The number of transactions TPS (transaction per second) per second executed on behalf of, a transaction may be calculated based on the number of complete test period. For example, the user performs six transactions per minute, the TPS is 6 / 60s = 0.10 TPS. At the same time we will know the response time of the transaction (or beat), this example, 60 seconds to complete the six transactions also represents the response time of each transaction or beat 10 seconds.
Little laws (Little's law):
The law by Professor MIT Sloan School of Management (MIT Sloan School of Management) of John Little, in the year of 1961 and presented proof. It is about a simple mathematical formula and in advance of the product relationship, this law is to improve the direction of lean production points the way.
Little rule as described in Equation: Lead Time (output time) = Availability × number of tact or TH (production efficiency) = the WIP (stock number) / CT (cycle time)
PS: Later, we'll list the load model application of the formula Little laws.
We look at how this law is the action that occurs in a production environment by two examples of view.
Example 1: Suppose we have developed concurrent server, concurrent access rate is: 1000 client / min, each customer spends on average 0.5 minutes on the server, little's law in accordance with the rules, at any time, the server will assume 1000 × 0.5 = 500 customer volume business processes. I assume that after some time, due to the increased customer base, concurrent access rate up to 2000 client / min. Under such circumstances, how do we improve the performance of our system? According to little's law rules, there are two options:
First: concurrent processing server improve traffic, i.e. up to 2000 × 0.5 = 1000. or
Second: reduce the average processing time of the server requests the client, i.e., reduced to: 2000 × 0.25 = 500.
Example 2: Suppose you line up to visit some scenic spots, the scenic spot is fixed Capacity: 60 people. The average time for each person in the scenic spot to stay is: 3 minutes. Suppose also ranked ahead of you have 20 people, ask: Do you estimate how much time you are about to enter this and other scenic spots.
Answer: 1 hour (3 × 20 = 60), and the number of receiving fixed regardless of interest.
Little by law to study the load model, we should first understand two factors: response time (Response time) and beat (Pacing). Actually beat surpass affect the response time of the TPS.
Example 1: Beat 0 seconds, 0 seconds think time
5 and the user performs a transaction response time per transaction is 10 seconds, it takes 50 seconds to complete the transaction 5, i.e. 5/50 = 0.1 TPS (TPS is here controlled by the response time).
Example 2: the rate of 15 seconds think time 0 seconds
5 and the user performs a transaction response time per transaction is 10 seconds, but the response time is greater than the actual beat because, it is better than the response time of the frequency control transaction occurs. 5 to complete the transaction needs to 5 * 15 = 75 seconds, to produce 5/75 = 0.06667 TPS.
In the second example, the average response time is less than the beat 15 seconds, 75 seconds required to complete 5 iterations produced 0.06667 TPS.
The above two examples, we assume that think time is 0 seconds. If you think time is 2 seconds, the total time is still 75 seconds to complete five iterations, resulting in 0.06667 TPS.
Beat is 0 seconds, the number of users = TPS * (reaction time + time Reflection)
Beats is not greater than 0 seconds and the response time and the think time, the number of users = TPS * (rate)
TPS is in fact the rate wrt transaction time, it is also called the throughput (throughput).
Little rule so as to explain the load model: the average number of users in the system throughput = average response time *
N = ( R + Z ) * X
N, the number of users
R, average response time (and possibly rate)
Z, think time
X, throughput (e.g., TPS)
Such as: N (number of users) = 1500, R (average response time) = 10, Z (think time) = 0, then X (throughput) = 1500/10 = 150 TPS