Chapter 5 - Validating Performance Optimizations ¶

We, as humans, are very bad at understanding how a computer works and how code is executed. It is impossible to know which implementation of an algorithm is going to be the fastest without a deep understanding of operating systems and your programming language implementation. Moreover, a patch improving one part of the code could potentially have a negative impact on another part of the application, making the overall speed actually worse.

Instead of guessing, we need hard numbers. Take a profile using the browser extension again on https://fix1.book.b7e.io/sighting/135, where we applied our 2 patches from the previous chapter:

• Implement a new countForUser method in the CommentRepository class:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

diff --git a/src/Repository/CommentRepository.php b/src/Repository/CommentRepository.php
--- a/src/Repository/CommentRepository.php
+++ b/src/Repository/CommentRepository.php
@@ -5,6 +5,7 @@ namespace App\Repository;

    use App\Entity\Comment;
    use Doctrine\Bundle\DoctrineBundle\Repository\ServiceEntityRepository;
    use Doctrine\Common\Persistence\ManagerRegistry;
+   use App\Entity\User;

+   public function countForUser(User $user, \DateTimeImmutable $sinceDate): int
+   {
+       return (int) $this->createQueryBuilder('comment')
+          ->select('COUNT(comment.id)')
+          ->andWhere('comment.owner = :user')
+          ->andWhere('comment.createdAt >= :sinceDate')
+          ->setParameter('user', $user)
+          ->setParameter('sinceDate', $sinceDate)
+          ->getQuery()
+          ->getSingleScalarResult();
+    }

• Refactor the countRecentCommentsForUser method in the CommentHelper class:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34

diff --git a/src/Service/CommentHelper.php b/src/Service/CommentHelper.php
index c806c1a..97a7aef 100644
--- a/src/Service/CommentHelper.php
+++ b/src/Service/CommentHelper.php
@@ -79,6 +79,12 @@ namespace App\Service;

    use App\Entity\User;
    +use App\Repository\CommentRepository;

     class CommentHelper
     {
+        private $commentRepository;
+
+        public function __construct(CommentRepository $commentRepository)
+        {
+            $this->commentRepository = $commentRepository;
+        }

        public function countRecentCommentsForUser(User $user): int
        {
-            $comments = $user->getComments();
-            $commentCount = 0;
-            $recentDate = new \DateTimeImmutable('-3 months');
-            foreach ($comments as $comment) {
-                if ($comment->getCreatedAt() > $recentDate) {
-                    $commentCount++;
-                }
-            }

-            return $commentCount;
+            return $this->commentRepository
+                ->countForUser($user, new \DateTimeImmutable('-3 months'));
        }
    }

Without the patch, the profile summary looked like this:

And here is the new one with the patch:

The wall time went down from 781ms to ... 1.24s, which may comes at a surprise. Was is really a bad idea? If profiling other sighting pages show nice improvements, we are a clear degradation of performance for /sighting/135.

Let's try to understand, then fix, this side-effect.

The call graph tells us that even if App\Repository\CommentRepository::countForUser() is still called 629 times, it is retrieving directly scalar values from the DB without the need of hydrating full entities:

We have just done a "manual comparison": we compared numbers and we looked for expected changes in the call graph. But there is a better way with Blackfire: profiles comparison.

Click on the Blackfire logo to view your Blackfire "Dashboard":

Click first on the "Compare" button on the second listed profile (the countForUser profile without the patch) and then click on "Compare" again on the first listed profile (the countForUser profile after applying the patch). Blackfire now displays the comparison of the two profiles. The web interface looks very similar to the profile one but the call graph is now a visual representation of the comparison: faster nodes in blue, slower nodes in red:

The summary indicates that there is a 58% time deterioration between the two profiles. Even with clear improvements for the Peak Memory (-43%) and Network (-73) usage, the graph is mostly made of slower red nodes.

