100 concurrent users without DynaCache
Table | -3 shows the resource usage for the baseline test without DynaCache.
Table 9-3
| CPU usage | Minimum JVM heap usage | Hard disk usage increase | |
|---|---|---|---|
| Commerce server 1 | 46.27% | 300MB | |
| Commerce server 2 | 47.39% | 300MB | |
| Database server | 100% | 23MB | |
| Web server | 2.72% |
Resource usage in the test with 100 users without DynaCache
As we can see, the database server has reached its capacity with 100% CPU usage. Since, according to the test scenario, browse only accounts for 85% of the site activity, the database update volume should be very low, and the bulk of database activity should come from retrieval. To narrow down the most common or expensive types of retrieval, we next look at the response times per activity.
Table 9-4
| Web action | Average response time (second) | Count | Total time (second) |
|---|---|---|---|
| Overall | 6.524 | 35,459 | 231334.516 |
| Select shipping address | 0.465 | 447 | 207.855 |
| Add shipping address | 0.301 | 321 | 96.621 |
| Display category | 2.277 | 16,137 | 36743.949 |
| Log off | 0.29 | 795 | 230.55 |
| Log on | 0.508 | 473 | 240.284 |
| Go to log on page | 0.376 | 473 | 177.848 |
| Add item to cart | 1.235 | 1,078 | 1331.33 |
| View billing address | 1.14 | 865 | 986.1 |
| Process order | 1.368 | 92 | 125.856 |
| View order summary | 2.219 | 192 | 426.048 |
| Add payment info | 1.314 | 92 | 120.888 |
| Display product | 35.338 | 5,305 | 187468.09 |
| View shipping method | 0.618 | 863 | 533.334 |
| Display top category | 0.335 | 5,388 | 1804.98 |
| Visit store front | 0.2 | 2,128 | 425.6 |
| Register user | 0.61 | 405 | 247.05 |
| Go to register page | 0.373 | 405 | 151.065 |
Response times in the test with 100 users without DynaCache
From Table | -4 and Figure | -3, we can see that the response time for the product display page is huge (35.3 seconds) and the page is used 81% (=187468/231334x100%) of the total time. Without caching, the server has to retrieve product information from the database whenever a user browses the catalog. But product pages are user-independent, can be "shared" by users, and are likely to remain unchanged for days or even months. This makes the product display page an excellent first candidate for dynamic caching.
Figure 9-3 Total time for URL actions in the100 users test without DynaCache