From physical storage, indexing and query optimization to transactions, concurrency control, reliability and active databases — the full Politecnico di Milano Data Base 2 course, rebuilt as an interactive, exam-focused study guide.https://www.polistudy.me/enhttps://www.polistudy.me/db2/db-architecture-acid/https://www.polistudy.me/db2/db-architecture-acid/What Data Base 2 adds to the SQL you already know — the layered machine inside the DBMS, the transaction as an atomic unit of work, the four ACID properties, and exactly which module is on the hook for each. Plus the shape of the written exam, read off fourteen real papers.DBMS architecturesecondary store managerbuffer manageraccess method managertransaction managerreliability managerconcurrency control systemtransactionOLTPACIDatomicityconsistencyisolationdurabilitycommitrollbackhttps://www.polistudy.me/db2/triggers/https://www.polistudy.me/db2/triggers/Event–Condition–Action rules that let the database react on its own. SQL:1999 syntax, transition variables, the BEFORE/AFTER × row/statement grid, execution order, cascading and termination — then the two things exams actually ask for: a correct event inventory and incrementally-maintained derived data.active databaseECA paradigmtriggertransition variabletransition tablerow-level granularitystatement-level granularityBEFORE triggerAFTER triggercascadingrecursive cascadingtriggering graphtermination analysismaterialized viewincremental view maintenanceSIGNALWHEN clausehttps://www.polistudy.me/db2/physical-storage-indexing/https://www.polistudy.me/db2/physical-storage-indexing/How tables actually sit on disk, and what each layout costs. The block-counting cost model, sequential and hash structures, the primary/secondary/clustering index taxonomy, B+ trees, composite indexes and the leftmost-prefix rule — the substrate every query-cost exercise is built on.pageblockI/O cost modelblock factorentry-sequenced structuresequentially-ordered structurehash structurebucketload factoroverflow chainindexsearch keydense indexsparse indexprimary indexsecondary indexclustering indexB+ treeB-treefan-outcomposite indexleftmost prefixhttps://www.polistudy.me/db2/query-optimization/https://www.polistudy.me/db2/query-optimization/How one SQL statement becomes one execution plan, and how to price the alternatives. Relation profiles and selectivity, the four join methods and their cost formulas, external merge sort, and the discipline of writing a complete cost formula — the single largest points pool after concurrency.query optimizeralgebraic optimizationrelation profileselectivitycardinality estimationnested-loop joinindexed nested loopscan and lookupmerge-scan joinhash joinexternal merge sortcost-based optimizationcompile-and-storebuffer poolLRUEXPLAINjoin eliminationhttps://www.polistudy.me/db2/schedules-serializability/https://www.polistudy.me/db2/schedules-serializability/The five anomalies concurrency can produce, and the two formal notions of correctness that rule them out — view-serializability and conflict- serializability. The conflict graph, the class hierarchy, and the exam instruction that makes it all pay off: use class inclusion wherever possible.anomalylost updatedirty readnonrepeatable readphantom updatephantom insertscheduleserial scheduleserializable schedulecommit projectionview equivalencereads-fromfinal writeVSRconflictconflict equivalenceCSRconflict graphblind writeclass inclusionhttps://www.polistudy.me/db2/locking-2pl/https://www.polistudy.me/db2/locking-2pl/The pessimistic answer to concurrency: locks, the two-phase rule that makes them sufficient, strictness that makes aborts safe, the SQL isolation levels built from them, and what to do when transactions wait for each other forever — including Obermarck's distributed detection and hierarchical intention locks.shared lockexclusive locklock tableconflict tablelock escalationtwo-phase lockingstrict 2PLlong-duration lockpredicate lockisolation levelread uncommittedread committedrepeatable readserializabledeadlockwait-for graphwait-diewound-waitObermarck's algorithmupdate lockhierarchical lockingintention lockhttps://www.polistudy.me/db2/timestamp-mvcc/https://www.polistudy.me/db2/timestamp-mvcc/The optimistic answer to concurrency: order transactions by birth date and kill whoever arrives out of order. RTM/WTM bookkeeping, Thomas rule, multiversion timestamps in the exact variant the exercises use, snapshot isolation and write skew — plus where each class sits against 2PL, CSR and VSR.optimistic concurrency controltimestampRTMWTMTS-monoThomas ruleobsolete writemultiversion concurrency controlversionTS-multisnapshot isolationwrite skewfirst-committer-winsLamport timestamphttps://www.polistudy.me/db2/reliability/https://www.polistudy.me/db2/reliability/How atomicity and durability survive a crash. Stable memory, the buffer policies that decide what recovery must do, the log and its two write rules, checkpoints and dumps, and the warm-restart procedure that rebuilds a consistent database from the log.reliability managerstable memorydumpbufferfixunfixforceflushSTEALNO-STEALFORCENO-FORCElogbefore-stateafter-stateUNDOREDOidempotencewrite-ahead logcommit rulecheckpointwarm restartcold restarthttps://www.polistudy.me/db2/ranking-topk/https://www.polistudy.me/db2/ranking-topk/Finding the best k objects when "best" is a scoring function over several criteria. Rank aggregation and instance optimality, the geometry of weighted scores, and the four middleware algorithms — B0, FA, TA and NRA — with the stopping rule and access count each one is graded on.multi-objective optimizationrank aggregationBordaCondorcetKendall tauSpearman footruleMedRankinstance optimalitytop-k queryscoring functionmonotone functionscore spacesorted accessrandom accessB0Fagin's AlgorithmThreshold Algorithmthreshold pointNRAmiddleware costhttps://www.polistudy.me/db2/skylines/https://www.polistudy.me/db2/skylines/Finding the good objects without inventing weights. Dominance and the skyline, BNL and Sort-Filter-Skyline with the sorting precondition that makes SFS correct, the k-skyband, and the sharp distinction recent exams keep testing — skyline points a weighted sum can never make win.dominanceskylinePareto optimalitymaximal vectorspotential optimalityBNLSort-Filter-Skylinetopological orderk-skybandconvex hullanti-correlated datasethttps://www.polistudy.me/db2/jpa-orm/https://www.polistudy.me/db2/jpa-orm/Mapping objects to tables without writing SQL. The JDBC boilerplate JPA removes, entities and the persistence context, and the four decisions every exam asks you to justify from data cardinalities — owner, mapped-by, fetch policy and cascade — plus the entity lifecycle and container transactions.three-tier architectureJDBCimpedance mismatchORMentitypersistence unitpersistence contextmanaged entityentity managerJPQLjoin columnjoin tablemappedByfetch policyEAGERLAZYcascadeorphan removalentity lifecyclecontainer-managed transaction