Chapter 19 Distributed Databases Heterogeneous and Homogeneous Databases

Chapter 19: Distributed Databases Heterogeneous and Homogeneous Databases Distributed Data Storage Distributed Transactions Commit Protocols Concurrency Control , Availability <- omitted #Distributed query processing Heterogeneous distributed database Cloud-based databases Database System Concepts - 6 th Edition 19. 1

Distributed Database System A distributed database system consists of loosely coupled sites that share no physical component. Database systems that run on each site are independent of each other. Transactions may access data at one or more sites. Database System Concepts - 6 th Edition 19. 2

Types of Distributed Databases In a homogeneous distributed database l All sites have identical software l Are aware of each other and agree to cooperate in processing user requests. l Each site surrenders part of its autonomy in terms of right to change schemas or software l Appears to user as a single system In a heterogeneous distributed database l Different sites may use different schemas and software 4 Difference in schema is a major problem for query processing 4 Difference in software is a major problem for transaction processing l Sites may not be aware of each other and may provide only limited facilities for cooperation in transaction processing Database System Concepts - 6 th Edition 19. 3

Distributed Data Storage Assume relational data model Replication l System maintains multiple copies of data, stored in different sites, for faster retrieval and fault tolerance. Fragmentation l Relation is partitioned into several fragments stored in distinct sites Replication and fragmentation can be combined l Relation is partitioned into several fragments: system maintains several identical replicas of each such fragment. Database System Concepts - 6 th Edition 19. 4

Data Replication A relation or fragment of a relation is replicated if it is stored redundantly in two or more sites. Full replication of a relation is the case where the relation is stored at all sites. Fully redundant databases are those in which every site contains a copy of the entire database. Database System Concepts - 6 th Edition 19. 5

Data Replication (Cont. ) Advantages of Replication l Availability: failure of site containing relation r does not result in unavailability of r if replicas exist. l 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. Disadvantages of Replication l Increased cost of updates: each replica of relation r must be updated. l l Increased complexity of concurrency control: concurrent updates to distinct replicas may lead to inconsistent data unless special concurrency control mechanisms are implemented. 4 One solution: choose one copy as primary copy and apply concurrency control operations on primary copy Database System Concepts - 6 th Edition 19. 6

Data Fragmentation Division of relation r into fragments r 1, r 2, …, rn which contain sufficient information to reconstruct relation r. Horizontal fragmentation: each tuple of r is assigned to one or more fragments Vertical fragmentation: the schema for relation r is split into several smaller schemas l All schemas must contain a common candidate key (or superkey) to ensure lossless join property. l A special attribute, the tuple-id attribute may be added to each schema to serve as a candidate key. Database System Concepts - 6 th Edition 19. 7

The account relation branch_name account_number balance Hillside A-305 500 Hillside A-226 336 Hillside A-155 62 Valleyview A-177 205 Valleyview A-402 10000 Valleyview A-408 1123 Valleyview A-639 750 Database System Concepts - 6 th Edition 19. 8

Horizontal Fragmentation of account Relation branch_name Hillside account_number A-305 A-226 A-155 balance 500 336 62 account 1 = branch_name=“Hillside” (account ) branch_name Valleyview account_number A-177 A-402 A-408 A-639 balance 205 10000 1123 750 account 2 = branch_name=“Valleyview” (account ) Database System Concepts - 6 th Edition 19. 9

The customer_info relation tuple_id Account Balance _number Customer_ name Branch_name 1 A-305 500 Lowman Hillside 2 A-226 336 Camp Hillside 3 A-177 205 Camp Valleyview 4 A-402 10000 Kahn Valleyview 5 A-155 62 Kahn Hillside 6 A-408 1123 Kahn Valleyview 7 A-639 750 Green Valleyview Database System Concepts - 6 th Edition 19. 10

Vertical Fragmentation of customer_info Relation branch_name customer_name tuple_id Lowman 1 Hillside Camp 2 Hillside Camp 3 Valleyview Kahn 4 Valleyview Kahn 5 Hillside Kahn 6 Valleyview Green 7 Valleyview deposit 1 = branch_name, customer_name, tuple_id (customer_info ) account_number balance tuple_id 500 A-305 1 336 A-226 2 205 A-177 3 10000 A-402 4 62 A-155 5 1123 A-408 6 750 A-639 7 deposit 2 = account_number, balance, tuple_id (customer_info ) Database System Concepts - 6 th Edition 19. 11

Practice How to get the original account relation from the two fragments? How to get the original customer_info relation from the two fragments? Database System Concepts - 6 th Edition 19. 12

