Strict CO .28SCO.29 Commitment ordering

read-write conflict: sco vs. ss2pl. duration of transaction t2 longer ss2pl sco. ss2pl delays write operation w2[x] of t2 until t1 commits, due lock on x t1 following read operation r1[x]. if t time units needed transaction t2 after starting write operation w2[x] in order reach ready state, t2 commits t time units after t1 commits. however, sco not block w2[x], , t2 can commit after t1 commits. (raz 1991c)

strict commitment ordering (sco; (raz 1991c)) intersection of strictness (a special case of recoverability) , co, , provides upper bound schedule s concurrency when both properties exist. can implemented using blocking mechanisms (locking) similar used popular ss2pl similar overheads.

unlike ss2pl, sco not block on read-write conflict possibly blocks on commit instead. sco , ss2pl have identical blocking behavior other 2 conflict types: write-read, , write-write. result, sco has shorter average blocking periods, , more concurrency (e.g., performance simulations of single database significant variant of locks ordered sharing, identical sco, show this, approximately 100% gain transaction loads; identical transaction loads sco can reach higher transaction rates ss2pl before lock thrashing occurs). more concurrency means given computing resources more transactions completed in time unit (higher transaction rate, throughput), , average duration of transaction shorter (faster completion; see chart). advantage of sco significant during lock contention.

the sco vs. ss2pl performance theorem


sco provides shorter average transaction completion time ss2pl, if read-write conflicts exist. sco , ss2pl identical otherwise (have identical blocking behavior write-read , write-write conflicts).

sco practical ss2pl since ss2pl provides besides serializability strictness, utilized basis efficient recovery of databases failure. ss2pl mechanism can converted sco 1 better performance in straightforward way without changing recovery methods. description of sco implementation can found in (perrizo , tatarinov 1998). see semi-optimistic database scheduler.

ss2pl proper subset of sco (which explanation why sco less constraining , provides more concurrency ss2pl).