Caching in

As the Internet continues its explosive growth, industry is moving to offer

information technology managers alternatives to provide employees with quick

and easy access to World Wide Web site content without devouring precious

agency bandwidth.

Web caching stores copies of commonly accessed Web pages locally on a server

to eliminate the need to download them from the Internet every time a user

requests them. One industry analyst group predicted that the caching market

will grow from $100 million in 1998 to $2 billion by 2002.

Companies vying to tap this growing market are approaching Web caching with

various methods. Caching can be accomplished by running software over general-purpose

operating systems such as Unix or Microsoft Corp.'s Windows NT. Or it can

be done by using server or network caching devices, commonly called appliances,

to redirect Web requests away from internal Web servers to the cache.

Cisco Systems Inc. looks at Web caching and its Web-caching engine as a

way to enable users to scale their Web connections, said Bob Deutsch, engineering

director of Cisco's federal operations.

In addition to caching Web pages, Cisco's caching engine also can be used

to load balance across multiple servers, intelligently noting which servers

are faster or slower. Overall, the engine is designed to reduce the network

load and to better address the ever-increasing demand that users place on

the Internet, he said.

One of the problems that has long been associated with Web caching is keeping

the stored data fresh and newly updated on the actual origin server of the

page. Cisco's cache engine can be programmed to cache dynamic information,

such as stock prices, and designate that the information will expire after

a certain amount of time, such as one minute.

Another problem that has plagued the caching industry is how to redirect

traffic destined for a Web server to the cache. Often, users have to reconfigure

all desktops and browsers or they have to install an additional switch,

called an L4 switch to do this. However, Cisco routers have a protocol that

enables them to recognize that a user's request should be redirected to

the cache without requiring any reconfiguration of the end user's machine,

Deutsch said.

InfoLibria Inc. also offers an Internet device called DynaCache that distributes

Web pages around the Internet geographically closer to users, where pages

can be delivered quickly and without congestion.

Solom Heddaya, InfoLibria's vice president of research and architecture,

said DynaCache is unique because it can be installed as a server device

or as a network device. Most competing cache devices can be installed only

as a server. Installing Web caching as a server device requires additional

configuration of browsers or desktops or the purchase of an L4 switch, while

installing a network device requires only that it be plugged into the network.

The federal government is an ideal candidate for using Web caching to post

publications that may be in high demand, such as Independent Counsel Kenneth

Starr's report to Congress on President Clinton last year, Heddaya said.

Documents can be cached and made available to the public without subjecting

internal agency Web servers to the intense demand that those in the House

of Representatives and the White House suffered after the release of that

controversial document.

"Neither the computer at the House of Representatives nor at the White House

was designed to meet the level of load the Internet offered," Heddaya said.

"If there was a network infrastructure with caches all over the world, then

the Starr report would have been automatically replicated...within seconds

of its release."

Network Appliance Inc. has taken yet another approach to Web caching. The

company combined caching software with a stripped-down version of a general-purpose

operating system designed for a server appliance and came up with a plug-and-play

appliance called NetCache.

In addition, NetCache has built an architecture infrastructure into its

products to enable them to cache various types of traffic, such as streaming

media, that users will demand as the Internet grows.

CacheFlow Inc., which recently has partnered with Marzik Inc. to offer all

of its products on the General Services Administration schedule, offers

a caching operating system called CacheOS. Unlike caching devices that refresh

data from Web servers only when the administrator or the user requests it,

CacheOS uses a technology designed to work independently of user requests

to learn what the most popular objects and pages are and to check those

pages for fresh content.

"Content has to be fresh," said Dan Brigati, CacheFlow's federal region

director. "It's information that might be strategic. It's information that

might involve new federal regulations. It might even be images from an intelligence

perspective. That information has to be up to date."

In addition, because many user requests are stopped at the CacheFlow device,

desktop machines do not have to be continuously connected to the Internet

and are less likely to be accessed by a hacker or other unauthorized users,

Brigati said.

Security was one of the reasons officials at NASA's Johnson Space Flight

Center outsourced the Web hosting functions ? including caching ? of its

human space flight site. Kelly Humphries, manager of the site, said officials

realized that NASA did not have sufficient bandwidth on its network to handle

both internal operations traffic and public traffic to the site, which logs

10 million hits per week.

In addition, contracting out Web caching should insulate the agency's internal

servers from external users, who could present security threats.

The Food and Drug Administration's Center for Food Safety and Applied Nutrition

has been using CacheFlow's caching device since February. Laurence Dusold,

chief of telecommunications and scientific computer support at the center,

said that before installing the device, the center's site was receiving

heavy traffic when food safety alerts were issued and at other random times.

For example, last year the site received 83,000 hits in three hours after

the release of a report about contaminated raspberries.

"We were very unprepared for that at the time," Dusold said. "The caching

device definitely helps that situation, helps smooth it out. That has added

a very large amount of capacity to our site."

Pu Xiang, senior analyst with Gartner Group marketing firm Dataquest, noted

that CacheFlow's operating system effectively eliminates the paramount problem

of refreshing data.

"The CacheOS automatically goes back to the origin server to refresh itself,"

Xiang said. "When a user requests a page, they're guaranteed [that] the

content is fresh."

Akamai Technologies Inc. has taken another approach to Web caching. Instead

of offering a product, the company offers Internet content delivery services.

Through its network of 1,000 servers in 15 countries, the company has moved

content closer to end users and combined this with intelligent routing requests

designed to eliminate problems caused by Web congestion or bottlenecks.

The company, which began offering its services in April, counts some of

the most popular Web sites, such as Yahoo, among its customers. While Web

site operators are its customers, Akamai offers to install for free one

of its servers at federal government installations to bring government users

closer to some of the world's most popular Web sites.

"We basically rename select objects on a Web page so that when you visit

the Web page, your browser is directed to our network to obtain the content,"

said Kieran Taylor, senior product manager for Akamai. Peter Christy, vice president of the Internet Caching Research Center, said although many caching devices can deliver only 40 percent of Web pages requested from users, Akamai, through its dispersed servers, can deliver 100 percent

of the pages requested.

Not all agencies that have looked at caching technology have decided to

embrace it warmly. GSA is examining various technologies to better serve

internal and external Web users while maintaining bandwidth, but the agency

remains uncertain as to whether Web caching is the best answer, said Mark

Kaprow, program expert in GSA's CIO office.

"We try to track the activity on our Web server; how will having a caching

server affect that ability?" Kaprow said. "Maybe we would be better served

by load balancing between the West Coast and the East Coast."

Kaprow's indecision may represent a prevalent feeling among users in most

federal agencies, where Web caching still has not been widely deployed.

It appears that the industry still has a lot of work ahead of it to persuade

federal users of Web caching's benefits.

Harreld is a free-lance writer based in Cary, N.C.