another dynamic limit need

[diemus]

I have multiple resources that are being requested, each with varying capacities. Therefore, I need to implement rate limiting based on the capacity of each individual resource.

[myml]

This can be easily implemented. Example of the Gin framework

// 5-S
middleware1 := mgin.NewMiddleware(limiter.New(store, rate))
// 100-S
middleware2 := mgin.NewMiddleware(limiter.New(store, rate))

router.Use(func(ctx *gin.Context) {
    if(ctx.Request.URL == "/api/v1") {
       middleware1(ctx)
       return
    }
    if(ctx.Request.URL == "/api/v2") {
       middleware2(ctx)
       return
    }
})

[diemus]

thanks for the example, but what if some user want higher quotas for the same resource? should i generate all possible middleware ?

[myml]

This depends on the specific business. Below is pseudocode provided as a reference:

var cache sync.Map
router.Use(func(ctx *gin.Context) {
    uid := getUserID(ctx)
    if v,ok := cache.Load(uid); ok {
        h := v.(gin.HandlerFunc)
        h(ctx)
        return
    }
    rate := genRateByUserID(uid)
    h = mgin.NewMiddleware(limiter.New(store, rate))
    cache.Store(uid, h)
    h(ctx)
})

Please pay attention to the store used by the limiter. If you are using Redis for storage, you will need to use different prefixes. However, if you are using in-memory storage, you should not encounter any issues.

[quentinlesceller]

Thank you for this very nice library. I was wondering, performance wise, how scalable is this approach? For let's say 10k users. My current understanding is that my program would need to store all the middleware in memory.

[myml]

Regarding the performance and scalability of this approach, I haven't personally tested it. However, using Redis as a middleware storage can be a viable option for achieving scalability.

[novln]

Thank you @myml for answering these questions.

[quentinlesceller]

Thank you for your answer. Indeed that could be interesting. Gonna try with the local approach one and then see how well it scales.

[diemus]

@myml Thank you for your answer, I guess probably using lower level code is more elegant, I checked the store interface. it's looks more flexibile to do something like dynamic rate limit. I can just pass in different users id as keys and set different rate.

[ibarryyan]

@myml In the above way, I will still find that the request for interface /api/v1 may be calculated on interface /api/v2