How does "Request Coalescing" work? #263
-
|
The docs explain that a cache stampede is avoided using "Request Coalescing", but there's no hint how that works. Does it simply lock the data-loading factory? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 7 replies
-
|
Hi @lonix1 , yes that is basically it. Request Coalescing simply means that multiple requests for the same data at the same time (to be more precise: in the same small time window) are handled as logically the same. To do that, usually a lock is involved, and in FusionCache that is the case with one lock per cache key, to not block a request for product 1 when one for product 2 is being executed, and vice versa. Hope this helps. |
Beta Was this translation helpful? Give feedback.
-
Thanks @lonix1 , after all the effort put into FusionCache words like these mena a lot, so thanks really!
Glad it has been helpful. Oh, and there are other features I'm working on so no, I haven't though of everything everything 😬
Yeah I'm collaborating with the team on that, where "collaborating" means I'm sharing my experiences related to FusionCache with the team and give them some suggestions/ideas, and they are doing the actual work 😅 (also, I suggested the name itself). |
Beta Was this translation helpful? Give feedback.
-
|
Will FusionCache be updated to encapsulate HybridCache, or will they be "competitors"? (But I don't think they can compete with this, feature-wise. The framework classes usually do only the "basics". If they solve the locking/stampede issue it would be nice though.) |
Beta Was this translation helpful? Give feedback.
-
Potentially both, in a way. The new HybridCache from Microsoft will in theory be 2 things: a sort of lingua franca that others (like FusionCache) can implement and provide additional features, and a concrete out-of-the-box implementation ready to be used. So on one hand the plan is that I'll create (read: already working on it) an HybridCache implementation that will act as an adapter for FusionCache, so people and other libs may be able to either depend directly on FusionCache to get all the features, or on a common abstraction (the HybridCache abstract class) while using a different implementation (like FusionCache) without missing some extra features. On the other hand though they may "compete" in the sense that there's the risk of people just going with "the Microsoft way" and be done with it: they would miss on some extra features (fail-safe, soft/hard timeouts, auto-recovery, etc) but this is their choice. In the end I'll keep evolving FusionCache to give people new features, extra resiliency and basically a choice, while I hope Microsoft will recognize the existence of other libs like mine, maybe even suggesting them in their blog posts or docs. We'll see 🤞
I don't know about future versions of their HybridCache, but their current plan is to support L1/L1+L2, cache stampede protection, compression and eviction by tag (this last one is still to be decided). FusionCache on the other hand has more resiliency features (fail-safe, soft/hard timeout, eager refresh), more flexible features (conditional refresh, named caches, adaptive caching), self-healing (auto-recovery) and more so their current feature set is admittedly smaller... for now. Btw note that I'm also experimenting with compression, and I'm actively working on eviction by tag, we'll see... Hope I've been able to give you an overall picture. |
Beta Was this translation helpful? Give feedback.
-
|
That was a great answer, thank you. One day you should copy-paste that into the repo's README, because it's very useful! Seems to me it will be like the difference between Having lots of fun using FusionCache so far, many thanks! |
Beta Was this translation helpful? Give feedback.
-
|
Hi @lonix1
Agree, and I kinda did that, by answering here: I put a lot of details and tried to clarify some nuance, if you have some time let me know what you think. I think I'll also link that in the main README to be easily discoverable for eveyone wondering about it.
Rest assured, I will not 😬
Thanks, happy you're liking it! |
Beta Was this translation helpful? Give feedback.
Hi @lonix1 , yes that is basically it.
Request Coalescing simply means that multiple requests for the same data at the same time (to be more precise: in the same small time window) are handled as logically the same. To do that, usually a lock is involved, and in FusionCache that is the case with one lock per cache key, to not block a request for product 1 when one for product 2 is being executed, and vice versa.
Hope this helps.