FRAMINGHAM (05/01/2000) - Web caching products can improve performance and reduce costs-- but only if you know how to use them By Shari Weiss Delta Air Lines Inc. watched the load on its web servers and firewalls surge.
Recently added site-personalization features eliminated static webpages in favor of fresh, dynamically created content--but they also put the company's network infrastructure into overdrive. Rather than add extra servers and boost bandwidth, however, Delta took a more subtle approach--caching.
Today Delta's servers create commonly accessed pages once. The pages then reside on the company's new caching appliance, from which browsers can retrieve the pages multiple times--without increasing server loads. The results have been dramatic; the cache currently handles some 92 percent of the site's page requests.
Easing the server's life wasn't the only benefit. The cache also lives outside the corporate firewall, which helps cut back on the traffic through the company's security systems. The overall reduction in load eliminated Delta's need to purchase additional servers and firewalls, cutting overall equipment costs while simultaneously simplifying support. "Our return on investment was instantaneous," says Mark Kraieski, director of electronic commerce services for Delta.
Put simply, caching is the temporary storage of web objects (such as HTML pages) for later retrieval. External caching applications--such as Delta's--provide faster response times to end users and reduce internal server loads. IT departments can also deploy caching on an organization's intranet to accelerate performance and reduce bandwidth requirements for both internal and external sites.
The technology works. And according to a 1999 report by the Internet Research Group, cache usage is evolving rapidly in hosting environments such as internet service providers and web hosts. The real potential, however, lies in the enterprise segment, which the report's author--analyst Peter Christy--claims will represent two-thirds of the market by 2001.
EFFICIENCY EXPERT Implemented properly, caching makes more efficient use of local bandwidth, is less expensive than WAN bandwidth and reduces overall network infrastructure needs--benefits that should appeal to every CIO.
But while today's IT managers are likely to know about web caching from the hype generated on Wall Street by caching startups, many companies are not taking full advantage of this technology to solve their network challenges, says Greg Howard, principal analyst at The HTRC Group.
Part of the problem is that some companies get their caches more by serendipity than by selection. Many enterprise customers often see caching capabilities as a bonus that comes with some other product, such as a firewall or proxy server, says Christy. If the product makes a noticeable improvement and doesn't cause problems, he says, customers are happy and go on to other, noncaching issues without fully appreciating the technology's potential.
Product selection is critical, however, because caching implementation strategy depends on the problem to be solved, Howard says. For instance, if the issue is a flood of HTTP traffic inside an organization that is drowning the current network, then client-side caching is appropriate. Deploying caches close to users and allowing those caches to handle web requests both cuts response times and saves network bandwidth.
When the issue is how to accelerate content to a target customer base outside corporate walls, then server-side caching is often the best approach. In this case, the cache acts as a front end for the server farm and handles all inbound requests. Commonly requested objects, catalog pages, for instance, go directly to the browser from the cache, while live transactions, such as sales requests, pass to the back-end servers. Simply using the cache that came with your firewall or other network component might work fine--but it might not be the best choice.
A cache used badly may also introduce additional complexities to an already complex network. Pages can go stale if the cache doesn't refresh content from the source servers frequently enough, for instance. And inaccurate data delivered quickly will not solve anyone's problem, says Greg Govatos, marketing director at CacheFlow, which manufactures the appliances used by Delta.
Companies must also be careful to optimize caches so that users get sent to the proper ones. And IT departments must make sure that sensitive information--such as credit card numbers or proprietary corporate data--doesn't end up stranded in the open on an improperly configured cache.
CACHE RICH But not every IT exec is cache-ignorant. For instance, John Puckett, CIO of ToySmart .com, got the task of building the online toy vendor's network from scratch--in less than a year. And caching has become a critical part of the job. "Organizational infrastructures are just not adequate to handle the increasing loads on their systems," he says. "Caching is all about taking data and reducing the amount that has to go back and forth."
Puckett's company has tried different approaches to caching over the last few months and has now adopted a multitier, phased-in strategy. "Currently we are implementing a redesign of our caching solutions utilizing local caching servers as well as switching equipment to manage the traffic to and from the cache."
