Distributed Databases @ KIDS Labs 1
Distributed Database System A distributed database system consists of loosely coupled sites that share no physical component Appears to user as a single system Database systems that run on each site are independent of each other Processing maybe done at a site other than the initiator of request @ KIDS Labs 2
Reasons for Distributed Database Business unit autonomy and distribution Data sharing Data communication costs Data communication reliability and costs Multiple application vendors Database recovery Transaction and analytic processing @ KIDS Labs 3
Figure 13-1 Distributed database environments (adapted from Bell and Grimson, 1992) @ KIDS Labs 4
Distributed Database Options Homogeneous - Same DBMS at each node Autonomous - Independent DBMSs Non-autonomous - Central, coordinating DBMS Easy to manage, difficult to enforce Heterogeneous - Different DBMSs at different nodes Systems With full or partial DBMS functionality Gateways - Simple paths are created to other databases without the benefits of one logical database Difficult to manage, preferred by independent organizations @ KIDS Labs 5
Distributed Database Options (cont.) Systems - Supports some or all functionality of one logical database Full DBMS Functionality - All distributed DB functions Partial-Multi database - Some distributed DB functions Federated - Supports local databases for unique data requests Loose Integration - Local dbs have their own schemas Tight Integration - Local dbs use common schema Unfederated - Requires all access to go through a central, coordinating module @ KIDS Labs 6
Homogenous Distributed Database Systems All sites have identical software They are aware of each other and agree to cooperate in processing user requests It appears to user as a single system @ KIDS Labs 7
An Homogenous Distributed Database Systems example A distributed system connects three databases: hq, mfg, and sales An application can simultaneously access or modify the data in several databases in a single distributed environment. @ KIDS Labs 8
Figure 13-2: Homogeneous Database Identical DBMSs Source: adapted from Bell and Grimson, 1992. @ KIDS Labs 9
Heterogeneous Distributed Database System In a heterogeneous distributed database system, at least one of the databases uses different schemas and software. A database system having different schema may cause a major problem for query processing. A database system having different software may cause a major problem for transaction processing. @ KIDS Labs 10
Figure 13-3: Typical Heterogeneous Environment Non-identical DBMSs Source: adapted from Bell and Grimson, 1992. @ KIDS Labs 11
Major Objectives Location Transparency User does not have to know the location of the data Data requests automatically forwarded to appropriate sites Local Autonomy Local site can operate with its database when network connections fail Each site controls its own data, security, logging, recovery @ KIDS Labs 12
Significant Trade-Offs Synchronous Distributed Database All copies of the same data are always identical Data updates are immediately applied to all copies throughout network Good for data integrity High overhead slow response times Asynchronous Distributed Database Some data inconsistency is tolerated Data update propagation is delayed Lower data integrity Less overhead faster response time @ KIDS Labs 13
Advantages of Distributed Database over Centralized Databases Increased reliability/availability Local control over data Modular growth Lower communication costs Faster response for certain queries @ KIDS Labs 14
Disadvantages of Distributed Database Compared to Centralized Databases Software cost and complexity Processing overhead Data integrity exposure Slower response for certain queries @ KIDS Labs 15
Distributed Data Storage Replication System maintains multiple copies of data, stored in different sites, for faster retrieval and fault tolerance. Fragmentation Relation is partitioned into several fragments stored in distinct sites Replication and fragmentation can be combined Relation is partitioned into several fragments: system maintains several identical replicas of each such fragment. @ KIDS Labs 16
Advantages of Replication Availability: failure of site containing relation r does not result in unavailability of r is replicas exist. Parallelism: queries on r may be processed by several nodes in parallel. Reduced data transfer: relation r is available locally at each site containing a replica of r. @ KIDS Labs 17
Disadvantages of Replication Increased cost of updates: each replica of relation r must be updated. Increased complexity of concurrency control: concurrent updates to distinct replicas may lead to inconsistent data unless special concurrency control mechanisms are implemented. One solution: choose one copy as primary copy and apply concurrency control operations on primary copy. @ KIDS Labs 18
Fragmentation Data can be distributed by storing individual tables at different sites Data can also be distributed by decomposing a table and storing portions at different sites called Fragmentation Fragmentation can be horizontal or vertical @ KIDS Labs 19
