In a mass-market world, catching the attention of, and keeping prospective customers on a Web site is a mix of getting a number of things right. Recently, Gartner asked some clients to take part in a panel discussion about building scalable, high-performance Web sites. Here are the top 10 recommendations from those experts, in increasing order of importance:
10. If legacy systems work, don't replace them. As a general rule, spend time and money on new functions, not on old ones.
9. Build multiple Web sites. For example, Charles Schwab & Co has multiple sites for its different lines of business, such as retail brokerage, institutional brokerage, and retirement services. Another company uses multiple sites to balance the needs of registered users vs unregistered, occasional visitors.
8. Invest in proactive capacity planning and change simulation tools. Because user volumes can't be predicted, maintaining excess capacity is key. Charles Schwab maintains capacity at three to five times peak volumes, whereas AskMe.com, another panellist, assumes that 60 per cent of peak daily traffic occurs in a four-hour period. Furthermore, the panellists strongly recommended using site performance simulation tools to understand the impact of increased user volume and changes to the site.
7. Invest in real-time availability and performance management. Monitoring real-time site availability and response time as well as component availability and performance statistics is vital to understanding site behaviour.
6. Provide the appearance of 24x7 availability. Sometimes planned downtime can't be avoided. But designing the Web site to enable planned downtime for a portion of functions while the balance are available to users minimises the overall impact of an outage. As a result, users have access to the site around-the-clock but may infrequently be denied access to some functions.
5. Invest in business continuity planning. Don't wait until a problem occurs to decide on a course of action. Instead, plan ahead for unforeseen events that could interrupt site access or degrade performance.
4. Design the site for no single point of failure. Designing the application architecture and technology infrastructure with redundancy minimises the impact of component failures and resulting unplanned site downtime. Architectures should be n-tier, with redundancy in each tier from both a hardware and a software perspective. In addition, the physical data centre should be designed for availability to reduce the impact of physical infrastructure outages due to maintenance or failures.
3. Design the entire architecture for horizontal scalability. By partitioning Web site functionality into components that live on different systems, enterprises achieve higher levels of availability and 'pluggable' scalability. The design goal is to be able to add more capacity by deploying another server and load balancing a portion of the connections to it.
2. Don't ignore people and process issues. Good design does not eliminate downtime, because on average about 80 per cent of site availability problems are caused by people and process issues, not by technology failures. So, to mitigate downtime risks, become adept at hiring and training.
1. Clearly articulate business priorities. Employees must understand business priorities (such as "the customer is always right"). Breed a culture that encourages teamwork, effective leadership, and solid communication. Furthermore, panellists recommended instilling fun and camaraderie into the culture to keep employees focused on goals. Barb Gomolski is a research director at Gartner Group.