Summary
| Date | Viewers | Samples | Avg | Med | 90%ile | Errors |
| 21/02/2021 | 1000 | 500,000 | 44.93 | 30.50 | 113.50 | 0 (0) |
| 19/02/2021 | 2000 | 1,000,000 | 45.69 | 30.50 | 112.50 | 7 (0.0007) |
| 20/02/2021 | 3000 | 1,500,000 | 61.02 | 30.50 | 117.95 | 4117 (0.027) |
| 22/02/2021 * | 3000 * | 1,500,000 | 48.25 | 30.50 | 112.50 | 45 (0.003) |
Viewers: Number of concurrent viewers
Samples: Number of response time measurements
Avg: Average response time (msec)
Med: Median of response time (msec)
90%ile: 90 percentile (msec), response time in a typical bad case.
Errors (Rate): Number of errors (Error rate %)
*：NetVirtue’s Resource Booster activated
Test Scenario
Using Apache JMeter, I measured response time for each web page of mjsf.com.au. Currently there are 50 static pages published (reachable from the home page). JMeter doesn’t visit contents out of MJSF site hosted by NetVirtue. So it doesn’t access movies as all the movies are out-sourced to external servers such as YouTube. It doesn’t access to third party pages such as “Raffle”.
The test scenario was set to simulate users behaviour. The JMeter visited each page one by one with interval around 10 sec (Gaussian distribution with 500msec deviation)
JMeter generates virtual web users up to selected number 1000, 2000 and 3000. They are increased from 0 to preset number in “Ramp-up time” so that the increasing rate was one per second. Each virtual user accesses all pages and repeats “visiting all” in “Loops” times.
To reduce the load on test client side, two test PCs were used. For example, testing 1000 concurrent user was actually done by two PCs and each one generated 500 concurrent users.
Detailed Reports
1. Concurrent Viewers: 1,000 ( 21, Feb 2021)
2. Concurrent Viewers: 2,000 ( 19, Feb 2021)
3. Concurrent Viewers: 3,000 ( 20 & 22, Feb 2021)
Detailed Reports:
before Booster 20/02/2021: PC1 PC2
after Booster 22/02/2021: PC1 PC2