Advantages of Fragmentation Horizontal: l allows parallel processing on fragments of a relation l allows a relation to be split so that tuples are located where they are most frequently accessed Vertical: l allows parallel processing on a relation l allows tuples to be split so that each part of the tuple is stored where it is most frequently accessed l tuple-id attribute allows efficient joining of vertical fragments Vertical and horizontal fragmentation can be mixed. l Fragments may be successively fragmented to an arbitrary depth. Database System Concepts - 6 th Edition 19. 13

Data Transparency Data transparency: users should be unaware of the details of how and where the data items are stored in a distributed system Several forms of transparency: l Fragmentation transparency 4 Users l Replication transparency 4 Users l need not know how a relation is fragmented view each data object as logically unique Location transparency 4 Users Database System Concepts - 6 th Edition need not know the physical location of the data 19. 14

Naming of Data Items - Criteria 1. Every data item must have a system-wide unique name. 2. It should be possible to find the location of data items efficiently. 3. It should be possible to change the location of data items transparently. 4. Each site should be able to create new data items autonomously. Database System Concepts - 6 th Edition 19. 15

Centralized Scheme - Name Server Structure: l name server assigns all names l each site maintains a record of local data items l sites ask name server to locate non-local data items Advantages: l satisfies naming criteria 1 -3 Disadvantages: l does not satisfy naming criterion 4 l name server is a potential performance bottleneck l name server is a single point of failure Database System Concepts - 6 th Edition 19. 16

Use of Aliases Alternative to centralized scheme: each site prefixes its own site identifier to any name that it generates i. e. , site 17. account. l Fulfills having a unique identifier, and avoids problems associated with central control. l However, fails to achieve network transparency. Solution: Create a set of aliases for data items; Store the mapping of aliases to the real names at each site. The user can be unaware of the physical location of a data item, and is unaffected if the data item is moved from one site to another. Database System Concepts - 6 th Edition 19. 17

Distributed Transactions Transaction may access data at several sites. Each site has a local transaction manager responsible for: l Maintaining a log for recovery purposes l Participating in coordinating the concurrent execution of the transactions executing at that site. Each site has a transaction coordinator, which is responsible for: l Starting the execution of transactions that originate at the site. l Distributing subtransactions at appropriate sites for execution. l Coordinating the termination of each transaction that originates at the site, which may result in the transaction being committed at all sites or aborted at all sites. Database System Concepts - 6 th Edition 19. 18

Transaction System Architecture Database System Concepts - 6 th Edition 19. 19

System Failure Modes Failures unique to distributed systems: l Failure of a site. l Loss of massages 4 Handled by network transmission control protocols such as TCP-IP l Failure of a communication link 4 Handled by network protocols, by routing messages via alternative links l Network partition 4 A network is said to be partitioned when it has been split into two or more subsystems that lack any connection between them – Note: a subsystem may consist of a single node Network partitioning and site failures are generally indistinguishable. Database System Concepts - 6 th Edition 19. 20

Commit Protocols Commit protocols are used to ensure atomicity across sites l a transaction which executes at multiple sites must either be committed at all the sites, or aborted at all the sites. l not acceptable to have a transaction committed at one site and aborted at another The two-phase commit (2 PC) protocol is widely used The three-phase commit (3 PC) protocol is more complicated and more expensive, but avoids some drawbacks of two-phase commit protocol. This protocol is not used in practice. (omitted) Database System Concepts - 6 th Edition 19. 21

Two Phase Commit Protocol (2 PC) Assumes fail-stop model – failed sites simply stop working, and do not cause any other harm, such as sending incorrect messages to other sites. Execution of the protocol is initiated by the coordinator after the last step of the transaction has been reached. The protocol involves all the local sites at which the transaction executed Let T be a transaction initiated at site Si, and let the transaction coordinator at Si be Ci Database System Concepts - 6 th Edition 19. 22

Phase 1: Obtaining a Decision Coordinator asks all participants to prepare to commit transaction T. l Ci adds the records <prepare T> to the log and forces log to stable storage l sends prepare T messages to all sites at which T executed Upon receiving message, transaction manager at that site determines if it can commit the transaction l if not, add a record <no T> to the log and send abort T message to Ci l if the transaction can be committed, then: 4 add the record <ready T> to the log 4 force all log records for T to stable storage 4 send ready T message to Ci Database System Concepts - 6 th Edition 19. 23

Phase 2: Recording the Decision T can be committed if Ci received a ready T message from all the participating sites: otherwise T must be aborted. Coordinator adds a decision record, <commit T> or <abort T>, to the log and forces record onto stable storage. Once the log record is on stable storage, it is irrevocable (even if failures occur) Coordinator sends a message to each participant informing it of the decision (commit or abort) Participants take appropriate action locally. Database System Concepts - 6 th Edition 19. 24