ToySmart's next phase will extend it into enterprise-level caching through services like Akamai, a company that provides businesses with high-performance delivery of rich web content and internet applications via distributed caching servers. (Akamai recently helped handle the flood of requests for FoxSports.com's Superbowl coverage and the October 1999 NetAid online benefit concert.) "This [phased] strategy will flush out technical issues with our site architecture and ensure that we are scaling a solution that works," Puckett says. "We did not want to be resolving problems in a large distributed environment until we fully understood all of the technical issues."
THE POWER OF CACHE Standard e-commerce transactions and webpages can be reason enough for caching. More advanced content--such as streaming audio and video--only increase the need.
Streaming media delivery becomes a problem when multiple end users try to access the same content at the same time. Typically, each user then receives a stream from the origin server, resulting in unnecessary congestion of both the server and the network. By using a cache, the server can send a single stream, with the cache--or caches--redistributing it to individuals, thus freeing server and network bandwidth for other activities.
When Merrill Lynch was looking for a cost-effective solution to speed the delivery of critical intranet content--including streaming media--to its international branch offices, it found caches ready to help. By deploying a caching appliance in each of the branches, the organization improved response times for the users while simultaneously saving on the high costs of international bandwidth, according to Peter Danka, vice president of international private client architecture at Merrill Lynch.
And if caching is important now, it will be critical in the near future. Within three years, most new software will be delivered as an internet-based service, according to Christy. "Any CIO who isn't trying to fully understand internet-based software solutions probably needs to retire early," he says, adding that caching will play a fundamental role in maintaining the performance benefits of distributed computing.
But even while analysts promote caching as the wave of the future, it shouldn't be taken lightly, warns ToySmart's Puckett. "In our quest to improve performance and reduce costs, we must ensure that we don't neglect scalability and, even more important, security," he says. "[Caching] is an area where CIOs need to be careful to not shoot themselves in the foot."
Is your website cache rich or cache poor? Tell Technology Editor Chris Lindquist about it at firstname.lastname@example.org. Shari Weiss, an Oakland, Calif.-based freelance writer, can be reached at email@example.com.
REVISIT EMBEDDED LOGIC Overhyped, then much maligned, expert systems find their niche inside web applications By fred hapgood In the mid-'80s a technology appeared with a striking marketing story: a product that promised to capture the expertise of skilled humans and then codify that expertise into a program. Transformed into such a program (so to speak), an expert would be continuously available and ubiquitously deployable, even if he or she had long since left the company.
The defining property of these so-called expert systems was that they made it easy to define, express, monitor and alter sets of rules associating inputs with outputs. (Ordinary programs, by contrast, tended to bury these rules deep within the code.) A programmer would interview an expert, present various possibilities and code up his or her solutions as a set of if/then rules.
Thereafter a user entering an "if" condition would be shown the action that the expert would have taken in that situation.
Despite a huge hype attack by the media, the first five years of the expert system business were lean ones. According to an article published by CIO in December 1990, the problem was that these early systems were used to address overly complex business problems, were incompatible with the rest of the corporate computing environment and required their own experts to operate them.
Where was the logic in dispensing with the need for one expert by acquiring a need for another?
At the time, we were much more enthusiastic about the prospects for a simpler and more modest generation of expert systems appearing on the market. These new systems were being used for highly defined applications addressing everyday issues like credit verification, underwriting approvals, resource scheduling and network diagnostics. According to the article, the payback periods were measured in months or less.
Nonetheless, the adoption curve of the technology stayed modest. According to Nalini Elkins, CTO and cofounder of Applied Expert Systems, a network management tools vendor, one problem was the name. "If you called something an expert system, people's expectations rose. Companies would try it out on their toughest problem and usually the program would fail." Vendors could not just scale up the intelligence of their products, since programs that can deal with every situation take forever to build. Elkins says that as a practical matter, expert systems are best suited for routine cases; tougher problems should be kicked "upstairs" to human beings. "We talked a lot about how we could rename the technology," said Ulf Strom, CEO of Design Power, a company that uses expert systems to design work processes and mechanical and chemical engineering processes.
A second problem, according to Tod Loofborrow, president and CEO of Authoria, an expert systems development house specializing in human resources and health-care information, was that functions like credit verification required high levels of interoperability with other corporate information. Years of work were needed to achieve compatibility.
During the last decade, expert systems have become so integrated that they have turned into parts of processes. Vendors continue to sell them embedded into various products but seldom mention their presence. Programs are sold based on their functionality for a given application, not on whether they are "expert" or not. The new robot mind became part of the plumbing.
