The Internet phenomenon and its growth have been unprecedented in any other industry. The growth, in terms of users adapting to the Internet and bandwidth, outperforms even Moore’s law that was defined for the growth of the semiconductor industry. Internet traffic is doubling every four months and e-commerce revenue is expected to be $300 billion by 2002. Approximately 150 million people around the world are connected to the Internet as compared to a single-digit number just few years ago. More and more people have become dependent on the Internet for their daily communication and information exchange.

It took 50 years for the telephone industry and 25 years for the computer industry to reach maturity levels that the Internet has taken only six years to reach. This means the next generation of Internet users will have very little tolerance. They will want the Internet to perform with the reliability of a dial tone.

All this puts a lot of pressure on service providers and networking vendors to offer quality services. The problem needs to be solved together between content providers, Internet Service Providers (ISPs), and networking vendors.

Most of the content today is distributed to the end users from a single location. This mechanism of distributing the content from a single site worked well as most content providers did not like the concept of having their servers located in a far off place. Single site placements also offer single point of failures due to natural calamities, equipment failures, software bugs, etc.

End-users’ access speed is increasing with the advent of newer technologies. Just about three years ago 28.8 Kbps was the highest speed that one could get from home. Today modem speeds can go up to 56 Kbps. xDSL, which is becoming very affordable and popular, can offer speeds from 128 Kbps to 2 Mbps. Cable modems can offer up to 10 Mbps from the desktop. As the speed to the desktop increases, people’s desire to access the content at a faster pace obviously increases. With such high speeds and more users, the content that has to traverse through congested and more expensive long-distance lines causes higher latency as well as packet loss resulting in slower response times.

Today, there are many technology solutions that content providers can adopt to overcome this problem in a cost-effective way.

Servers Located at Distant Internet Data Centres

Redundancy and reliability can be achieved through highly robust data centres which are built to handle power failures, network failures and even monitoring of the servers for uptime using different technology-based solutions. However, if an entire data centre fails for reasons like natural calamity, the servers will not be accessible.

By having Internet servers located in multiple, geographically dispersed data centres, the content providers can achieve a way of distributing the local traffic as local as possible.

Intelligent Routing

Once the servers are located in several geographically dispersed data centres, the concept of intelligent routing will help to keep a request coming locally to be redirected to a local server. For example, let’s assume that a content provider has hosted one set of servers in North India, say New Delhi, and another set of servers in South India, say Bangalore, both carrying the same content. Now, a request coming from Punjab should be directed to the servers located in Delhi based on uptime and response time. Similarly, requests coming from Hyderabad must be directed to servers in Bangalore. Intelligent routing helps achieve this and thus balances the load. This feature also provides redundancy element to an extent when one set of servers are down the requests will be directed to the servers that are functional.

Content Replication

Once the servers are located in multiple locations to achieve redundancy and to balance traffic, it is important to make sure that the content in these servers are replicated very efficiently. Content replication products are just being introduced in the market. That means, as content changes on the origin server, these solutions will "push" the new content to various replicated servers almost immediately. An example of this would be news published by newspapers. If a newspaper web server has replicated sites and as the news changes it must be reflected in all the servers almost immediately.

Some of the products that are solving these issues are TIBCO, Webspective, Veritas, and DB Vendors.

As content gets replicated, it adds another dimension to the problem. If the sites use transaction servers, then the transactions need to be synchronized also. Most sites prefer to keep single site transaction servers to minimize synchronization issues.

Content replication can be made more effective if IP Multicasting is enabled on the network. This helps to use the network and load on the origin server more effectively, especially when there are multiple replicated sites.

The main drawback of content replication is the cost associated with replication and investment in several servers.

Caching

Caching is a very important technology for better delivery of content. In network caching, the content provider will launch cache engines at various locations where ISPs exchange traffic. Most frequently used content will very likely be cached, and hence, will be served to the end users almost instantly from the cache engine. The content that is not cached (such as dynamic web pages, DB transactions) will always be served from the origin server.

There are two major limitations of caching.

• Most sites would like to know who their end users are. This information is crucial to rate the sites based on number of unique users and hits. Most caching technologies have solved the problem by providing log information. This information must be integrated between multiple cache servers as well as the origin server.

• Caching is not useful if a site has mostly dynamic content. Although preliminary data pointers indicate that most sites can be cached up to 70 percent. In a dynamic web page up to 30 percent content can be cached.

Caching solution is provided by companies such as Inktomi, CacheFlow, Network Appliance, Cisco, etc., and is provided as a service by Akamai, Exodus’ Mirror Image, and Speedera. The companies that provide caching as a service will have caching servers located globally across many networks, and hence, the content that is most frequently used will be sitting as close to the end user as possible by giving the end user a great user experience.

Bandwidth

Bandwidth issue can be classified as "long distance" and "last mile". The long distance service providers who offer either clear bandwidth or IP services must do proper capacity planning to accommodate the demand and reliability aspects. Cost of bandwidth is inversely proportional to volume. As the volume increases the cost per megabit will drop substantially. Without enough bandwidth, this new industry will not take off.

In the "last mile", there are many ways of solving the problem. Existing 2-wire copper can provide up to 56 Kbps on regular modems or by using DSL can go up to 2 Mbps depending on the distance to the central office. Cable modems can also offer shared 10 Mbps. Wireless technology is evolving so rapidly that it can provide up to 2 Mbps today and perhaps up to even 6 Mbps.

As the speed of the "last mile" access increases, it is absolutely essential to provide justice to the end users to make sure there is enough bandwidth to provide better user experience. That will translate into users doing more things on the Internet and perhaps even see e-commerce activities really take off.

Summary

Internet is not just technology, but it is a new way we conduct our lives. As the Internet usage continues to grow exponentially,
content providers must use a combination of replication and caching. These solutions will help deliver the content more efficiently with superior performance and better user experience.