#Handling of Failures - Site Failure When site Sk recovers, it examines its log to determine the fate of transactions active at the time of the failure. Log contains <commit T> record: site executes redo (T) Log contains <abort T> record: site executes undo (T) Log contains <ready T> record: site must consult Ci to determine the fate of T. l If T committed, redo (T) l If T aborted, undo (T) Log contains no control records concerning T: it implies that Sk failed before responding to the prepare T message from Ci l since the failure of Sk precludes the sending of such a response, Ci must abort T l Sk must execute undo (T) Database System Concepts - 6 th Edition 19. 25

#Handling of Failures-Coordinator Failure If coordinator fails while the commit protocol for T is executing then participating sites must decide on T’s fate: 1. If an active site contains a <commit T> record in its log, then T must be committed. 2. If an active site contains an <abort T> record in its log, then T must be aborted. 3. If some active participating site does not contain a <ready T> record in its log, then the failed coordinator Ci cannot have decided to commit T. Can therefore abort T. 4. If none of the above cases holds, then all active sites must have a <ready T> record in their logs, but no additional control records (such as <abort T> of <commit T>). In this case active sites must wait for Ci to recover, to find decision. Blocking problem: active sites may have to wait for failed coordinator to recover. Database System Concepts - 6 th Edition 19. 26

#Handling of Failures - Network Partition If the coordinator and all its participants remain in one partition, the failure has no effect on the commit protocol. If the coordinator and its participants belong to several partitions: l Sites that are not in the partition containing the coordinator think the coordinator has failed, and execute the protocol to deal with failure of the coordinator. 4 No harm results, but sites may still have to wait for decision from coordinator. l The sites that are in the same partition as the coordinator, think that the sites in other partitions have failed, and follow the usual commit protocol. 4 Again, Database System Concepts - 6 th Edition no harm results 19. 27

Trading Consistency for Availability Consistency in Databases (ACID): l Database has a set of integrity constraints l A consistent database state is one where all integrity constraints are satisfied l Each transaction run individually on a consistent database state must leave the database in a consistent state Consistency in distributed systems with replication l Strong consistency: a schedule with read and write operations on a replicated object should give results and final state equivalent to some schedule on a single copy of the object, with order of operations from a single site preserved l Weak consistency (several forms) Database System Concepts - 6 th Edition 19. 28

Availability Traditionally, availability of centralized server For distributed systems, availability of system to process requests l For large system, at almost any point in time there’s a good chance that 4 a node is down 4 Network partitioning Distributed consensus algorithms will block during partitions to ensure consistency l Many applications require continued operation even during a network partition 4 Even Database System Concepts - 6 th Edition at cost of consistency 19. 29

Brewer’s CAP Theorem Three properties of a system Consistency (all copies have same value) Availability (system can run even if parts have failed) Via replication Partition-tolerance (network can break into two or more parts, each with active systems that can’t talk to other parts) Brewer’s CAP Theorem: You can have at most two of these three properties for any distributed system. Very large systems will partition at some point Choose one of consistency or availability Traditional database choose consistency Most Web applications choose availability Except for specific parts such as order processing Database System Concepts - 6 th Edition 19. 30

BASE properties BASE (Basically Available, Soft state, Eventual consistency), as opposed to ACID Soft state: copies of a data item may be inconsistent Eventually Consistent – copies becomes consistent at some later time if there are no more updates to that data item When no updates occur for a long period of time, eventually all updates will propagate through the system and all the nodes will be consistent Widely used in social-networking systems, etc. 4 PNUTS allows inconsistent reads to reduce latency – Critical for many applications 4 But update protocol (via master) ensures consistency over availability Database System Concepts - 6 th Edition 19. 31

#Distributed Query Processing For centralized systems, the primary criterion for measuring the cost of a particular strategy is the number of disk accesses. In a distributed system, other issues must be taken into account: l The cost of a data transmission over the network. l The potential gain in performance from having several sites processing parts of the query in parallel. Database System Concepts - 6 th Edition 19. 32

#Query Transformation Translating algebraic queries on fragments. l It must be possible to construct relation r from its fragments l Replace relation r by the expression to construct relation r from its fragments Consider the horizontal fragmentation of the account relation into account 1 = branch_name = “Hillside” (account ) account 2 = branch_name = “Valleyview” (account ) The query branch_name = “Hillside” (account ) becomes branch_name = “Hillside” (account 1 account 2) which is optimized into branch_name = “Hillside” (account 1) branch_name = “Hillside” (account 2) Database System Concepts - 6 th Edition 19. 33