This self-effacement may have delivered them to their true role, that of web-based interfaces to corporate data resources. "Expert systems can personalize inquiries to the nth degree," says Loofborrow, whose systems let employees explore their benefits policies on the web. Akeel Al-Attar, of Attar Systems, thinks expert systems are a natural support system for e-commerce, since consumer "interviews" are rule bound and require much back and forth with corporate databases. He gives as an example a Japanese pump manufacturer and Attar Systems customer, Ebara Manufacturing, that produces several thousand kinds of pumps for many industries. Traditionally, customers would just ask for a pump. Sales personnel had to figure out what kind.
As products got more numerous and sophisticated, this manual system started to break down. Ebara fixed this problem with an online expert system. The system brings customers through a series of questions that connect their needs to specific products, often in less than a minute.
Al-Attar points out a subtle edge enjoyed by this new generation of outward-facing expert systems. One reason why the programs CIO wrote about 10 years ago did not enjoy the success we expected, he suggests, was that they were internal systems; all they did was reduce costs. E-commerce applications bring in revenue. While it might be true that a saved dollar contains the same number of pennies as an earned dollar, technologies that make money tend to get front-office attention. In the end, this may prove to be an even better marketing story than capturing the wisdom of experts.
Want to brag about an expert system? Technology Editor Chris Lindquist wants to know (firstname.lastname@example.org).
PREDICTIONS E-MAIL E-MAIL AVALANCHE If your organization's e-mail traffic isn't growing exponentially, then you're probably CIO at Luddites Inc. and you don't need to read this. The rest of you are likely seeing huge increases in both the volume and size of electronic messages. Yet according to San Francisco-based Ferris Research, IS departments are underestimating how much network bandwidth and e-mail storage capacity they will need to keep up with the increased e-mail flow. A recent survey of IS departments at 24 organizations found that most expected their message stores to grow by roughly 50 percent over the ensuing 12-month period. But given that new message volume is projected to grow from 60 percent to 80 percent during that time, message sizes will likely double or triple, and users are loathe to delete old messages. IS departments should really be planning for a 100 percent to 150 percent increase in storage and a three- to fivefold increase in WAN bandwidth.
Why are IS department e-mail estimates coming up short? David Ferris, president of Ferris Research, notes that most IS departments don't have adequate tools to monitor and predict e-mail capacity needs; he recommends that they deploy such tools to avoid future surprises. To get a handle on old e-mail storage, Ferris recommends using tools to manage, archive and delete messages. And the future may offer another solution: During the next five years, Ferris predicts e-mail storage will gradually become integrated with file storage systems, especially as e-mail becomes used more and more frequently to send and store voice, fax, graphics and other types of files. "It's just crazy to have these two separate hierarchical storage places where everyone is keeping information," Ferris says. -Sari Kalin UNDER DEVELOPMENT INTERNET SECURITY UNPLUGGING CYBERCRIME With internet terrorists assaulting organizations ranging from NASA to Yahoo, many CIOs wish they could just pull the plug on cybercrime. Now, a new, airtight method of safeguarding internal networks has the potential to turn beleaguered CIOs' wishes into reality.
Unlike firewalls, which filter out unwanted packets or redirect connections, Whale Communications' Air Gap technology physically isolates servers from the internet to keep hackers from getting their hands on critical information. The system uses a nonprogrammable appliance, called an e-Gap switch, to rapidly switch data in and out of a SCSI-based e-disk that features solid-state memory.
The system works something like a fortified checkpoint in which security agents can investigate travelers--some innocent, others terrorists--without the risk of a hidden bomb or other weapon wreaking havoc on the local landscape. In operation, one host writes data onto the e-disk. The switch then quickly disconnects the e-disk from that host and connects it to the other host, which reads the data. The arrangement shuttles information back and forth between public and local networks without ever connecting the two systems in a way that could expose sensitive information assets.
Elad Baron, cofounder and CEO of Fort Lee, N.J.-based Whale, says Air Gap's strength lies in its total disregard for protocols. "It uses no TCP/IP, no operating system and no physical connection between networks, thus eliminating many of the vulnerabilities with which firewalls have to struggle." He notes that Air Gap allows only a safe, narrow path for specific data or transaction exchanges, thereby preventing any protocol or operating system attack on a user's internal network.
