Reputation: 9863
I built an app in php where a feature analyzes about 10000 text files and extracts stuff from them and puts it into a mysql database. The code itself is just a for loop where every file is loaded through file_get_contents()
and after the end of that iteration, its unset()
from memory. The file analysis is a cron job and a single php file does all this processing.
The problem however is that the app was built (initially) entirely on a shared server everything worked seamlessly really well. I didn't notice any delays or major lags neither did users however in order for it to be able to handle more of a load, I moved everything to an EC2 server (the micro instance).
The problem I am having now is that every time I run the cronjob (process the files on hourly basis) it slows the entire server down so much that a normal page takes about 5-8 seconds to load, which sort of defeats the purpose of moving it to EC2.
The cron itself is a very long process. Here are some tests results of the script process (every hr)
SQL Insertion Time: 23.138303995132 seconds
Memory Used: 10.05 MB
Execution: 411.00507092476 seconds
But on the top of every hour the server slows down so much for 7 minutes despite of having more dedicated hardware acceleration compared to a shared server (I think at least). The graphs from EC2 dashboard show that the CPU usage is close to 100% but I don't understand how it gets to that level.
Can anyone help me determine the reason as to why this could be happening? I have noticed not even the slightest lag when the cron runs on the shared server but the case is completely different for EC2.
Please feel free to ask me anything I missed mentioning.
Upvotes: 0
Views: 234
Reputation: 1954
Well EC2 is designed to be scalable.
Since your code is running in 1 loop to open each file one after another, it does not make for a scalable design.
Try changing your codes to break them up so that the files are handled concurrently by different instances of the php script. That way, each copy of the script can run in a thread by itself. If you have multiple servers (or instances of servers in EC2), you can run them on different machines to speed it up even more.
Upvotes: 0
Reputation: 15159
Micro instances are pretty slow. If you use a larger instance, it'll run a lot faster.
We use EC2 for all of our production boxes. I can't say enough good things about that platform. I'll never go back to another host.
Also, if you want to write your code in C++, it'll run A LOT faster. I wrote a simple mysql insert with this code here. It's multi-threaded, so you can asyncronously run mysql updates or inserts.
Please let me know if you need any help with it, but I'm sure you'll be able to just use a micro instance still and get great speeds.
Hope that helps...
PS. I'd be willing to help you write a C++ version for your uses... just because it's fun! :-)
Upvotes: 2