✳flâneur — a map of the web's best reading

https://pdos.csail.mit.edu/6.824/notes/l-farm.txt

pdos.csail.mit.edu · 1,918 words · saved by 1 readers

6.5840 2023 Lecture 15: FaRM, Optimistic Concurrency Control why are we reading about FaRM? another take on transactions+replication+sharding this is still an open research area! optimistic concurrency control exploiting huge performance potential of RDMA NICs the overall setup all in one data center configuration manager, using ZooKeeper, chooses primaries/backups sharded w/ primary/backup replication P1 B1 P2 B2 ... can recover as long as at least one replica of each shard i.e. f+1 replicas tolerate f failures transaction clients (which they run in the servers) transaction code acts as two-phase-commit Transaction Coordinator (TC) the goal: millions of distributed transactions per second so the time budget is tens of microseconds very challenging! how do they get high performance? sharding over many servers (90 in the evaluation) data must fit in total RAM (so no disk reads) non-volatile RAM (so no disk writes) one-sided RDMA

6.5840 2026 Lecture 14: FaRM, Optimistic Concurrency Control why are we reading about FaRM? another take on transactions+replication+sharding this is still an open research area! optimistic concurrency control exploiting huge performance potential of RDMA NICs the overall setup all in one data center configuration manager, using ZooKeeper, chooses primaries/backups sharded w/ primary/backup replication P1 B1 P2 B2 ... can recover as long as at least one replica of each shard i.e. f+1 replicas tolerate f failures transaction clients (which they run in the servers) transaction code acts as two-p

Explore this link on the map →

related reading