 |
A Methodology for Website
Performance Optimization Often the first approach to optimizing website performance is to add more hardware. This works as long as hardware is the limiting factor, but may have little or no effect if the bottleneck is really the database behind the site. Faced with rapid website user growth, a shortage of skilled technical staff, and the careful scrutiny of Venture Capital investors, many dot.com startups are struggling to find effective ways to optimize the performance of their websites. This is a growing concern among IT managers, especially those who oversee high-performance, high-availability, rapidly growing web-based systems. The ultimate goal of a system manager is to stay ahead of the growth curve while providing optimal service. Many issues can be resolved by a competent internal IT staff. However, knowing when to bring in outside technical experts and how to work effectively with them can mean the difference between quick and efficient solutions or unnecessary expenditures and possible system failure.
This paper details the Rare & Dear Performance Team’s approach to diagnosing performance bottlenecks, including the strategies and tactics it employs to resolve performance limiting issues in a quick and effective manner. The methodology is neither startling nor proprietary, rather it is a well-planned, focused strategy that they developed to intervene in performance-critical situations. Knowledge of a process for identifying problems, implementing short-term tactical solutions with longer-term system strategies, and understanding the critical success factors for website performance remediation, are fundamental for every IT professional.
The following points describe key success factors that lead
to victory . This section describes a typical engagement by the Rare
& Dear Performance team. As the first step in getting to know the Client system, Rare
& Dear provides the Client with a Questionnaire asking for detailed
information about technical aspects of the web system and its performance
characteristics. As the second step in
the process, the Performance team installs a proprietary Remote Diagnostic
package on the Client system to gather preliminary system and database
statistics. The data from a 24 hour
period is analyzed in light of the information provided by the Client. Both static and dynamic measurements are
taken to highlight OS and database activity as well as website response time
(from the users’ perspective). These
measurements are charted, graphed, compared with each other, and compared
against “reasonable” values. The result
of this is a report that indicates the most likely areas to investigate more
deeply for solutions to performance issues. After the Remote Diagnostics are run, a decision is made as
to whether or not the remediation work can be accomplished remotely or whether
is requires the Performance team to work on-site. If the team must work on-site, a command center is set up as
described above. The team is then
briefed by top management and introduced to key personnel who will facilitate access
to systems and information. For remote
projects this is accomplished via teleconference or even
video-teleconference. As typifies all dot.com projects, this activity needs to be
defined and implemented quickly, with maximum communications but minimum impact
on the Client’s existing systems and staff.
To get a quick handle on the current system performance with an eye to
identifying bottlenecks, the Performance team takes a broad look at a number of
different factors with one sub-team assigned to investigate at a macro level
(top down) while a second sub-team is assigned to investigate at the micro
level (bottom up). The entire
Performance team meets every two hours to share findings and fine-tune the
investigation. Any existing tools for gathering system performance data are
reviewed. A Client may have a wide
variety of tools in place and still not have the staff skilled enough to
identify a complex problem. The Performance team interviews key staff to gain knowledge
of the historical site performance and growth patterns, recent system changes
and planned system changes and additions. The Performance team initially surveys the entire system for
performance and capacity metrics. The
depth and breadth of the Performance Team’s knowledge enables them to quickly
focus on critical issues that eventually lead to optimal solutions. Their fresh, detached perspective supports
the Client’s staff to question all
assumptions and hypotheses formed by both internal staff as well as Performance
team members. A number of questions are asked of each application that is
having performance problems, for example: After synchronizing performance graphs from the various
sources, a picture often emerges showing a high correlation between certain
database activity and site performance degradation. A theory of the cause of the bottleneck, can be
hypothesized. More examination of the
apparent bottleneck will either prove or disprove the theory. One or more rounds of hypothesis and
examination may be required to clearly identify the problem(s). Often there is more than one problem that
contributes to poor site performance. Within a day or so, the Performance team will usually have
isolated the problem and provide a temporary work-around to achieve acceptable
performance (if the live site is experiencing unacceptable performance). After the Client has implemented the
work-around, the Performance team continues to observe system behavior over the
next 24 hours. If performance is
acceptable, they will develop both short term and more robust long term
recommendations to make the improved performance permanent. After a de-brief to Client management, the Performance team
will terminate the intense site scrutiny.
However, at the Client request, they may continue to monitor the
performance for a longer period to assure no other problems arise. A full report of the team’s findings and
recommendations is also provided to the Client. Recommendations may include hardware, software, development
methodology, operations, and even management approach. The Performance team is then in a position
to provide direction and oversight to Client staff in implementing any of the
recommendations. If you recognize the need for outside intervention, you must
provide full support down the chain of command to facilitate quick turnaround
and a good return on your investment.
By using an external team of database and system performance experts,
you can often leverage internal knowledge with an experienced and complementary
external team to implement a performance improvement plan that can save
hundreds of thousands of dollars in avoided hardware upgrades. © Rare & Dear, Inc. 2002
|
