3 ms·
The whole article is a straw man. Yes, if things further down your stack are not capable of handling the load in decent time your queue is going to overflow (a
by danielrhodes 12y ago
The whole article is a straw man.
Yes, if things further down your stack are not capable of handling the load in decent time your queue is going to overflow (assuming a fixed capacity). No, it does not make your entire stack faster -- it just defers processing in a way you can manage and tame it. What can become faster is things like front-end requests, which are no longer held up by blocking operations or a starved CPU. Either way, it buys you some time to actually re-engineer your stack to work faster and at greater scale.
- danudey 12y ago1. That's not what 'straw man' means. 2. The author isn't arguing anything opposed to what you're saying; yes, it can speed up front-end requests by deferring work, but if you build an image uploading system that can't handle 10,000 image post-processing tasks per minute, then deferring image post-processing into a queue won't solve your bottleneck. The author isn't talking about individual components of a system, he's talking about the entire system as a whole. Even when you're using it to speed up your front-end, you still need the capacity to do the post-processing on the backend.