![]() (Note: In Reuters, Reuters will multiply the throughput obtained in the second step by 66.7% as the soft alarm line of the system, 80% as the hard alarm line of the system, and the limit value is only used to carry the peak of the burst.) Performance testing also requires selective testing of these two scenarios based on the actual situation. Network packets can cause bandwidth consumption and performance problems. For example, TCP_NODELAY is not enabled, which results in latency increases (see TCP_NODELAY for details). Latency increases at low throughput levels. Low throughput and network packet testing. ** Collect system data: CPU, memory, hard disk/network IO, etc., observe their curves and corresponding response times to ensure the system is stable. Perform 5 minutes with the throughput obtained in step 2, 1 minute with the maximum value obtained in step 4, 5 minutes with the throughput obtained in step 2, 1 minute with the permission value obtained in Step 4, and so on for a period of time, say 2 days. For example, with a 100% success rate (regardless of response time), the system can last 10 minutes of throughput.ĭo a Burst Test. So, this value is the performance of the systemįour, find the limit value of the system. For example, the CPU is stable and the memory usage is stable. ** Then collect CPU, memory, hard disk/network IO, and other indicators to see if the system is stable. For example, use the second step to Soak the throughput for 7 consecutive days. It is best to use test data from the production line. Test data, need to have large and small data of various sizes, and can be mixed. Second, find the highest throughput within this response time constraint. For example, the Reuters definition: 99.9% response times must be less than 1ms, the average response time is less than 1ms, and 100% of requests succeed. The following performance tests are basically from my old employer, Thomson Reuters, which makes real-time financial data systems.įirst, you need to define a system response time latency, TP99 is recommended, and success rate. Generally speaking, performance testing takes into account several factors: Thoughput throughput, Latency response time, and resource utilization (CPU/MEM/IO/Bandwidth.). ![]() For some critical systems, the number of successful requests must be 100%, no ambiguity at all. The tolerance of failure rates for performance tests should be very low. For example, I say my system can have 100,000 concurrent requests, but the failure rate isĤ0%, then, 100,000 concurrent messages is a complete joke. This should be understandable, if the request is not successful, there is no performance test. Why is response time throughput linked to success rate This means that we constantly test on different concurrency numbers to find the maximum throughput when the software is most stable. For example, when TP99 is less than 100ms, the maximum number of concurrent requests supported by the system is 1000qps. ** Therefore, the value of ** throughput must have response time to card.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |