Timeline for What is the difference between Leaky bucket and Token bucket approach during rate limiting?
Current License: CC BY-SA 4.0
16 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 11 at 11:15 | comment | added | Ewan | Ahh no i think i see where you are going wrong. The leaky buckets capacity is not used for burst | |
| Jun 11 at 7:55 | comment | added | Sankozi | Ok so this leaky bucket is a nice abstraction when dealing networking kind of problems. Still you are comparing leaky bucket with low/no capacity to a token bucket with above 1 capacity. | |
| Jun 6 at 16:08 | comment | added | Basilevs | Let us continue this discussion in chat. | |
| Jun 6 at 14:43 | comment | added | Ewan | I've edited to make clear there are multiple buckets | |
| Jun 6 at 14:43 | history | edited | Ewan | CC BY-SA 4.0 | added 417 characters in body |
| Jun 6 at 14:33 | comment | added | Ewan | really the question should be on the networking stack exchange I am just remembering the old days of everyone wanting to use the same ADSL | |
| Jun 6 at 14:31 | comment | added | Ewan | This kind of thing lets you segment out the bandwidth. You can also put in other rules to allow free bandwidth to be apportioned, but you don't want to be constantly changing the allowed rates | |
| Jun 6 at 14:27 | comment | added | Ewan | mmmmmm. I'll have to think of how to explain, it's not like there is a massive difference but say i put all the traffic through a leaky bucket with 90% of the bandwidth max outflow.. Now I know I have 10% free guaranteed. I instead I had a token bucket I might still hit the cap in bursts | |
| Jun 6 at 13:32 | comment | added | Basilevs | Still no difference with leaky buckets, because an anolog for them would be total leak adjusted to bandwidth. | |
| Jun 6 at 13:20 | comment | added | Basilevs | Ah, I see now. If token inflow matches the bandwidth AND tokens round-robin between buckets AND overflow is added to other buckets, then this answer makes sense. But could we add these clarifications to the answer body? It was hard to figure out. | |
| Jun 6 at 13:10 | comment | added | Basilevs | Also, the example is strange. Why did Bob stop complaining? He still effectively has 50% of bandwidth, even when Andy is offline. The load balancing imlies some kind of shared resource, and I do not see any shared buckets in the example. | |
| Jun 6 at 13:06 | comment | added | Basilevs | If I replace every token bucket in this answer with leaky bucket, nothing changes. Does this mean that there is no difference? | |
| Jun 6 at 12:33 | comment | added | Arseni Mourzenko | That's an outstanding, very clear explanation! Nice example with two users, with two buckets, one per user—when trying to figure it out myself, I somehow convinced myself that it's one bucket for everyone, which doesn't make too much sense, in retrospect. | |
| Jun 6 at 10:19 | history | edited | Ewan | CC BY-SA 4.0 | added 712 characters in body |
| Jun 6 at 10:02 | history | edited | Ewan | CC BY-SA 4.0 | added 533 characters in body |
| Jun 6 at 9:50 | history | answered | Ewan | CC BY-SA 4.0 |