The reason development, the need for throughput (TPS), QPS, the number of concurrent response time (RT) to do the next few concepts to understand, self-check Baidu Encyclopedia, recorded as follows:
1. Response Time (RT)
response time is the time of the system to respond to the request. Intuitively, this index and the subjective perception of software performance is very consistent, because it is a complete record of the time of the entire computer system to process the request. Since the system will typically provide a number of functions, and the different functions of the processing logic vary widely, and therefore the response time of different functions are not the same, the same time is not the same even in the case where different functions in response to input data. Therefore, when discussing a response time of the system, it generally refers to the maximum response time is the average time for all functions or all functions of the system. Of course, which often need to be discussed and the average response time, or maximum response time for each function in each group.
For applications not concurrent operation of stand-alone, it is generally considered to be a reasonable response time and accurate performance. It should be noted that the absolute value of the response time does not directly reflect the level of performance of the software, the level of software performance really depends on the user response time of acceptance. For a game software, the response time of less than 100 ms should be good, the response time of about one second may belong to barely acceptable, if the response time of 3 seconds to a totally unacceptable. As for the build system, the full compile a large-scale software source code may need tens of minutes or even longer, but the response time for users is acceptable.
2. Throughput (Throughput)
Throughput refers to the number system of processing requests per unit time. For non-concurrent applications, throughput and response time to a strict inverse relationship, in fact, this time the throughput is reciprocal response time. As already mentioned, for a single-user system, the response time (or response time and system latency application) may be a good measure of performance of the system, but for concurrent systems, as typically required with a certain performance.
For a multi-user system, the average response time of the system only if a user is t, when n number of your user, each user sees a response time not usually n × t, while often than n × T is much smaller (of course, in certain circumstances may also be larger than n × t, and even much larger). This is because the processing of each request needs to use a lot of resources, due to the processing of each request, there are many concurrent execution does not go hard, which leads to a specific point in time, the share of resources is often not much. That is when a single request at each time point may have many resources are idle, when processing a plurality of requests, if the rational allocation of resources, the average response time for each user does not see the number of users increases increases linearly. In fact, the average response time of the system with different number of users increase growth rate is not quite the same, which is mainly employed to measure the performance of certain concurrent systems. In general, the throughput is a more generic indicators, two systems having different number of users and user mode, the maximum throughput if it is consistent, it may be determined processing capability of the two systems consistent.
3. The number of concurrent users
of concurrent users is the number of system users using normal functions of the system can be carried simultaneously. Compared with certain number of concurrent users it is more intuitive, but a more general performance. In fact, the number of concurrent users is not a very accurate indicator, because different users different usage patterns can cause a user issues a different number of requests per unit of time. A site system as an example, assume that the user can use after only registered, but registered users are not using the site all the time, so the specific time only a part of registered users at the same time online, online users browse the site will spend a lot time to read the information on the site, and therefore a specific time only part of the online users also issued a request to the system. In this way, we have three systems for site statistics on the number of users: the number of registered users, the number of online users and the number of users simultaneously send requests. Because of registered users for a long time may not go to the website, using the number of registered users cause great error as performance metrics. The number of online users and colleagues requesting the number of users can be used as performance indicators. In contrast, online users more intuitive as the performance index, and the number of users simultaneously requesting more accurate as the performance index.
4. QPS query rate per second (Query Per Second)
QPS QPS rate is a particular query server within the specified time the measure of how much traffic processing on the Internet, as the performance of the domain name system server machine is often used to measure queries per second rate. Corresponds fetches / sec, i.e., in response to requests per second, that is, the maximum throughput. (It appears to be similar to the TPS, only apply to the throughput of a particular scene)
[Reserved] link https://www.cnblogs.com/data2value/p/6220859.html