MARKETING WHITE PAPER

JUDITH BERCK

(Text only - final layout and printed collateral available upon request.)

____________________________________________________________________________________

Intel's Quick Web Technology is the smart, easy way to speed up your Internet connection. Quick Web Technology can cut a user's wait for web page content in half.

The Internet has come of age brimming with a fantastic amount of content. Home pages, news, e-commerce, live television from halfway across the globe - the choices for information and entertainment are astonishing.

But anything successful and popular creates its own set of problems.

Low bandwidth: The telephone lines that connect our phone systems together were designed to transmit voice traffic only. Voice information requires a relatively slow transmission rate, or bandwidth, of 9.6 Kbps to travel almost instantaneously over dial-up lines and retain high quality. But digital images and other graphics are much denser kinds of information which take longer to transmit. Neither the wires in the phone lines, nor the routers and switching centers that direct data over long distances are fast enough to transmit it efficiently. The bandwidth is simply to low.

The current solutions are not good enough. One could rewire the phone lines to increase bandwidth, but that would be prohibitively expensive for the telephone companies. One could use a faster, more sensitive modem to make the most of what's available, but there is an upper limit to modem speed and effectiveness. Modems are limited by electromagnetic interference from everything else running on or near the phone lines, and by the degradation of signal strength over long distances.

Latency: The Internet, or rather the multiple networks that comprise it, suffer from another problem as well -- high latency. It takes a long time for data to travel through the Internet. Why? For one thing, the telephone networks themselves are antiquated and limited because they began as local and regional networks. Internet requests for content may travel through many points around the world on the round-trip from the user to an Internet Service Provider (ISP) to a web content server and back again. The telephone networks transmit information over such long distances using routers, or switching circuits. Each of these "hops" slows the round-trip down.

Plus, the telephone systems are more congested than ever due to the exponential growth of the Internet. Web content servers can't respond to user requests fast enough to keep up with the demand. This is especially true with popular web sites, where the congestion is greatest and the servers are busiest. Frequent visits by multiple users can cause a virtual "traffic jam." Latency thus resembles traffic on a highway, where travel is slow regardless of what the bandwidth, or "speed limit" is.

The problem from an ISP's point of view is network traffic. It's not the ISP's fault that the telephone system it leases its lines from wasn't designed for the burgeoning Internet. But the ISPs must still contend with the problems produced for its customers, including long waits and busy signals.

Simply put, Intel engineers have found a way to cut the wait for web content in about half, without requiring any changes to the current telephone system. How?

The technology builds on the basic concept that you don't want or need every bit of data that you request. Think of the mail you receive every day in your mailbox. First, you sort through the junk mail and toss out the Grand Prize Winner notifications and the Venetian blind discount coupons. You save only the mail you want and need, such as letters and bills. But even now, you only want certain pieces of this mail. Many parts of the mail can be thrown out, such as envelopes, boxes and other packaging, or the money-saving tip sheet inserted into your electricity bill.

What if somebody could do this sorting and tossing for you before the mail even gets to your mailbox? Somebody who could divert the stream of mail headed your way from around the world, who could look closely at each piece and throw away the useless parts, so that only the most valuable material remained? They could sit in your neighborhood mail truck, or in your local post office. What if they could do this for everyone so there would be less mail to sort and carry in the first place? The end result would be faster delivery of what you really want and need.

If what you really want and need is faster delivery of web pages, Intel's Quick Web is that enabler. The results are clear: a study by Inverse Network Technologies found that the average page download waits using Quick Web were about half (47%) the wait without it. This translates into an increase in speed for the end-user by a factor of 1.9x (mileage varies depending on the particular connection).

In addition to reduced delays, Quick Web gives the user direct control over the speed of web content through a patented pop-up JAVA applet called the Dynamic Client.

The rest of this document goes into a more detailed explanation of how Quick Web technology enhances web download speed through real-time bitstream analysis, sophisticated data compression and caching, and at the same time gives the user direct control over the process.

The fundamental concept behind Quick Web software is simple. In standard web usage, when a user makes a request for web content, the request goes to one of the ISP's servers, which then "fetches" the content from the requested web site and sends it back to the user. But if a user has signed up to use the Quick Web service, all of his or her requests for web content go instead to a Quick Web-enabled server. The Quick Web server then acts as a go-between, making the content request to the web site, receiving the content and sending it on to the user.

The difference is what happens at the Quick Web server as it is receiving the web content. Like the hypothetical sorter at the post office, the Quick Web server pulls data out of the bitstream of data packets it's receiving, examines and modifies it before sending it on to a user. This modification can mean dropping bits to improve a user's download speed (compression). Or it can mean putting frequently-used bits aside in local storage (caching). It can also mean adding additional bits to the bitstream to enrich the user's web experience (the Dynamic Client JAVA applet). It does all this in real-time, or "on the fly," using techniques that comply with industry Internet standards.

How does Quick Web analyze the bitstream in real-time without slowing it down? Labels. Packets of data are "packaged" using the industry standard HTTP format and TCP/IP protocols. These protocols use certain standard labels and codes to give instructions for each packet. From the labels, Quick Web can quickly tell where a chunk or packet of data is headed, what kind of data is in the packet, and any special instructions the packet might have.

