Optimize document and media seeding by looking up from database in batches

[AndyButland]

Prerequisites

• I have added steps to test this contribution in the description below

Relates to discussion in #18892 around boot time.

Description

When Umbraco boots it will seed the hybrid cache based on configured settings. It currently does this via the hybrid cache's GetOrCreate method, which means that on start up, when the memory cache is empty, each seeded item makes a database request to get a single value back. This can add up to a lot of SQL calls if a large number of items are seeded.

This PR optimizes this to handle the seeding in batches of 100. For each batch it'll check for each if they are available in the cache. For all not found it'll make one database request to get all the documents or media to cache, and then cache them one at a time. In doing so this we change to fewer, larger database calls from many small ones.

Support for "Exists" in Hybrid Cache

Note that the hybrid cache API doesn't support an "Exists" check currently, but based on the discussion here it looks to be possible to create one. I've used this code to add an ExistsAsync extension method. I've done this a little differently though. The shared code worked by setting flags on the HybridCacheEntryOptions to disable writing of the cache, which had this downside:

... since we are disabling writing on L1, if the value is present in L2 we are not taking it in L1 for the next access, meaning we'll go to L2 every single time...

Which isn't what we want for seeding. By seeding we are saying we do want these values in memory, so in the L1 cache. As such I've modified the code such that, if we don't find a value, we'll write a "null" but then immediately remove it again. That seems right for our needs where we aren't expecting to cache null values for documents and media.

Performance

To test this adds benefit I've created an Umbraco setup with increasing number of content nodes selected for seeding, and look to get positive results:

Number of Documents Seeded	Time Taken (Before)	Time Taken (After)
500	928 ms	260 ms
1000	1629 ms	328 ms
2000	3218 ms	514 ms

Testing

Once Umbraco has started up with seeding, verify pages load without SQL requests using MiniProfiler.

[Copilot]

Pull Request Overview

This PR optimizes document and media seeding by replacing single database requests with batch processing to improve Umbraco's boot time performance. The optimization reduces SQL calls from many individual requests to fewer, larger batches of 100 items.

• Implements batch processing for cache seeding with 100-item groups
• Adds HybridCache extension methods for existence checking and value retrieval
• Introduces bulk database methods for fetching multiple content/media sources

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
HybridCacheExtensions.cs	New extension methods providing ExistsAsync and TryGetValueAsync functionality for batch operations
HybridCacheExtensionsTests.cs	Comprehensive test coverage for the new extension methods
DocumentCacheService.cs	Refactored seeding logic to use batch processing and existence checking
MediaCacheService.cs	Applied similar batch optimization as document cache service
IDatabaseCacheRepository.cs	Added interface methods for bulk content/media retrieval
DatabaseCacheRepository.cs	Implemented bulk database operations for efficient batch queries
DocumentHybridCacheMockTests.cs	Updated tests to reflect new batch processing behavior
DocumentBreadthFirstKeyProviderTests.cs	Minor formatting change adding blank line

[Zeegaan]

Looks good 🚀

[enkelmedia]

@AndyButland Seems like a great improvement! Did you pick 100 just as a "random number" or why was that number picked? I'm wondering if it could be beneficial to be able to change this number or if it could be calculated based on the seed size in any way.

We have some sites where we will have to seed over 20k nodes and (guessing) that it might be faster if we loaded them in batches of 1000 or something. Did you play with different batch sizes?

[Zeegaan]

Good point @enkelmedia , We already had a similar concept with the NuCacheSettings.SqlPageSize, so we could apply the same concept here 😁

[AndyButland]

Yes, good idea to have some flexibility here. I've proposed moving this to configuration in #19894