FastQueue perf compared to SDK

[yevhen]

How does FastQueue achieves "up to 20x" faster than SDK provided client? What is the secret sauce?

[Scooletz]

I have two implementations, the SDK and the fast one: https://github.com/Scooletz/QueueBatch/blob/develop/src/QueueBatch/Impl/Queues/FastCloudQueue.cs

My fast approach uses:

• my Span based xml parser: https://github.com/Scooletz/QueueBatch/blob/develop/src/QueueBatch/Impl/Queues/FastXmlAzureParser.cs
• memory pooling for obtaining the response
• a custom HttpContent for sending messages https://github.com/Scooletz/QueueBatch/blob/develop/src/QueueBatch/Impl/Queues/FastCloudQueue.cs#L145

So basically, no allocations, no encoding and no XmlReader. Benchmarking might be affected by distance to DC (so for long hops it won't be 20x), still, this non-alloc approach pays off a lot.

If this answers it, please close the issue @yevhen

[yevhen]

Yes. Thank you!