#Example Query (Cont. ) Since account 1 has only tuples pertaining to the Hillside branch, we can eliminate the selection operation. Apply the definition of account 2 to obtain branch_name = “Hillside” ( branch_name = “Valleyview” (account ) This expression is the empty set regardless of the contents of the account relation. Final strategy is for the Hillside site to return account 1 as the result of the query. Database System Concepts - 6 th Edition 19. 34

#Simple Join Processing Consider the following relational algebra expression in which the three relations are neither replicated nor fragmented account depositor branch account is stored at site S 1 depositor at S 2 branch at S 3 For a query issued at site SI, the system needs to produce the result at site SI Database System Concepts - 6 th Edition 19. 35

#Possible Query Processing Strategies Ship copies of all three relations to site SI and choose a strategy for processing the entire locally at site SI. Ship a copy of the account relation to site S 2 and compute temp 1 = account depositor at S 2. Ship temp 1 from S 2 to S 3, and compute temp 2 = temp 1 branch at S 3. Ship the result temp 2 to SI. Devise similar strategies, exchanging the roles S 1, S 2, S 3 Must consider following factors: l amount of data being shipped l cost of transmitting a data block between sites l relative processing speed at each site Database System Concepts - 6 th Edition 19. 36

#Semijoin Strategy Let r 1 be a relation with schema R 1 stores at site S 1 Let r 2 be a relation with schema R 2 stores at site S 2 Evaluate the expression r 1 r 2 and obtain the result at S 1. Compute temp 1 R 2 (r 1) at S 1. 2. Ship temp 1 from S 1 to S 2. 3. Compute temp 2 r 2 temp 1 at S 2 4. Ship temp 2 from S 2 to S 1. 5. Compute r 1 Database System Concepts - 6 th Edition temp 2 at S 1. This is the same as r 1 19. 37 r 2.

#Formal Definition The semijoin of r 1 with r 2, is denoted by: r 1 r 2 it is defined by: R 1 (r 1 Thus, r 1 r 2 ) r 2 selects those tuples of r 1 that contributed to r 1 In step 3 above, temp 2=r 2 r 2. r 1. For joins of several relations, the above strategy can be extended to a series of semijoin steps. Database System Concepts - 6 th Edition 19. 38

#Join Strategies that Exploit Parallelism Consider r 1 r 2 r 3 r 4 where relation ri is stored at site Si. The result must be presented at site S 1. r 1 is shipped to S 2 and r 1 shipped to S 4 and r 3 S 2 ships tuples of (r 1 S 4 ships tuples of (r 3 r 2 is computed at S 2: simultaneously r 3 is r 4 is computed at S 4 r 2) to S 1 as they produced; r 4) to S 1 Once tuples of (r 1 r 2) and (r 3 r 4) arrive at S 1, (r 1 r 2) (r 3 r 4 ) is computed in parallel with the computation of (r 1 r 2) at S 2 and the computation of (r 3 r 4) at S 4. Database System Concepts - 6 th Edition 19. 39

Heterogeneous Distributed Databases Many database applications require data from a variety of preexisting databases located in a heterogeneous collection of hardware and software platforms Data models may differ (hierarchical, relational, etc. ) Transaction commit protocols may be incompatible Concurrency control may be based on different techniques (locking, timestamping, etc. ) System-level details almost certainly are totally incompatible. A multidatabase system is a software layer on top of existing database systems, which is designed to manipulate information in heterogeneous databases l Creates an illusion of logical database integration without any physical database integration Database System Concepts - 6 th Edition 19. 40

Advantages Preservation of investment in existing l hardware l system software l Applications Local autonomy and administrative control Allows use of special-purpose DBMSs Step towards a unified homogeneous DBMS l Full integration into a homogeneous DBMS faces 4 Technical difficulties and cost of conversion 4 Organizational/political difficulties – Organizations do not want to give up control on their data – Local databases wish to retain a great deal of autonomy Database System Concepts - 6 th Edition 19. 41

Unified View of Data Agreement on a common data model l Typically the relational model Agreement on a common conceptual schema l Different names for same relation/attribute l Same relation/attribute name means different things Agreement on a single representation of shared data l E. g. , data types, precision, l Character sets 4 ASCII 4 Sort vs EBCDIC order variations Agreement on units of measure Variations in names l E. g. , Köln vs Cologne, Mumbai vs Bombay Database System Concepts - 6 th Edition 19. 42

