Review Plan Requirements


In the planning step, you estimated your average and peak loads. Always base capacity plans on peak usage projections . This provides you with the capacity for peak loads and, of course, also supports your average needs. During the planning phase, in addition to estimating your peak load, you specified a target objective for response time. In this capacity planning exercise, you take the site usage projections and plan requirements and map them to the performance test results.

Review Load, Throughput, and Response Time Objectives

Begin the capacity planning phase by revisiting the requirements specified in your test plan. You want to understand the load, throughput, and response time requirements you specified in this plan. Response time is a critical objective for your capacity planning and usually becomes the limiting factor in your capacity planning. Your web site may support a large number of users without failure on a relatively small system; however, if this configuration produces poor response time, you must upgrade the hardware. Be prepared to spend to support good response times.

Start your capacity plan by filling out a Hardware Sizing worksheet from Appendix A if you haven't already. Capacity planning also requires some of the calculations from this worksheet. We'll use the same test plan data discussed in the example in Chapter 6. Table 15.1 shows the first part of the worksheet.

Table 15.1. Hardware Sizing Worksheet: Initial Input Data
  Input Data Source Your Data
1. Concurrent users Line 12, Capacity Sizing worksheet 5,000
2. Throughput: page rate Line 13, Capacity Sizing worksheet 42 pages/sec
3. Throughput: requests Line 14, Capacity Sizing worksheet 252 requests/sec
4. Response time Marketing 2 sec

Incorporate Headroom

Next, decide how much headroom you want in case you misjudged the load or to allow for future growth. How much headroom to use depends on how confident you are in your performance test requirements and performance test results. If your usage projections came from your existing web site, you might choose a low buffer factor. Also, if you expect your user community to start small, you may use less headroom and rely on collecting actual production performance data to adjust capacity planning targets. However, if you anticipate an aggressive web site launch, you need to incorporate a significant amount of headroom.

For example, say you want to allow for 50% headroom. This means at peak load, you use only 50% of the capacity of your systems, or, if your estimates are incorrect, your systems should be able to accommodate twice the estimated load. Headroom also gives your web site capacity for expected growth. For example, if you plan based on needs for the first three months, but you expect a steep ramp-up in site usage, incorporate headroom into your calculations to meet these near- term requirements.

To incorporate a headroom buffer into capacity planning, proportionally increase the number of concurrent users and the requests per second required. Keep your response time objective the same. To support 50% headroom for the example shown in Table 15.1, double the requirements to 10,000 concurrent users and 500 requests/sec with the same two-second response time. Table 15.2, a continuation of the Hardware Sizing worksheet, shows the specific calculations.

Table 15.2. Hardware Capacity Worksheet: Headroom Objectives
  Input Data Source Your Data
5. Headroom factor Marketing/planning 50%
  Calculated Data Equation Total
6. Concurrent user requirement (with headroom) 5,000 users + 5,000/(100/50-1) 10,000
7. Throughput requirement (with headroom): page rate 42 pages/sec + 42 / (100 / 50 “ 1) 84 pages/sec
8. Throughput requirement (with headroom): requests 252 requests/sec + 252 / (100 / 50 “ 1) 504 requests/sec
9. Response time (with headroom) Response time = Response time 2 sec

The planning objectives from lines 6, 8, and 9 in the Hardware Capacity worksheet transfer directly to Part 1 of the Capacity Sizing worksheet, also found in Appendix A. Now, let's continue on to Part 2 of the Capacity Sizing worksheet.



Performance Analysis for Java Web Sites
Performance Analysis for Javaв„ў Websites
ISBN: 0201844540
EAN: 2147483647
Year: 2001
Pages: 126

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net