PEOPLE, PLEASE CHECK OUT MY NEW BLOG.
So i had a dream that Bruce Lee looked at me and said:
It is not the technology that matter, it is the professional administering it.
I have been thinking about this for some while now. A good sys admin will use which ever tools in his disposal to make what YOU want happen. Well, this is a sweeping overstatment!, but there is some truth to it.
Keeping that in mind, it would be a nifty idea to create a competition for sys admins which consist of a well defined problem, some hardware, and some time to put them together. We have had programming competitions in university, but they were very algorithmic in nature, and very pgorammy.
I think the dude from plentyoffish.com would do a good job at them!. Geees, 2 million hits an HOUR, with ONE WINDOWS WEB SERVER!. How the hell (checkout plentyoffish.wordpress.com)!!!!
There few types of sys admins, i’m talking about the ones who design systems. Not just maintain it. Usually they differ by years of experience and Seniority.
Okay people, i often heard the words “Requests Per Second” with respect to a CMS installation on a server. After doing a BIT of work, i learned what tools will allow you to measure that. Which could become very useful if you are trying to see how well your server bEe hIvE under high load.
There are few tools out there, i’m going to talk about one:
AB Which abriviates Apache Benchmark. Very simple command line tool which just does the trick quick, and easy. Here is how i use it.
ab -n 100 -c 5 http://www.example.com
- -n: Is the number of hits which it will generate.
- -c: Number of threads which it will run to simulate the number of people.
You can get fancy with this by using the -p switch to specify a file which contains post values. So
ab -n 100 -c 5 -p fancypostfile.txt http://www.example.com
where your fancypostfile.txt looks like this
?action=edit&post=32
?action=delete&post=52
?action=dothis&post=432
Okay, so assume that you would like to add a username and a password authentication. You can do that by using the -A switch like so.
ab -n 100 -c 5 -p fancypostfile.txt -A username:password http://www.example.com
You can also play around with jmeter and grinder as an alternative to AB. But they are probably for more advance use. Unless your doing something of a REALLY large scale, ab will suffice.
UPDATE:
You can run this from any shell promt around the world!, HOWEVER, i recommend using this test on your hosted server machine because it will eleminate the factors of network delays etc and will allow you to evaluate your application more clearly.
You would excecute these commands on your shell accounts. So you would have to have an SSH username/password/server and you would something like Putty to login and excecute these commands (If your a windows user). Another thing i would like to make clear, THIS DOES NOT SPEED UP YOUR APPLICATION. It will simply tell you how fast/slow it is right now!.
Okay!, so few interesting stuff here!. So apperently plentyoffish.com runs on a microsoft server!, so mentioned in plentyoffish.wordpress.com. Also, he is doing two million page views an hour on ONE machine!. I mean, i know i’m new to all of this, but i KNOW this is really good!.
Filed under: Uncategorized | Tags: Scaling, Social Networking, Web Development
Okay, so i’m involved in a project where i have to scale wordpress MU to 50,000+ users. I’m new to this whole scaling game, but i do get the bottom line of it and i have enough skills to make it happen. So i did a bunch of reading, and here are the options which i found out of googling for an hour. PLEASE, we NEED FEEDBACK 🙂
1: The most simple solution was presented in this MU support thread, As showen in the diagram on the side. His main stratagy is to install “out of the box” MU on a whole bunch of machine, and to find a way to distrubte the traffic between all of them using apache rewriting rules. To achive that, he is using a local DNS which would be a MUCH better idea than using IPes. This idea is simple to implement and it requires NO messing around with any code.
2:Â The second way (dont have a picture for it yet) which i’m thinking of doing is a bit more involved but it does make more sense to me. So, you will have a load balancer in the front, distributing the load among X amount of web servers. The webservers will have nothing but apache and MU on them. They will store no information other than sessions. Mounted to these servers, we’re going to have file servers which have LOTS of space on them, this is where all the user files will be kept. And finally the DB’s, this is where most of the work is going to happen. THe idea is to create LOTS of databases, and to determine what user should be using what DB based on thier user name. Best explained by Matt.
It would also be possible to have some sort of domain based key with random distribution, for example you could take an MD5 hash of the incoming username, then use the first character of that hash to go to one of 16 clusters or databases. If you grow beyond that start using the first two characters, which increases your number of machines by a factor. Then when you grow more use the first 3 characters… 🙂
The second way requires fiddling with the source code. Which i dont mind doing, but i just want to make SURE that i HAVE to. 🙂
Virus In Training 