Comparing SQL queries might give us more information:

The cardinality is interesting. 629 calls to the countForUser method caused only 100 SELECT queries when we now have 629 SELECT count. The third query on the list teach us that there are 100 users in the database.

This means we were protected by Doctrine internal caching mechanism Data has been retrieved and the entities hydrated only once, then stored for a possible later use.

Let's then add a simple caching strategy:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44

diff --git a/src/Twig/AppExtension.php b/src/Twig/AppExtension.php
index 2cd5e63..ab9c696 100644
--- a/src/Twig/AppExtension.php
+++ b/src/Twig/AppExtension.php
@@ -4,6 +4,8 @@ namespace App\Twig;

use App\Entity\User;
use App\Service\CommentHelper;
+use Symfony\Contracts\Cache\CacheInterface;
+use Symfony\Component\Cache\CacheItem;
use Twig\Extension\AbstractExtension;
use Twig\TwigFilter;
use Twig\TwigFunction;
@@ -11,10 +13,12 @@ use Twig\TwigFunction;
class AppExtension extends AbstractExtension
{
    private $commentHelper;
+    private $cache;

-    public function __construct(CommentHelper $commentHelper)
+    public function __construct(CommentHelper $commentHelper, CacheInterface $cache)
    {
        $this->commentHelper = $commentHelper;
+        $this->cache = $cache;
    }

    public function getFilters(): array
@@ -25,6 +29,17 @@ class AppExtension extends AbstractExtension
    }

    public function getUserActivityText(User $user): string
+    {
+        $key = sprintf('user_activity_text_%s', $user->getId());
+
+        return $this->cache->get($key, function (CacheItem $item) use ($user) {
+            $item->expiresAfter(3600);
+
+            return $this->calculateUserActivityText($user);
+        });
+    }
+
+    private function calculateUserActivityText(User $user): string
    {
        $commentCount = $this->commentHelper->countRecentCommentsForUser($user);

Let's trigger a profile using the browser extension on https://fix1withcache.book.b7e.io/sighting/135, which contain the simple caching mechanism on top of the first fix:

The figures seem more in line with our expectations. Let's compare with the initial profile to have precise information on the outcome:

The summary indicates that there is a 57% time improvement between the two profiles, with our two rather simple patches, which is not too bad.

Having a look at the numbers on the left and the colors on the comparison graph, we can conclude that there is now no performance regression elsewhere:

In the previous chapter, we used the homepage and the project page profile summaries to select the page we should optimize first. Then we looked at a profile of the project page to find optimizations.

Instead of doing this manually, we could have used the Blackfire's comparison feature to highlight performance differences:

Not surprisingly, the comparison shows that the biggest slow down comes from the Doctrine\ORM\UnitOfWork::createEntity calls.

Blackfire provides detailed data on many dimensions: wall time, CPU time, I/O time, memory, and network. Focusing on the wall time only, like we did on Finding Bigfoot, is just a first step.

A function consuming a lot of memory or getting a lot of data from the network has a direct impact on performance as well. The wall time already includes the impact of memory consumption or the impact of the amount of data retrieved from the network... but depending on the machine specifications or the network topology, the impact may vary widely.

Data retrieved over a network is an excellent example. On your local development machine, all project components are probably installed locally, so loading big payloads probably won't have a significant impact on performance as latency is very low in this situation. On production servers, where data can be hosted on a different server or even on a different network, latency and network speed can have a much bigger impact on the overall performance of your application. That's one of the reasons why you should optimize all dimensions, not just the wall time.

Always check all dimensions when looking for performance issues.

57% improvement on the Fiding Bigfoot sighting page is impressive, but can we do better? Profiling is a never-ending process. Whenever you fix a bug or add a new feature, you need to check the performance impact of that change.

Take 10 minutes now to look again at the homepage profile again and try to find some more function calls that could be optimized.

Done? Did you find something? I did!

25 calls to PDOStatement::execute() represents 6.68% of the wall time, almost 18% of the I/O Wait time and over 92% of the Network .

