Real Story Group. Make Better Technology Decisions.

Kas Thomas

Cache as Cache Can

24-Jan-2008

Tags: Web Content Management, Information Architecture

Despite what eager (often junior) developers may tell you, snappy website performance is hard to achieve when you're serving complex, dynamic content. This is especially true when a page contains multiple AJAX widgets and/or multiple portlets serving up personalized content

It may sound obvious, but a Web Content Management or Portal solution that demos well on a salesperson's laptop doesn't necessarily serve complex content quickly in production, under real-world loads. Many customers are finding this out the hard way as they try to scale systems that weren't designed with dynamic content, social apps, AJAX, or mashups in mind, let alone quirky traffic spikes.

In talking to vendors recently in connection with our Web CMS Report 2008, it seems many of them, as it happens, are revisiting their cache architectures for the first time in years (or in some cases for the first time ever).

Good performance is almost impossible without good caching, but unfortunately many CMS and Portal vendors seem to be stuck in 1997. The page-level caching models envisioned by HTTP/1.1 (RFC 2616) were plenty good enough to get the job done back then -- and still are, for many static-HTML scenarios.

But now pages are more granular and content bits are more dynamic. Just-in-time assembly of highly complex, componentized pages is becoming more of a norm. The advent of AJAX means that many pages are now making multiple asynchronous HTTP requests, behind the scenes, while a user interacts with a single page. All of this has led to something of a caching crisis.

Customers shopping for a Web Content Management solution need to be aware of the importance of caching, and understand that not all vendors have designed their spiffy-looking, "highly personalizable" systems with good caching in mind.

When shopping for a system, demand that vendors demonstrate acceptable performance of their product under load, with complex content being served up continuously to multiple concurrent users. Don't be afraid to ask vendors about caches and cache architecture. (Bring your best developers with you and have them ask the hard questions.) Be wary of metrics involving page-views per second. It matters a great deal what kind of page is being viewed. And always make sure you understand how cache invalidation works (your editors really care about this -- even if they don't know how it works).

A good overview of this topic for webmasters is Mark Nottingham's Caching Tutorial.

Our advice? Do a little homework here. Save yourself many bottles of aspirin later.