DB access on demand VS XML writing

stephdumais

Hi, I'm currently in the process of planning an application that needs to output XML for external websites, an RSS feed actually. The XML info is pulled from a database which is updated on an irregular basis. The external websites pull this info once a day.

My question is, should I setup a cron tab that will write to an XML file or should I simply have a script output the XML and make requests to the database everytime it is accessed?

My concerns are for performance and optimization of course.

Any suggestions?

CoderDan

I voted for the cronjob, but it does depend on a couple of things:
1) how often does the data change?
2) how important is it that the feed is up to date to a period more accurate than 24h max age?
3) how many feed requests do you get per day?

stephdumais

CoderDan;10923215 wrote:
I voted for the cronjob, but it does depend on a couple of things:
1) how often does the data change?
2) how important is it that the feed is up to date to a period more accurate than 24h max age?
3) how many feed requests do you get per day?

The data can change many times in one day while not at all on other days, but the 24h max age is not critical. i.e.: some data can wait for the next day before showing up.
As for requests, 3 websites will be feeding on this data, with at least a thousand hits each day.

ps: in my initial post I mentioned the external website pull the info once a day but that is innacurate, since that depends on their hits and is most certainly more than once a day.

CoderDan

Well, I'm thinking that due to caching of scripts, etc, you are best to calculate the data once, however I'd also look at a method to ensure that the expiration date of that file is sent, so that any requestors know that it is accurate from now until midnight (or whenever the script runs) - this will lower your load even more.

stephdumais

CoderDan;10923240 wrote:
Well, I'm thinking that due to caching of scripts, etc, you are best to calculate the data once, however I'd also look at a method to ensure that the expiration date of that file is sent, so that any requestors know that it is accurate from now until midnight (or whenever the script runs) - this will lower your load even more.

Thanks! I was hoping the better solution was the dynamic way but you make a valid point regarding server load.

CoderDan

well, it's gonna have to be both. a) don't generate it 1000's of times a day, and b) better yet, don't even need to serve it 1000's of times a day, if it's getting cached right