Professor Armando Fox:
From Bit Plumbing to Information Delivery -- Internet Applications Grow Up

     The internet is ubiquitous in our everyday life. A person can do almost everything he or she wants online. It seems logical that the next step is to make access to the internet as easy as possible. Indeed, Professor Armando Fox and his research group here at Stanford addresses this problem. In a talk that he gave on Thursday February Fourth entitled "From Bit Plumbing to Information Delivery: Internet Applications Grow Up", Professor Fox described the next generation of internet applications.

Background
     Professor Fox began his presentation with a brief history of the internet. The internet started as a government funded military project in 1969 during the Cold War. At that point, it was called the ArpaNet and consisted of four key sites connected by phone lines. The goal of the project was to create a network that would not go down even if one or more of its routers were hit. The TCP/IP protocol was then introduced in 1973 which allowed different networks to communicate with each other. Larger networks could be formed from clusters of small ones. After U.C. Berkeley released a free version of TCP/IP in 1983, the backbone of the internet could be put in place. Beginning in 1993 where Tim Bernes Lee at CERN proposed the World Wide Web - linked text and images where everyone in the world can be a publisher and consumer of rich data -- this backbone has grown into what we know as the Internet today.

THE NEXT STEP
     So what's next for the Internet? Professor Fox believes that in a "Post-PC" world where Palm Pilots and wireless communication devices such as cellular phone dominate our lives as much as PCs, the next project is to give these devices access to the Internet. The Internet has become a mass market phenomenon and "access is the killer app." Access to the information infrastructure has become just as important as the computers and hand-held devices.

ADAPTATION BY PROXY
     Several problems must be overcome before this next step can be taken. Most of all, the Internet was not made for these devices. The information format on the Internet is tailored to the high end PCs and modems. The information must be translated before it can be understood by pagers and cellular phones. In his lecture, Professor Fox proposed adaptation on the fly by proxy. A proxy is an intermediate between the internet and the client device. As an example, a proxy can take the big screen presentation of a web page -- the way a web page would normally be seen on a computer monitor -- scale down the images, throw away the color, reformat the text and make the web page appear on a much smaller device, such as the Palm Pilot. Adaptation by Proxy has several advantages: the clients do not need to change and the servers do not need to change. Designers for web pages can continue to make pages for high end PC users and manufacturers of hand-held and wireless devices can continue with their own internal formats.

Properties fo the Proxy
     One may then ask 1) is adaptation by proxy fast enough? and 2) what kind of properties must the proxy have? Professor Fox believes that adaptation by proxy is fast enough. Since proxy are used mostly to scale down big-screened images - make web pages appear normal on a much smaller screen with much lower resolution - the delay from the operation performed by the proxy should not be significant compared to the network latency required to download the page. In his experience, proxies can perform the necessary translations in one to two seconds, a delay that users rarely notice because it takes much longer for the page to download across the network. The second question is slightly more difficult. Because the Internet is so essential in providing information, access has gained the status of being mission critical. If proxies were to be used for translation, they must be reliable and running 24 hours a day 7 days a week. In addition to being reliable, a proxy must also be scalable since internet traffic is growing exponentially.

     In answer to these questions, Professor Fox argued for the value in exploiting cluster computers. Cluster computers are basically a locally networked set of computers used as a supercomputer. This architecture provides incremental scalability, essential to accommodate the growing clientele, and hardware redundancy, essential for reliability. In order for a cluster to act as a proxy, the cluster must contain three attributes: they must be highly available; they must be robust against bugs; and they must have an interface which enable users to easily write applications.
 
     As part of his research at U.C. Berkeley, Professor Fox built a network that satisfied all the requirements. SNS (Scalable Network Systems) is a cluster of computers that acted as a proxy and translated information from the internet into palm top devices. SNS consisted of 3 to 10 UltraSparc servers. The cluster harnessed and released machines as they were needed. Some applications that ran on SNS were a web accelerator and the first graphical web browser for the palm pilot. In the duration of the research project, SNS has grown from supporting just a few users to supporting more than 15,000 Palm Pilot web browsers.

     In conclusion to his talk, Professor Fox mentioned several projects related to his area of interest. One of them is a project in U.C. Berkeley, Ninja - infrastructure for scalability internet services (E. Brewer, D. Culler et al) and another is Mary Baker's project at Stanford, Mobile People. He added that all these projects are trying to solve the general problem of intelligent delivery of information and with the layer of the proxy, this problem can be solved for the internet and palm top devices.