Why use Fragmentation? Usage - in general applications use views so it s appropriate to work with subsets Efficiency - data stored close to where it is most frequently used Parallelism - a transaction can divided into several sub-queries to increase degree of concurrency Security - data more secure - only stored where it is needed Disadvantages: Performance - may be slower Integrity - more difficult @ KIDS Labs 20
Horizontal Fragmentation Each fragment, T i, of table T contains a subset of the rows Each tuple of T is assigned to one or more fragments. Horizontal fragmentation is lossless @ KIDS Labs 21
Horizontal Fragmentation Example A bank account schema has a relation Account-schema = (branch-name, account-number, balance). It fragments the relation by location and stores each fragment locally: rows with branch-name = `Hillside` are stored in the Hillside in a fragment @ KIDS Labs 22
Vertical Fragmentation Each fragment, T i, of T contains a subset of the columns, each column is in at least one fragment, and each fragment includes the key: T i = attr_list i (T) T = T 1 T 2.. T n All schemas must contain a common candidate key (or superkey) to ensure lossless join property. A special attribute, the tuple-id attribute may be added to each schema to serve as a candidate key. @ KIDS Labs 23
Vertical Fragmentation Example A employee-info schema has a relation employee-info schema = (designation, name, Employee-id, salary). It fragments the relation to put information in two tables for security concern. @ KIDS Labs 24
Types of Data Replication Push Replication updating site sends changes to other sites Pull Replication receiving sites control when update messages will be processed @ KIDS Labs 25
Types of Push Replication Snapshot Replication - Changes periodically sent to master site Master collects updates in log Full or differential (incremental) snapshots Dynamic vs. shared update ownership Near Real-Time Replication - Broadcast update orders without requiring confirmation Done through use of triggers Update messages stored in message queue until processed by receiving site @ KIDS Labs 26
Issues for Data Replication Data timeliness high tolerance for out-of-date data may be required DBMS capabilities if DBMS cannot support multi-node queries, replication may be necessary Performance implications refreshing may cause performance problems for busy nodes Network heterogeneity complicates replication Network communication capabilities complete refreshes place heavy demand on telecommunications @ KIDS Labs 27
Horizontal Partitioning Different rows of a table at different sites Advantages - Data stored close to where it is used efficiency Local access optimization better performance Only relevant data is available security Unions across partitions ease of query Disadvantages Accessing data across partitions inconsistent access speed No data replication backup vulnerability @ KIDS Labs 28
Vertical Partitioning Different columns of a table at different sites Advantages and disadvantages are the same as for horizontal partitioning except that combining data across partitions is more difficult because it requires joins (instead of unions) @ KIDS Labs 29
Figure 13-6 Distributed processing system for a manufacturing company @ KIDS Labs 30
Distributed DBMS Distributed database requires distributed DBMS Functions of a distributed DBMS: Locate data with a distributed data dictionary Determine location from which to retrieve data and process query components DBMS translation between nodes with different local DBMSs (using middleware) Data consistency (via multiphase commit protocols) Global primary key control Scalability Security, concurrency, query optimization, failure recovery @ KIDS Labs 31
Figure 13-10: Distributed DBMS architecture @ KIDS Labs 32
Local Transaction Steps 1. Application makes request to distributed DBMS 2. Distributed DBMS checks distributed data repository for location of data. Finds that it is local 3. Distributed DBMS sends request to local DBMS 4. Local DBMS processes request 5. Local DBMS sends results to application @ KIDS Labs 33
Figure 13-10: Distributed DBMS Architecture (cont.) (showing local transaction steps) 1 2 3 5 4 Local transaction all data stored locally @ KIDS Labs 34
Global Transaction Steps 1. Application makes request to distributed DBMS 2. Distributed DBMS checks distributed data repository for location of data. Finds that it is remote 3. Distributed DBMS routes request to remote site 4. Distributed DBMS at remote site translates request for its local DBMS if necessary, and sends request to local DBMS 5. Local DBMS at remote site processes request 6. Local DBMS at remote site sends results to distributed DBMS at remote site 7. Remote distributed DBMS sends results back to originating site 8. Distributed DBMS at originating site sends results to application @ KIDS Labs 35
