Re: PHP Performance/Scalability

Hello

1. We tried increasing the number of php workers. It didn't help.

2. We increased the number of vm's (horizontal scaling) and that worked.
But I am more interested in Vertical-scaling. As management of
horizontal-scaling becomes cumbersome.

3. The DB server performed ok. It is not the bottleneck. In Java
application, the number of same queries gave excellent response times.

4. I tried increasing the memory of PHP but it didn't help.

5. The code is fairly simple (a few images, a simple query) so I doubt that
is creating the performance problem.

Btw,

We are using: Apache 2.4.41, PHP 7.2, CodeIgniter 3.1, MySQL 8.0.27
(Could CodeIgniter be the problem?)

Deepak
"The greatness of a nation can be judged by the way its animals are treated
- Mahatma Gandhi"

+91 73500 12833
[email protected]

Facebook: https://www.facebook.com/deicool
LinkedIn: www.linkedin.com/in/deicool

"Plant a Tree, Go Green"

Make In India : http://www.makeinindia.com/home


On Tue, Oct 26, 2021 at 4:58 PM shiplu <[email protected]> wrote:

> It seems like the average response time is at the expected level when the
> number of concurrent users is below 1500. Then it's better you increase the
> number of PHP workers. Now if you have W workers create 6xW workers. If
> each worker receives a max of 1000 requests then you achieve the target
> concurrency.  What does this worker mean depends on your deployment? If you
> have 10 workers behind 1 apache/nginx instance in a single VM, then either
> you increase that 10 PHP workers to 60 or use 6 VMS instead of 1. This is
> the easiest way to solve this problem.
>
> The hardest way is to line up all the components for a single request and
> then find which component is getting hurt when concurrency hits 1500. It
> can be the database. It can be the memory in the DB server, it can be a
> memory in the PHP application server, it can be inefficient code written by
> a junior developer.
>
> Let me know if that helps.
> ―
> Shiplu.Mokadd.im
> Senior Unix Hacker
>
> *Innovation distinguishes between follower and leader*
> —  Steve Jobs
>
>
> *An expert is a person who has made all the mistakes that can be made in a
> very narrow field.*—  Niels Bohr
>
>
> On Tue, Oct 26, 2021 at 12:22 PM Deepak Goel <[email protected]> wrote:
>
>> Hello
>>
>> When we throw 5 to 10 concurrent requests/sec, the response times are in
>> ms for php as well as java. As we increase the load to 1500
>> concurrent users for php, the response times climb to 2-3 seconds. We would
>> like to serve 6000+ concurrent users with millisecond response times for
>> php (same as java).
>>
>> Thanks.
>>
>>
>> Deepak
>> "The greatness of a nation can be judged by the way its animals are
>> treated - Mahatma Gandhi"
>>
>> +91 73500 12833
>> [email protected]
>>
>> Facebook: https://www.facebook.com/deicool
>> LinkedIn: www.linkedin.com/in/deicool
>>
>> "Plant a Tree, Go Green"
>>
>> Make In India : http://www.makeinindia.com/home
>>
>>
>> On Tue, Oct 26, 2021 at 3:48 PM shiplu <[email protected]> wrote:
>>
>>> > Average Response Time for PHP pages is 2-3 seconds for 1500 concurrent
>>> users (150 requests/sec).
>>>
>>> This is where the problem lies. The requests are not being finished and
>>> then new requests are coming in. The requests are not being finished
>>> because the first request is slow.
>>>
>>> What is the average response time when you send 5 concurrent requests?
>>> Is it sub 200ms?
>>>
>>> How many users do you want to serve in how many seconds?
>>>
>>> ―
>>> Shiplu.Mokadd.im
>>> Senior Unix Hacker
>>>
>>> *Innovation distinguishes between follower and leader*
>>> —  Steve Jobs
>>>
>>>
>>> *An expert is a person who has made all the mistakes that can be made in
>>> a very narrow field.*—  Niels Bohr
>>>
>>>
>>> On Tue, Oct 26, 2021 at 12:10 PM Deepak Goel <[email protected]> wrote:
>>>
>>>> Hello
>>>>
>>>> 1. Concurrent php requests/sec is 150.
>>>>
>>>> 2. Average Response Time for PHP pages is 2-3 seconds for 1500
>>>> concurrent users (150 requests/sec). If we increase the load beyond 1500
>>>> concurrent users, the response time goes up to 10 seconds for 2000+
>>>> concurrent  users. We also start getting 500 errors as we increase the load
>>>> beyond 2000 concurrent users.
>>>>
>>>> 3. For Java, the response times are in milliseconds even for 6000+
>>>> concurrent users (for the same functionality)
>>>>
>>>> Cheers!
>>>>
>>>> Deepak
>>>> "The greatness of a nation can be judged by the way its animals are
>>>> treated - Mahatma Gandhi"
>>>>
>>>> +91 73500 12833
>>>> [email protected]
>>>>
>>>> Facebook: https://www.facebook.com/deicool
>>>> LinkedIn: www.linkedin.com/in/deicool
>>>>
>>>> "Plant a Tree, Go Green"
>>>>
>>>> Make In India : http://www.makeinindia.com/home
>>>>
>>>>
>>>> On Tue, Oct 26, 2021 at 3:28 PM shiplu <[email protected]> wrote:
>>>>
>>>>> Hello Deepak,
>>>>>
>>>>> 1500 as concurrency is very very high. It means there are 1500
>>>>> *ongoing* HTTP requests. They started and didn't finish yet at the
>>>>> same time. This can only happen when your average response time is high.
>>>>> What is your average response time?
>>>>>
>>>>> ―
>>>>> Shiplu.Mokadd.im
>>>>> Senior Unix Hacker
>>>>>
>>>>> Meeting Calendar: https://calendar.x.ai/shiplu
>>>>>
>>>>> *Innovation distinguishes between follower and leader*
>>>>> —  Steve Jobs
>>>>>
>>>>>
>>>>> *An expert is a person who has made all the mistakes that can be made
>>>>> in a very narrow field.*—  Niels Bohr
>>>>>
>>>>>
>>>>> On Tue, Oct 26, 2021 at 11:55 AM Deepak Goel <[email protected]>
>>>>> wrote:
>>>>>
>>>>>> Hello
>>>>>>
>>>>>> I have a very simple PHP application which I develop on
>>>>>> Apache-Web-Server. However the application does not scale beyond 1500
>>>>>> concurrent users (150 requests/sec), even though there is plenty of
>>>>>> hardware available (CPU is getting utilized 15%, Memory is getting utilized
>>>>>> 15%).
>>>>>>
>>>>>> On a similar note, when I develop the same application in Java, it
>>>>>> scales to more than 6000 concurrent users (600 requests/sec) on the same
>>>>>> hardware.
>>>>>>
>>>>>> Any pointers on how to address this scalability issue?
>>>>>>
>>>>>> Thanks!
>>>>>>
>>>>>> Deepak
>>>>>> "The greatness of a nation can be judged by the way its animals are
>>>>>> treated - Mahatma Gandhi"
>>>>>>
>>>>>> +91 73500 12833
>>>>>> [email protected]
>>>>>>
>>>>>> Facebook: https://www.facebook.com/deicool
>>>>>> LinkedIn: www.linkedin.com/in/deicool
>>>>>>
>>>>>> "Plant a Tree, Go Green"
>>>>>>
>>>>>> Make In India : http://www.makeinindia.com/home
>>>>>>
>>>>>