Speeding content delivery

The task of content acceleration is a challenge as old as the Internet, which earned its worldwide-wait reputation for a reason.

Although the Web is more capable today, content has grown in complexity, and consumer expectation is high. Government Web sites are dealing with a richer variety of content than ever before, from streaming media to Internet-based applications. The public sector's Web outposts must also answer to e-government and the call for expanded citizen service.

For example, look at the Education Department's newly renovated Federal Student Aid Web site. About 133,000 users visit the site each day to research schools, submit college applications and apply for financial aid. "It's such a diverse group of customers," said Adam Essex, special assistant to the Students Channel chief-of-staff in Education's Federal Student Aid program. But Essex said he has found one common thread: "high expectations for performance."

One answer to the demand for faster service is Web caching. The Web cache, essentially a proxy server, stores frequently requested content physically closer to requesting users, thereby accelerating response time.

Agencies can build their own Web-cache solutions, taking advantage of numerous products available from server manufacturers and specialized caching appliance vendors. Another option is to hire an external service provider. Content delivery networks (CDNs) provide the necessary infrastructure for a monthly subscription fee. Service providers include Akamai Technologies Inc., Cable & Wireless plc, Warp Solutions Inc., and Mirror Image, a subsidiary of Xcelera Inc.

Web caching traditionally has involved information that doesn't change much over time, hence it is called static content. But vendors and CDNs have expanded their offerings to deal with dynamic content, which is trickier to cache. In a related thrust, vendors also seek to move Web applications and dynamic content closer to end users.

Close to the edge

Agencies seeking to accelerate the flow of information have numerous options when it comes to the do-it-yourself Web-cache approach.

At the high end, a data center-class Web cache operates at the gateway between an organization's network and the Internet. Often described as an edge server, the Web-caching engine offloads often-requested content from one or more centralized origin servers. The edge server boosts the flow of information because user requests don't have to go all the way to the origin server. Caching solutions can also be positioned within remote offices or points of presence so that the consumer is that much closer to the content.

Server vendors such as IBM Corp. and Hewlett-Packard Co. offer caching products. Other vendors offer CDN solutions for enterprises and service providers in which caching is embedded. Cisco Systems Inc., Network Appliance Inc. and Nortel Networks Ltd. offer turnkey CDN products, said Henry Goldberg, a senior analyst with In-Stat/MDR.

Such solutions may involve a sizable infrastructure investment. Web-caching appliances start at around $1,000 for a remote office solution but can reach six-figure sums for enterprise gear.

The alternative to internal deployment is working with a CDN service provider.

The Office of the Secretary of Defense's Web program, which includes the heavily used DefenseLink site, is using Akamai's CDN. The organization's initial interest was optimization for streaming broadcasts, said Terry Davis, manager for the public Web program. But Davis said he has discovered other benefits, including the ability to block denial-of-service attacks.

The Defense Department's budget benefits, too. Employing a service provider, Davis said, is much cheaper than trying to internally replicate CDN capabilities. "To have the infrastructure in place to deal with the bandwidth alone, we would have had to spend almost $1 million in the first year," he said. The organization's CDN service contract is about a quarter of that cost, he added.

Education's student aid site, built and managed by Accenture, also takes advantage of a CDN. Kelly Tate, a partner with Accenture's Government Operating Group, said the Web site had been performing well, but the company recently implemented Akamai's service to improve it.

Akamai's EdgeSuite service caches static content, delivering information "much more quickly for Web site users," said Jeff Robinson, account manager for infrastructure services at Computer Sciences Corp. CSC runs the Federal Student Aid program's virtual data center.

Keith Johnson, vice president of the public sector at Akamai, said the government is increasingly accepting service providers. The company had a handful of accounts when he joined in February 2002. Now, the company has more than 35 public-sector accounts.

Beyond the static cache

CDNs and server makers hope to entice more customers with offerings that go beyond static content caching. Some vendors have launched initiatives to cache dynamic content and even applications.

Akamai is the main driver behind one such effort, known as Edge Side Includes. ESI is a markup language that breaks Web pages into fragments with a profile describing their ability to cache items, according to ESI.org. Fragments may be labeled as cacheable for days, minutes or seconds. Thus, information once thought to be beyond caching technology, such as stock quotes, can be cached.

Upon request, the fragments are assembled into an HTML page at the CDN's edge. Only the page fragments deemed impossible to cache are retrieved from the origin server. The result, proponents say, is faster delivery of dynamic content.

Although ESI is fairly new — the specification was submitted to the World Wide Web Consortium in 2001 — some industry executives believe it has already been superseded by other methods.

One knock against ESI is the time needed to retag pages to identify cacheable content. Paul Stolorz, director of product management for CDN at Cable & Wireless, cited "substantial overhead involved in setting up the dynamic capabilities."

Other attempts to address the dynamic content issue include AT&T Labs Research's Client Side Inclusion (CSI), which is not yet commercially available. CSI is backwardly compatible with ESI, but they differ from each other in that dynamic page assembly takes place at the browser rather than the edge server. When a user requests information from a CSI-enabled Web site, the browser automatically loads a JavaScript. The script lets the browser cache frequently requested page elements for assembly.

"Just [as] edge assembly reduced bandwidth from the origin server to the edge, we push the savings over the last mile," said Misha Rabinovich, technology consultant at AT&T Labs Research.

Other vendors, meanwhile, are emphasizing edge computing. This Java-based approach aims to migrate Web applications from the data center to the network's edge, prompting some CDNs to recast themselves as application delivery networks (ADNs).

Akamai has enlisted IBM's WebSphere application server in its ADN effort. WebSphere servers let Akamai's customers distribute their Web application workload to the network, which results in better response time, said Stefan Van Overtveldt, IBM's director of WebSphere product integration.

Government agencies, however, aren't snapping up edge solutions — ESI, CSI or otherwise. DOD's public Web program has not yet pursued edge computing, Davis said.

Robinson said CSC is evaluating both ESI and edge computing.

How far will agencies push to the edge? That remains to be seen.

Moore is a freelance writer based in Syracuse, N.Y.

Featured

Reader comments

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above