Figure 13-10: Distributed DBMS architecture (cont.) (showing global transaction steps) 2 1 8 3 7 4 6 5 Global transaction some data is at remote site(s) @ KIDS Labs 36
Distributed DBMS Transparency Objectives Location Transparency User/application does not need to know where data resides Replication Transparency User/application does not need to know about duplication Failure Transparency Either all or none of the actions of a transaction are committed Each site has a transaction manager Logs transactions and before and after images Concurrency control scheme to ensure data integrity Requires special commit protocol @ KIDS Labs 37
Two-Phase Commit Prepare Phase Coordinator receives a commit request Coordinator instructs all resource managers to get ready to go either way on the transaction. Each resource manager writes all updates from that transaction to its own physical log Coordinator receives replies from all resource managers. If all are ok, it writes commit to its own log; if not then it writes rollback to its log @ KIDS Labs 38
Two-Phase Commit (cont.) Commit Phase Coordinator then informs each resource manager of its decision and broadcasts a message to either commit or rollback (abort). If the message is commit, then each resource manager transfers the update from its log to its database A failure during the commit phase puts a transaction in limbo. This has to be tested for and handled with timeouts or polling @ KIDS Labs 39
Concurrency Control Concurrency Transparency Design goal for distributed database Timestamping Concurrency control mechanism Alternative to locks in distributed databases @ KIDS Labs 40
Query Optimization In a query involving a multi-site join and, possibly, a distributed database with replicated files, the distributed DBMS must decide where to access the data and how to proceed with the join. Three step process: 1 Query decomposition - rewritten and simplified 2 Data localization - query fragmented so that fragments reference data at only one site 3 Global optimization - Order in which to execute query fragments Data movement between sites Where parts of the query will be executed @ KIDS Labs 41
Evolution of Distributed DBMS Unit of Work - All of a transaction s steps. Remote Unit of Work SQL statements originated at one location can be executed as a single unit of work on a single remote DBMS @ KIDS Labs 42
Evolution of Distributed DBMS (cont.) Distributed Unit of Work Different statements in a unit of work may refer to different remote sites All databases in a single SQL statement must be at a single site Distributed Request A single SQL statement may refer to tables in more than one remote site May not support replication transparency or failure transparency @ KIDS Labs 43
Commit Protocols Commit protocols are used to ensure atomicity across sites Atomicity states that database modifications must follow an all or nothing rule. a transaction which executes at multiple sites must either be committed at all the sites, or aborted at all the sites. @ KIDS Labs 44
The Two-Phase Commit (2 PC) Protocol What is this? Two-phase commit is a transaction protocol designed for the complications that arise with distributed resource managers. Two-phase commit technology is used for hotel and airline reservations, stock market transactions, banking applications, and credit card systems. With a two-phase commit protocol, the distributed transaction manager employs a coordinator to manage the individual resource managers. The commit process proceeds as follows: @ KIDS Labs 45
Phase1: Obtaining a Decision Step 1 Coordinator asks all participants to prepare to commit transaction T i. C i adds the records <prepare T> to the log and forces log to stable storage (a log is a file which maintains a record of all changes to the database) sends prepare T messages to all sites where T executed @ KIDS Labs 46
Phase1: Making a Decision Step 2 Upon receiving message, transaction manager at site determines if it can commit the transaction if not: add a record <no T> to the log and send abort T message to C i if the transaction can be committed, then: 1). add the record <ready T> to the log 2). force all records for T to stable storage 3). send ready T message to C i @ KIDS Labs 47
Phase 2: Recording the Decision Step 1 T can be committed of C i received a ready T message from all the participating sites: otherwise T must be aborted. Step 2 Coordinator adds a decision record, <commit T> or <abort T>, to the log and forces record onto stable storage. Once the record is in stable storage, it cannot be revoked (even if failures occur) Step 3 Coordinator sends a message to each participant informing it of the decision (commit or abort) Step 4 Participants take appropriate action locally. @ KIDS Labs 48
Two-Phase Commit Diagram @ KIDS Labs 49
Costs and Limitations There have been two performance issues with two phase commit: If one database server is unavailable, none of the servers gets the updates. This is correctable through network tuning and correctly building the data distribution through database optimization techniques. @ KIDS Labs 50
@ KIDS Labs 51