Whale's initial customers, not surprisingly, include e-commerce, financial, insurance and research organizations, as well as several government agencies.
But Baron says the technology is suitable for any enterprise that is concerned with network security, particularly organizations that use the internet to link telecommuters and business partners into internal systems. "Companies can place their most sensitive webpages and user password databases behind Air Gap, away from hackers' reach, while still enabling real-time transactions," he says.
A basic system sells for $43,000, but organizations with a large number of transaction networks will have to install an array of units with load-balancing software in order to prevent throughput degradation. The technology can also be used in conjunction with a firewall to provide an extra layer of security.
Frederick Avolio, president of Avolio Consulting, a computer and network security consulting company located in Lisbon, Md., likes Air Gap's simplicity.
"It solves a real problem and does so with elegance," he says. -John Edwards NEW PRODUCTS THE SMALL SCREEN One of the hurdles to rolling out web applications on cell phones, PDAs and other internet appliances is translating existing web data and applications into a format that can be viewed on the devices' tiny screens.
Enter IBM's WebSphere Transcoding Publisher, software that dynamically translates (or transcodes) web text, images and other information into a format that internet appliances understand. Data and applications written in HTML and XML can be converted to wireless markup language (WML) and other formats; graphics can be converted to formats that are viewable on the device at hand or to a hyperlink. The server-side software is priced at $20,000 per server. For more information, call 800 426-4968 or visit www-4.ibm.com/ software/webservers/transcoding.
FOX IN A BOX Looking to simplify the way you connect an office to the internet?
NetWolves offers FoxBox, an all-in-one internet gateway that can handle internet connections via a variety of means, including dial-up, DSL, ISDN, cable modem, 56K modem, dedicated T1 line or existing routed network. The package consists of the hardware and software needed to offer firewall security, web browsing and hosting, caching, access control, e-mail and Domain Name System (DNS) and Dynamic Host Configuration Protocol (DHCP) server functionality. Options include VPN functionality and e-mail archiving. The system is administered via a web-based interface. Prices range from $3,400 to $7,100, depending on net access bandwidth. For more information, visit www.netwolves.com or call 813 286-8644.
MANAGE YOUR WEB When web applications are a critical piece of a company's business, keeping those applications running reliably--and monitoring their performance--is equally crucial. WRQ's Holistix Web Manager can help. The software monitors website components--individual CPUs, web servers, URLs and the like--for their availability, reliability and responsiveness at any given moment; reports can also track the web applications' performance over time. Web components can be grouped into categories of the organization's choosing--for example, all the hardware, software and URLs related to customer service can be grouped and monitored together. Pricing starts at $1,800 per managed server CPU; the price includes maintenance, upgrades, standard technical support and unlimited network devices. For more information, call 800 872-2829 or visit www.holistix.net.
CONFERENCES ON THE HOUSE Tired of paying teleconference service bureaus to hook up complex conference calls? Lucent Technologies has developed a conference calling system that can handle calls with more than 1,000 participants across multiple locations. Conferences can be scheduled in advance or initiated on demand; the system lets participants schedule conferences via conference management software, a web-based meeting manager or an interactive phone system. The Lucent Audio Conferencing System 500 offers 24 to 576 ports, and pricing starts at $24,000. The Lucent Audio Conferencing System 1100 offers 192 to 1,152 ports, and pricing starts at $340,000. Both systems are based on an audioconferencing system from MultiLink, a PictureTel subsidiary. For more information, visit www.lucent.com or call 877 582-2632, Ext. 563.
TURNING TABLES Static HTML tables can now be brought to life. InterNetivity offers a new web-based service called Databeacon.com that adds interactivity to HTML tables. The service is based on InterNetivity's Databeacon software, a data analysis and reporting application. Here's how it works: When a website is linked to the Databeacon server (it takes a few lines of HTML to make the link), the server compresses data into the HTML table. The compressed data is sent back to the user in an HTML page, where the data can be sorted, analyzed, exported or printed. There is no charge for the basic service. A monthly $500 subscription fee lets publishers use secure socket layer (SSL) connections and gives their processing requests priority. The company also offers customization and the ability to host Databeacon.com on other servers, for an extra charge.
For more information, visit www.databeacon.com or call 613 729-4480.