Taking bits out of the stream -- Content Compression

One of Quick Web's techniques is compression of large image files. Quick Web monitors the labels on packets in a bitstream of data and picks out the large GIF and JPEG image files. Using intelligent algorithms, Quick Web then selectively drops pixels out of the image, reducing its file size. This compressed version is sent out to the user. The user receives this version in about half the time the original version would take, because less data takes less time to move. This compression technique addresses the delay caused by low bandwidth: instead of increasing bandwidth, Quick Web reduces the need for it in the first place.

Quick Web uses "lossy" compression algorithms to drop a certain percentage of pixels from an image without drastically affecting image quality. For instance, Quick Web can look at the similarity of the pixels in an image and "see" pixels of black next to gray and white -- and drop out only the gray pixels. Or see that there are a lot of white pixels with one black pixel in the middle - and drop out the black ones. The percentage of pixels dropped from an image depends on the size and detail of the image -- it can range anywhere from none in tiny images to 50-75% for very high quality images.

Another type of lossy compression technique is used for animated GIFS, which are series of images tiled together like frames in a very short movie. Even if some of the frames are dropped, the animation is still comprehensible.

The same labels that indicate which packets contain image data can also specify that an image not be compressed. Quick Web is intelligent enough to distinguish these images and leave them intact, or obey any other special instructions.

What happens to image quality when the images are compressed by Quick Web? It is altered perceptibly, but not so much that the image becomes incomprehensible to the human visual system. If at any time a user would like to restore the original image quality, Quick Web provides a way to do it with the click a button (see below, "Adding bits into the bitstream-- the Dynamic Client).

Storing bits outside the bitstream -- Caching

Part of the wait for web content is caused by latency delay. As mentioned above, it can take a long time for data to travel through the Internet. This is especially true for the most popular sites, where congestion from frequent visits by thousands or even millions of users can cause a "traffic jam." Under these conditions, web content servers can't respond fast enough to keep up with the demand.

Quick Web addresses the latency delay through caching. Caching works this way: when a user makes a request for content via Quick Web, the content's compressed version is cached, or stored, at a Quick Web server. This is similar to the way a web browser caches content on your computer's hard drive. This way, other users who request the same content can receive a cached copy of the compressed version directly from Quick Web, instead of waiting for busy web content servers to respond to them. Caching reduces the travel distance and the wait, resulting in faster downloads for the user.

Quick Web caching technology is intelligent enough to observe directions for particular content, including instructions not to compress particular images, and "time to live" instructions. For instance, CNN's home page content may have a "time to live" instruction of only 30 minutes. Quick Web can determine when content in the cache is outdated, and replace it with "fresh" content accordingly.

A user can bypass the caching process by reloading or refreshing his or her browser. In this case, the request goes all the way to the server. Quick Web's cache is then updated with that content, and any further user requests for that content are updated as well.

Normally, web content providers keep count of how many "hits" their site gets by users. Quick Web keeps track of its cache hits and forwards this information to the content providers, so they continue to have an accurate count of how often their site is "hit".

Adding bits into the bitstream - the Dynamic Client

As explained above, the main increase in speed comes from compression, based on the principle that users don't want to spend so much time downloading images. But most users will want to see a small percentage of web images in their uncompressed original state.

Quick Web provides a way to do this through its patented Dynamic Client JAVA applet. The Dynamic Client is simply a window that pops up next to a user's browser at each session. It has buttons that allow a user to choose between two speeds: fast speed for the compressed version of images, or standard speed for the original, unmodified version. All a user has to do to change speed is click on a button and reload or refresh the browser. In this way, the Dynamic Client's interactive buttons give a user direct control in the process, and the final say as to what version of images he receives.

What makes the Dynamic Client "dynamic" is that it is written in JAVA, a special programming language tailored to the Internet which allows programs or "applets" to be downloaded onto a user's computer and executed as a user is working in real-time. This makes Quick Web very easy to deploy: there is no software a user has to install. The JAVA installs itself seamlessly and safely on the user's machine, once at the beginning of a session -- and uninstalls itself without a trace at the end of the session. No permanent effect is made on a user's system.

Authentication, log-in and user management

Quick Web's Dynamic Client uses standard authentication mechanisms to protect who has access to its benefits - at the beginning of a session, pop-up windows help a user choose a name and password. User preferences are then stored in a database for automatic reference the next time a user logs in.

Quick Web "sits" at critical junctures within an ISP's network. This allows it to optimize performance for the maximum number of users. The server it uses at these junctures is called a "proxy server," because it is acting as a proxy for the ISP - when a user requests web content, the request goes to Quick Web servers instead of to the ISP servers.

Intel's Quick Web technology has developed a way to reduce the World Wide Wait and increase efficiency. But the implications are larger than that. Quick Web has the capability to take out or insert bits into the bitstream in real-time, which raises exciting possibilities for many other applications of this technology.

____________________________________

Legal Stuff © 1997 Intel Corporation