Caching of byte arrays or streams

[esskar]

I want to make the transformation to fusion cache; before I can decide, i want to know how it handles byte arrays?
We are currently only using redis as a cache, and we are caching the data like

var cache = await _cacheConnectionProvider.GetCacheDatabaseAsync(cancellationToken); // cache is a IDatabase from Redis
await cache.StringSetAsync(CreateDataKey(environment, path), RedisValue.CreateFrom(ms), _cacheDuration);

Redis will store the data as is, fusion cache will probably use the JSON serializer which i want to avoid. Is there a way to do so?

[jodydonetti]

Hi @esskar

I want to make the transformation to fusion cache; before I can decide, i want to know how it handles byte arrays?

A byte[] is handled as any other type, like string or Person.

We are currently only using redis as a cache, and we are caching the data like
[...]
Redis will store the data as is, fusion cache will probably use the JSON serializer which i want to avoid. Is there a way to do so?

FusionCache is not based on JSON, instead it can support any serializer out there (or you can create your own if you like) thanks to a generic IFusionCacheSerializer interface that models "a generic binary serializer".

For the most common ones I already created adapter packages, which you can find here: as you can see there are the usual suspects that are JSON-based, but there are also others which are natively optimized for binary representations like MessagePack, MemoryPack or Protobuf.

Usually JSON is the preferred approach for most data types, since it's good enough perf-wise and handles pretty well things like an evolvable schema, which can happen when you refactor your Person or Product class, add new properties and so on.

In your case, if you need to handle byte[]s, I'd suggest to simply use Protobuf, MessagePack or MemoryPack.

I just did a test locally, and with a 100K bytes blob, the serialized entry in Redis with Protobuf (just as an example) was only 24 bytes larger.
Of course if you use extra features (like fail-safe, eager refresh, etc) you may need a little bit more bytes, but that's it.

I'd suggest to try with Protobuf or one of the other 2 and see if the size/perf is good enough for you, but I think this is the way.

Oh, and since FusionCache supports multiple named caches, you'll not be forced to use Protobuf (or whatever) for everything: you can create 2 named caches, one for your common objects configured with a JSON-based serializer, and one for your binary data configured with Protobuf.

Finally, since you are starting right now, I also suggest to start directly with FusionCache V2 (I recently released the preview-4) so that when it will be officially out in a week or two you'll be already on the latest version.

Hope this helps.