as a php developer and a technical developer on a start-up team, choosing a development framework is a very difficult thing.
with thinkphp, you can get started by recruiting a developer who has just graduated from a training institution, but decoupling performance and post-code is a headache. however, many third-party functions do not need to be written by yourself, as many experts have already paved the way.
with laravel, legend is very comfortable to write and has sufficient scalability, but the learning cost is a bit high. you can’t give junior developers half a month to learn the framework. and it is said that people in the world have revealed that laravel’s performance is not very good, and the documentation is not particularly rich.
with yii, the syntax is a bit verbose, and the separation of front-end and back-end code is a little troublesome, but the performance is very good, and it is already used by many large domestic companies. if something unexpected happens, you can quickly find an expert to answer your questions.
it’s better to do a simple performance evaluation. you can’t go wrong by choosing one with excellent performance
performance test
test time: may 6, 2016
test tool: siege (because the error pr_socket_recv: connection reset by peer (54) always occurs in ab of mac system, so i chose siege)
test machine: macpro core i5 processor/8gb memory/256gb ssd flash memory
test environment: apache php5.6
framework version: thinkphp 3.2.3 laravel 5.2 yii2.0.5
testing principle: each test cycle is 5 times, and medium data is taken
additional note: all projects are placed in subdirectories
explanation of results
transactions: 2119 hits 处理请求总数量 availability: 96.85 % 可用性 elapsed time: 9.74 secs 运行时间 data transferred: 1.31 mb 数据传输量 response time: 0.60 secs 响应时间 transaction rate: 217.56 trans/sec 每秒处理效率 throughput: 0.13 mb/sec 每秒处理数据 concurrency: 130.28 并发 successful transactions: 2162 成功的请求 failed transactions: 69 失败的请求 longest transaction: 2.85 最长的单个请求 shortest transaction: 0.01 最短的单个请求
start testing
add a controller and enter helloworld 10,000 times in the controller
concurrency 50 cycles 10 times:
the first thing that came to us was thinkphp, which looked okay
transactions: 500 hits availability: 100.00 % elapsed time: 2.81 secs data transferred: 52.45 mb response time: 0.26 secs transaction rate: 177.94 trans/sec throughput: 18.67 mb/sec concurrency: 47.10 successful transactions: 500 failed transactions: 0 longest transaction: 0.48 shortest transaction: 0.03
then laravle, well, a little disappointed
transactions: 500 hits availability: 100.00 % elapsed time: 13.33 secs data transferred: 52.45 mb response time: 1.27 secs transaction rate: 37.51 trans/sec throughput: 3.93 mb/sec concurrency: 47.55 successful transactions: 500 failed transactions: 0 longest transaction: 3.64 shortest transaction: 0.07
yii, who was relatively optimistic before the final game
transactions: 500 hits availability: 100.00 % elapsed time: 4.84 secs data transferred: 52.45 mb response time: 0.46 secs transaction rate: 103.31 trans/sec throughput: 10.84 mb/sec concurrency: 47.65 successful transactions: 500 failed transactions: 0 longest transaction: 0.88 shortest transaction: 0.04
200 concurrent loops 10 times:
first is thinkphp
transactions: 1977 hits availability: 98.85 % elapsed time: 10.03 secs data transferred: 207.40 mb response time: 0.95 secs transaction rate: 197.11 trans/sec throughput: 20.68 mb/sec concurrency: 187.68 successful transactions: 1977 failed transactions: 23 longest transaction: 1.22 shortest transaction: 0.02
then laravel
transactions: 1890 hits availability: 94.50 % elapsed time: 51.85 secs data transferred: 198.27 mb response time: 4.88 secs transaction rate: 36.45 trans/sec throughput: 3.82 mb/sec concurrency: 178.00 successful transactions: 1890 failed transactions: 110 longest transaction: 26.01 shortest transaction: 0.07
finally yii
transactions: 1996 hits availability: 99.80 % elapsed time: 18.95 secs data transferred: 209.39 mb response time: 1.79 secs transaction rate: 105.33 trans/sec throughput: 11.05 mb/sec concurrency: 188.57 successful transactions: 1996 failed transactions: 4 longest transaction: 3.29 shortest transaction: 0.10
unexpectedly, without optimization, thinkphp was the fastest, yii was slightly faster, and laravel finished the test falteringly. the results are somewhat unexpected, but the above test only represents the development environment. the framework will be optimized below to simulate the online environment.
optimization framework
thinkphp:
app_debug改为false
laravel:
app_debug改为false php artisan route:cache php artisan optimize php artisan config:cache composer dumpautoload -o
yii:
yii_debug changed to false
composer dumpautoload -o
concurrency 200 loops 10 times:
thinkphp
transactions: 1655 hits availability: 82.75 % elapsed time: 8.21 secs data transferred: 173.62 mb response time: 0.69 secs transaction rate: 201.58 trans/sec throughput: 21.15 mb/sec concurrency: 139.29 successful transactions: 1655 failed transactions: 345 longest transaction: 7.83 shortest transaction: 0.00
laravel:
transactions: 1520 hits availability: 76.00 % elapsed time: 34.95 secs data transferred: 159.45 mb response time: 3.15 secs transaction rate: 43.49 trans/sec throughput: 4.56 mb/sec concurrency: 136.84 successful transactions: 1520 failed transactions: 480 longest transaction: 19.18 shortest transaction: 0.00
yii:
transactions: 1704 hits availability: 85.20 % elapsed time: 15.16 secs data transferred: 178.76 mb response time: 1.46 secs transaction rate: 112.40 trans/sec throughput: 11.79 mb/sec concurrency: 164.21 successful transactions: 1704 failed transactions: 296 longest transaction: 9.04 shortest transaction: 0.00
what is strange is that when the framework turns off debugging mode, there are more failures. however, it is obvious that after simple optimization, laravel's performance has been greatly improved, but it is still lower than the other two frameworks.
conclusion
during the test, i felt depressed and felt that my world view had collapsed to the point of collapse.
thinkphp’s performance is twice as high as yii and nearly four times higher than laravel
the performance of yii is relatively mediocre, but in the test, it is obvious that there are fewer request failures than the other two frameworks
laravel is still elegant, but its performance is worrying. it is better to describe it in the words of netizens: those who learn will live, those who use it will die
finally
my project was finally developed using a self-built framework. thanks to the power of composer, the development process went very smoothly. finally, i attached the data of the self-built framework running 200 concurrent cycles for 10 times. the framework is open source but the update is quite busy. welcome to search ppphp on github and build the framework with me.
Transactions: 1672 hits Availability: 83.60 % Elapsed time: 6.18 secs Data transferred: 175.40 MB Response time: 0.57 secs Transaction rate: 270.55 trans/sec Throughput: 28.38 MB/sec Concurrency: 153.16 Successful transactions: 1672 Failed transactions: 328 Longest transaction: 4.57 Shortest transaction: 0.01