > vs 10 000 Req/s for serving small static files with a 'ab -n 1000 -c 20') > As I was surprised, Cherokee claim to be the fastest, I was wondering wether > I have a bad linux environnement. > So I tried Nginx, and this is really surprising : > > For a small static file, Nginx is able to do 25 000 Req/s (ab -n 100000 -c > 20) and even 59 790 Req/s with the keep-alive (ab -k -n 100000 -c 20) !! > > In the same time, I tried to send keep-alive requests to cherokee (ab -k -n > 10000 -c 20) and I discovered that performance is falling from 10 000 Req/s > without keep-alive to only 501 Req/s with keep-alive !!! > > So, with keep-alive, on ubuntu 9.04, Nginx is really 119x faster than > cherokee : > > As I have seen some people in the mailing list having same kind of > slowlyness on Ubuntu 9.04 I wonder whether there is a referenced problem > with this Linux : do you know ? > > Regards, > > Eric > > _______________________________________________ > Cherokee mailing list > [email protected] > http://lists.octality.com/listinfo/cherokee > > -- /M:D M. David Peterson Co-Founder & Chief Architect, 3rd&Urban, LLC Email: [email protected] | [email protected] Mobile: (206) 999-0588 http://3rdandUrban.com | http://amp.fm | http://www.oreillynet.com/pub/au/2354 | http://broadcast.oreilly.com/m-david-peterson/ _______________________________________________ Cherokee mailing list [email protected] http://lists.octality.com/listinfo/cherokee
