/avatar.jpg

CS186-L18: Parallel Query Processing

Intro to Parallelism Architectures and Software Structures we will focus on the shared-nothing here 😋 Kinds of Query Parallelism side note: intra: single inter: multiple at the same level Parallel Data Acceess Data Partitioning across Machines Round robin means that each machine haves the same shuffled data parallel scans scan and merge $\sigma_p$ : an operator that skip entire sites that have no matching tuples in range or hash partitioning

CS186-L13: Transactions & Concurrency I

🎉 Intro transaction’s principle ACID Isolation (Concurrency) however, do not consider serial execution 😅 Atomicity and Durability Consistency Concurrency Control 基本符号表达 序列等价性: $Def1:$ Serial Schedule each transaction executes in a serial order, one after the other, without any intervening $Def2:$ schedules Equivalent involve same transaction each transaction’s actions are the same order both transactions have the same effect on the database’s final state $Def3:$ Serializable if a schedule is serializable, then it is equivalent to some serial schedule Conflict Serializability conflict operations?

CS186-L14: Transactions & Concurrency II

Two Phase Locking (2PL) Strict 2PL same as 2PL, but with stricter definition of release locks at once pink area is the Strict 2PL Lock Management there is a lock manager, which maintains a hash table keyed on names of objects being locked Deadlocks 🤔 why happen? side note: prioritize upgrades can avoid #2 unlike the OS which can have a fixed order of required sources…… avoiding deadlocks timeout first, TIMEOUT is a not so bad idea 🤔