What is it about?
When web platforms count things (ad clicks, likes, votes, etc.), they often store copies of the current count on multiple computers around the world, so that it's fast to access the count from anywhere. If you add the ability to "reset" the counter occasionally, say because you want to get a running count at the end of each day, this becomes a bit tricky: one computer might increment the counter at the same time as it gets reset, and we have to make sure that the reset doesn't clobber the new increment. We describe a way to do this that uses much less memory than previous solutions.
Featured Image
Why is it important?
Problems involving coordination among many computers are the backbone of all big web services, but are also tricky to get right, especially when they need to respond to users quickly. We solve a specific coordination problem identified by previous papers and systems in an especially efficient way.
Perspectives
I enjoyed how were able to take an existing solution and find a second solution that accomplishes the same thing, but works more efficiently under the hood, by using several neat tricks. I look forward to seeking out similar optimization opportunities for related problems.
Matthew Weidner
Carnegie Mellon University
Read the Original
This page is a summary of: An oblivious observed-reset embeddable replicated counter, April 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3517209.3524084.
You can read the full text:
Resources
Contributors
The following have contributed to this page
