search

vmware-archive / haret

A strongly consistent distributed coordination system, built using proven protocols & implemented in Rust.
461 stars 18 forks source link

Add RFC #2 for View Change Optimizations #106

Closed andrewjstone closed 7 years ago

andrewjstone commented 7 years ago

Rendered

Thanks to @evanmcc for the idea.

evanmcc commented 7 years ago

+1, I think that this improves upon my handwavy suggestion quite a lot.

justinsheehy commented 7 years ago

+1, I dig this.

tsantero commented 7 years ago

+1, the 8 byte overhead per node per prepare message is negligible to the cost of a full state transfer from each replica during failure recovery--especially considering that on contended networks the paper's specified recovery protocol could cause further disruption given a log with a significant number of entries.

andrewjstone commented 7 years ago

Thanks @evanmcc @justinsheehy @tsantero. I appreciate you taking the time to read this. Approval from legendary ex-bashos just earned this a merge.