Distributed Database Management Systems Lecture 28

In the previous lecture Types of Transaction Transaction in DDBS Serial Transactions Conflicting Ops/Tr.

In this Lecture Serializability Theory Serializability Theory in DDBS.

Equivalent Schedules Two schedules S1, S2 defined over same T are equivalent if they have same effect on the database, that is, leave database in same final state.

Formally, if for each pair of conflicting operations Oij and Okl (i ≠ k) if Oij <1 Okl then Oij <2 Okl The phenomena is also called conflict equivalence

Serializable Schedule

If it is conflict equivalent to a serial schedule, i. e If it is conflict equivalent to a serial schedule, i.e., the final state in which it leaves the database is equivalent to a serial schedule-

Ss ={W2(x), W2(y), R2(z), C2, R1(x), W1(x), C1, R3(x), R3(y), R3(z), C3} S1 = {W2(x), R1(x), R3 (x),W1(x),C1, W2(y), R3(y), R2(z),C2 ,R3(z), C3} No

Ss ={W2(x), W2(y), R2(z), C2, R1(x), W1(x), C1, R3(x), R3(y), R3(z), C3} S2 ={W2(x), R1(x), W1(x), C1, R3(x), W2(y), R3(y), R2(z), C2, R3(z), C3} yes

The function of the concurrency controller is to generate serializable schedule Local and Global Schedules.

Fragmented Databases

The serializability is straight forward. Local transaction are independent of each other; each concerns local data.

In case of global transactions Local subtransactions will be treated as different transactions-

Replicated Databases

T1: T2: Read(x) x = x + 5 Write(x) Commit x = x*10 LS1={R1(x), W1(x),C1, R2(x), W2(x), C2} LS2={R2(x), W2(x), C2, R1(x), W1(x), C1}

All values of replicated data should be same Local Schedule same Conflicting Ops in same relative order on all sites.

Logical and physical data items User issues Ops on logical data items Replica control maps to physical ones-

ROWA Protocol Reduces availability in case of failure Different alogos, different replications.

Concurrency Control Algorithms

Different categorizations possible Like, mode of distribution, network topology-

Synchronization primitive is the most common Locking and Ordering Pessimistic & Optimistic.

Pessimistic approach synchronizes transactions early Optimistic do this late in execution life cycle of transactions

Pessimistic Locking-based Centralized Locking Primary Copy Locking Distributed Locking-

Pessimistic Timestamp Ordering (TO) Hybrid Basic TO Multiversion TO Conservative TO Hybrid

Optimistic Locking-based Timestamp ordering-based.

Locking based Concurrency Control

Basic idea is that data items accessed by conflicting operations are accessed by one operation at a time Data Items locked by Lock Manager

Two major types of locks, read lock and write lock Transaction need to apply lock first.

For improved accessibility, compatibility of locks to be established rli(x) wli(x) rlj(x) Yes No wlj(x)

Locking is job of DDBMS, not the user Scheduler is the Lock Manager TM and LM interact.