Marcin Copik
Marcin Copik
We need an accounting system to easily conduct benchmarks. - [ ] Gather lambda runtimes. - [ ] Count queue events. - [ ] Aggregate DynamoDB operation counts. - [...
Sometimes we observe that the main node `/` has large data even if not set. This appeared with the S3 user data store. Likely, there is a bug in the...
When modifying the list of children in a node (`delete`, `create`), we should modify the counter `pFixSys`.
When trying to commit data from the distributor PoV, we need to handle transaction conflicts by retrying in a loop until we get a failure on success. We do not...
The current codebase is not sustainable, and we should improve it. - [ ] Split the logic of different requests into modules. - [ ] Abstract AWS-specific dependencies from the...
AWS S3 and Redis support pipelining, and we could use them to send a stream of requests from a client. - [ ] Implement HTTP-based requests to S3. - [...
Currently, a single writer function can only execute a batch of requests sequentially. However, the operations can be efficiently pipelined by pushing requests to storage in parallel while preserving the...
We noticed that when we create more watches, the epoch counters grow - but they do not shrink once watches are processed.
DynamoDB is not sufficient, we need to try SQS. - [x] Allocate SQS queue in resources. - [ ] Benchmark SQS queue. - [x] Expose configuration of polling parameters. -...
The implementation does not include the new type of queue and function. - [x] Change writer implementation to push to the new queue. - [x] Allocate queues. - [x] Implement...