Optimizing a function call can be done in two ways: optimizing some of the functions called by this method or reducing the number of calls to this method.

Browsing the call graph, we can see PDOStatement::execute() is related to a Twig file _sightings.html.twig. Use the search field on the left to find the _sightings.html.twig::doDisplay call then click on the function name to reveal the detailed panel:

Besides the main dimension numbers, we learn that this method has "1 caller" and calls "3 callees".

A caller is a function that called our method, a parent node in the call graph. Again, click on each caller:

What about the callees? A callee is a function called by our method, a child node in the call graph. Click on each callee arrow to review them all:

The main callee is twig_length_filter, and it comes from Twig, which we probably cannot optimize. From Twig documentation, we learn that the ...|length filter returns the number of items of a sequence or mapping, or the length of a string.

Let's step back to reflect on the situation. Within a Twig template, counting a variable with twig_length_filter ultimately cause the execution of 25 SQL queries and the hydratation of 25 more entities.

Let's check the code of the _sightings.html.twig Twig file to fully understand this specific issue now we know exactly where to look for that bottleneck.

1
2
3
4
5
6
7
8

{% for sighting in sightings %}
<tr>
    ...
    <td>
        <a class="text-white table-content text-center" href="{{ path('app_sighting_show', {id: sighting.id}) }}">{{ sighting.comments|length }}</a>
    </td>
</tr>
{% endfor %}

We are indeed iterating over the list of BigFootSighting entities and are trying to display the number of comments one sighting has.

It's a common issue called the N+1 problem, that occurs when managing 1-n relationships with an ORM. The related Comment entities are not fetched with the BigFootSighting one and are loaded only the first time they are accessed. This is the default (aka LAZY) behavior for associations in Doctrine.

In our case, we are fetching, and hydrating, them all only to count them. This sounds familiar, isn't it?

This case could be solved in different ways. One way would be to change the type of association the BigFootSighting entity has with the Comment one. Doctrine has an EXTRA_LAZY association type which allows counting the collection without having to fully load its members.

Let's refactor the BigFootSignthing Entity accordingly:

1
2
3
4
5
6
7
8
9
10
11
12
13

diff --git a/src/Entity/BigFootSighting.php
index 42836f9..cd0aeba 100644
--- a/src/Entity/BigFootSighting.php
+++ b/src/Entity/BigFootSighting.php
@@ -55,7 +55,7 @@ class BigFootSighting
         private $createdAt;

        /**
-        * @ORM\OneToMany(targetEntity="App\Entity\Comment", mappedBy="bigFootSighting")
+        * @ORM\OneToMany(targetEntity="App\Entity\Comment", mappedBy="bigFootSighting", fetch="EXTRA_LAZY")
         * @ORM\OrderBy({"createdAt"="DESC"})
         */
         private $comments;

By now, you know the story by heart. Go to https://fix2.book.b7e.io/, generate a profile with the browser extension, and compare it with the previous profile we made after the first fix:

These two patches improve the performance of the Finding Bigfoot homepage by 44% (from 122ms to 68.3ms) and the sighting pages by 57% (from 781ms to 339ms).

To sum up, finding and fixing performance bottlenecks always uses the same profiling methodology:

• Profile key pages;
• Select the slowest ones;
• Compare and analyze profiles to spot differences and bottlenecks (on all dimensions);
• Find the biggest bottlenecks;
• Try to fix the issue or improve the overall performance;
• Check that tests are not broken;
• Generate a profile of the updated version of the code;
• Compare the new profile with the first one;
• Take a moment to celebrate your achievement;
• Rinse and repeat.

You might think that optimizing an application like we've just done is enough and that Blackfire is not needed anymore. And you would be wrong. This is just the beginning of our journey. Performance management is a day-to-day activity. Whenever you fix a bug or add a new feature in your application, you should check that you have not introduced a performance regression.

The next chapter will guide you through the installation of Blackfire on your environment to let you profile your very own projects.