CMS caching problem

[swest]

I have a news site with roughly 100,000 articles, when I add a new article the headline link appears in the category list first, but if you click on the headline link to the article, the article is not viewable for around 30 mins. Is this normal?

[ergophobe]

No, that's not normal. If the link appears, the article should be available.

[swest]

Would a site with 100,000 articles and roughly 7000 unique visitors per day need caching?

[BillyS]

>>Would a site with 100,000 articles and roughly 7000 unique visitors per day need caching?

Not with a good host. That's not a lot of visitors.

[ergophobe]

Like BillyS said, under normal conditions, no way. If your peak hour is 600 visitors and if each one views five pages, that's only one page view every two seconds. So unless you have a really slow script, that shouldn't be a problem.

The relevant numbers are not number of pages on the site or number of unique visitors. What you need to know are

- number of raw page views (preferably at peak times)
- time required to render a page.

This would include bot traffic. So if you have some bad bots, they could be requesting thousands of pages and that won't even show up in something like Google Analytics.

So the question is, do you believe you have bottlenecks that need caching, or are you just doing it prophylactically in case you get slashdotted or dugg?

[swest]

Thanks for the replies, helpful.

[swest]

ergophobe,

Pageviews at peak times: 900
Typical page size: 380kb

Page download times, according to online speed checker:

ISDN 128K 37.89 seconds
T1 1.44Mbps 16.62 seconds

[ergophobe]

Page size is only one factor. The main thing is how long does it take your software on the server to create that 380 KB page. It's more a matter of number of DB queries, how complicated the JOINs are and things like that.

So if you want to test how much you can serve up, you want to look into something like Apache Bench or a host of other free tools that let you stress test your server and figure out how many concurrent users it can handle.

If you find that your code takes a long time to render, you typically want to profile it to figure out where the bottlenecks are (for PHP, for example, you use xDebug + CacheGrind and it tells you how long each piece of code is taking so you can figure out where your holdups are).

Finally, if you are finding that you're having bottlenecks, you can also look for slow DB queries. For MySQL you can generate a slow query log so you'll see what activities are killing the server.