ratcashdev edited a comment on issue #1334: Experimental settings to achieve 5 ms latency URL: https://github.com/apache/incubator-pulsar/issues/1334#issuecomment-408586469 Thank you for the detailed explanation, @sijie. So if I understand all of that correctly, 1. when we compare the two in their default configuration (default config being the `apple`), Kafka will probably have lower latency than Pulsar for publishing, since - at least up until know - it could not control fsync behavior. Users need to keep in mind the difference in behaviour at their default settings: `filesystem-persistence` (Kafka) vs `direct-to-disk-persistence` and forced fsync (Pulsar) 2. when we compare the two in a config that's comparable in behavior (here identical behavior being the `apple`), Kafka (with something resembling `direct-to-disk-persistence` will likely provide higher latency compared to Pulsar in its default setting in this regard. 3. Pulsar using `filesystem-persistence` was not possible up until today (can you please pinpoint the exact config that needs to be tweaked) , and it's likely that it will provide lower latencies, but not lower than with a `non-persistent` topic, will it? And yet back to option No 2 above: Kafka , by default, (first?) replicates the data, then writes to the filesystem. What's the order for Pulsar, when we disable forced fsync? This makes me wonder. If up until know Pulsar did not support delayed/disabled fsync and is the default for kafka, how exactly could they
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
