Open keithazzopardi opened 9 months ago
There is currently no mechanism for this, but if you have a proposal for an implementation which will not be prohibitively expensive at runtime, then we will consider it. Ideally, it should impact performance only if it is used.
I am also interested in learning more about the scenario. I see that this could be quite beneficial. Ideally, if we had cheap, possibly approximate counters per-grain and per-grain-type, we could use those to prioritize request or Task scheduling and/or to throttle grains which consume too much.
For inspiration, the JIT in .NET 8 implements scalable approximate counting.
Hi folks,
Let's say that we have 2 types of grains: Grain 1 and Grain 2. I would like that under heavy system load Grain 1 will have higher priority compared to Grain 2 and serves more throughput. What is the best way to achieve this? Perhaps throttling on grain type?
Thanks,