Performance: N+1 Query Problem

Phorge Contributor Documentation (Developer Guides)

How to avoid a common performance pitfall.

Overview

The N+1 query problem is a common performance antipattern. It looks like this:

$cats = load_cats();
foreach ($cats as $cat) {
  $cats_hats = load_hats_for_cat($cat);
  // ...
}

Assuming load_cats() has an implementation that boils down to:

SELECT * FROM cat WHERE ...

..and load_hats_for_cat($cat) has an implementation something like this:

SELECT * FROM hat WHERE catID = ...

..you will issue "N+1" queries when the code executes, where N is the number of cats:

SELECT * FROM cat WHERE ...
SELECT * FROM hat WHERE catID = 1
SELECT * FROM hat WHERE catID = 2
SELECT * FROM hat WHERE catID = 3
SELECT * FROM hat WHERE catID = 4
SELECT * FROM hat WHERE catID = 5
...

The problem with this is that each query has quite a bit of overhead. It is much faster to issue 1 query which returns 100 results than to issue 100 queries which each return 1 result. This is particularly true if your database is on a different machine which is, say, 1-2ms away on the network. In this case, issuing 100 queries serially has a minimum cost of 100-200ms, even if they can be satisfied instantly by MySQL. This is far higher than the entire server-side generation cost for most Phorge pages should be.

Batching Queries

Fix the N+1 query problem by batching queries. Load all your data before iterating through it (this is oversimplified and omits error checking):

$cats = load_cats();
$hats = load_all_hats_for_these_cats($cats);
foreach ($cats as $cat) {
  $cats_hats = $hats[$cat->getID()];
}

That is, issue these queries:

SELECT * FROM cat WHERE ...
SELECT * FROM hat WHERE catID IN (1, 2, 3, 4, 5, ...)

In this case, the total number of queries issued is always 2, no matter how many objects there are. You've removed the "N" part from the page's query plan, and are no longer paying the overhead of issuing hundreds of extra queries. This will perform much better (although, as with all performance changes, you should verify this claim by measuring it).

See also LiskDAO::loadRelatives() method which provides an abstraction to prevent this problem.

Detecting the Problem

Beyond reasoning about it while figuring out how to load the data you need, the easiest way to detect this issue is to check the "Services" tab in DarkConsole (see Using DarkConsole), which lists all the service calls made on a page. If you see a bunch of similar queries, this often indicates an N+1 query issue (or a similar kind of query batching problem). Restructuring code so you can run a single query to fetch all the data at once will always improve the performance of the page.