search

lkuper / CSE232-2020-10

A graduate course on distributed systems
https://decomposition.al/CSE232-2020-10/
11 stars 1 forks source link

11/16 Discussion for: Robbert van Renesse and Fred B. Schneider, "Chain Replication for Supporting High Throughput and Availability" (OSDI 2004) #16

Open katelynstone13 opened 3 years ago

katelynstone13 commented 3 years ago

Scribes: @Hmadkhali @meghnaburli @git-ram @versey-sherry @lakshmikrish-97

versey-sherry commented 3 years ago

@lkuper Our first draft is ready for you to review. Thank you!

lkuper commented 3 years ago

Some quick feedback:

Q1: looks OK, but there's some copypasta where "there are efforts that go into predicting workloads" appears twice. Also: "For systems that mainly write requests, we may opt for primary/backup." Why would we do that? (hint: write latency)

Q2: "Another case could be that the process/server may be dead but the heartbeat message component is still sending messages." -- that is probably what I said in class, but I can try to phrase it a little more carefully. What I mean is not that the process is dead, but that the component of the process that's actually supposed to apply updates is not working for some reason, say, because the disk is full. Meanwhile, the heartbeat mechanism is unaware of that. We haven't really talked about this kind of fault (because it isn't really a problem specific to distributed systems), but it is something to be aware of when building systems!

Q3: Hmm, so it seems like you really put a lot of emphasis on the homogenous vs. heterogenous here, instead of on LAN vs. WAN. I think you can drop the stuff about homogeneity. The point of the question (which may not have been obvious; my bad!) was really "will chain replication work well across the WAN?" And you pretty much have the right answer, which is "not really", but specifically, why is that the case, and why is it worse than primary-backup in this situation?

lkuper commented 3 years ago

P.S. for the figure, please add a note that it's taken from the paper. Thanks!

versey-sherry commented 3 years ago

P.S. for the figure, please add a note that it's taken from the paper. Thanks!

Resolved! Thanks!

versey-sherry commented 3 years ago

@lkuper the updated version is ready for you to review. Thank you!