Query Processing Several issues in query processing in a heterogeneous database Schema translation l Write a wrapper for each data source to translate data to a global schema l Wrappers must also translate updates on global schema to updates on local schema Limited query capabilities l Some data sources allow only restricted forms of selections 4 E. g. , l web forms, flat file data sources Queries have to be broken up and processed partly at the source and partly at a different site Removal of duplicate information when sites have overlapping information l Decide which sites to execute query Global query optimization Database System Concepts - 6 th Edition 19. 43

Mediator Systems Mediator systems are systems that integrate multiple heterogeneous data sources by providing an integrated global view, and providing query facilities on global view l Unlike full fledged multidatabase systems, mediators generally do not bother about transaction processing l But the terms mediator and multidatabase are sometimes used interchangeably l The term virtual database is also used to refer to mediator/multidatabase systems Database System Concepts - 6 th Edition 19. 44

Cloud computing A new concept in computing that emerged in the late 1990 s and the 2000 s. First, software as a service Vendors of software services provided specific customizable applications that they hosted on their own machines Then, generic computers as a service l Clients runs its own software, but runs it on vendor’s computers. l These machines are called virtual machines, which are simulated by software that allows a single real computer to simulate several independent computers l Clients can add machines as needed to meet demand release them at times of light load. l Other services l Data storage services, map services, and other services can be accessed using a Web-service API. Venders of cloud service l Traditional computing vendors, Amazon, Google Database System Concepts - 6 th Edition 19. 45

Data Storage on the Cloud-based database l Web applications need to store and retrieve data for very large numbers of users l Traditional parallel databases not designed to scale to 1000’s of nodes (and expensive) l Value availability and scalability over consistency l Storing and retrieving data items by key value is minimum functionality 4 Key-value stores Systems for data storage on the cloud l Bigtable from Google l HBase, an open source clone of Bigtable l Dynamo, which is a key-value storage system from Amazon l Simple Storage Service (S 3) from Amazon l Cassandra from Facebook l Sherpa/PNUTs from Yahoo! Database System Concepts - 6 th Edition 19. 46

Key value store It does not need extensive query language support. Two primitive functions of key-value stores: l put(key, value): store values with an associated key l get(key): retrieve the stored value associated with the specified key 4 Bigtable (later) additionally provide range queries on key values. 4 Multiple versions of data may be stored, by adding a timestamp to the key. An example application l The profile of a user needs to be accessible to many different application that are run by an organization. l The profile contains many attributes, and there are frequent additions to the attributes stored in the profile l Some attributes may contain complex data. Database System Concepts - 6 th Edition 19. 47

Data Representation It needs to provide flexibility in the set of attributes that a record contains (no schema), and the types of these attributes (complex types) l XML, JSON Example of JSON: { “ID”: “ 22222”, “name”: { “firstname”: “Albert”, “lastname” “Einstein” }, “deptname”: “Physics”, “children”: [ {“firstname”: “Hans”, “lastname”: “Einstein”}, {“firstname”: Eduard”, “lastname”: “Einstein”} ] } Big. Table has its own data model (the next page) Database System Concepts - 6 th Edition 19. 48

Big. Table A record is split into component attributes that are stored separately. The key for an attribute value consists of (record-identifier, attribute- name). Each attribute value is just a string. Example: A record with identifier “ 22222”, can have multiple attribute names such as “name. firstname”, “deptname”, “children[1]. firstname”, “children[2]. lastname”. To fetch all attributes of a record, a prefix-match query consisting of just the record identifier, is used. The record identifier can itself be structured hierarchically l A single instance of Bigtable can store data for multiple application, with multiple tables per application, by simply prefixing the application name and table name to the record identifier. Database System Concepts - 6 th Edition 19. 49

Partitioning and Retrieving Data Unlike regular parallel database, it is usually not possible to decide on a partitioning function ahead of time. Therefore, it partitions data into small units, called tablets (a tablet is a fragment of a table, usually hundreds of megabytes). The partitioning is done on the search key, so that a request for a specific key value is directed to a single tablet. A tablet controller site tracks the partitioning function, to map a get/put request to tablets, and map from tablets to sites. The site to which a tablet is assigned acts as the master site. l All updates are routed through this site, and then propagated to replicas The partitioning of data is not fixed, but happens dynamically. l if a tablet grows too big as data are inserted, it is broken into smaller parts l if the load (get/put operations) on a tablet is excessive, the tablet may be broken into smaller tablets, which can be distributed across two or more sites to share the load. Database System Concepts - 6 th Edition 19. 50

PNUS - a cloud data storage system Database System Concepts - 6 th Edition 19. 51

Challenges with Cloud-based Database advantages Do not need to build a computing infrastructures from scratch l Essential for certain applications Disadvantage l Additional communication cost like traditional distributed database system l The physical location of data is under the control of the vendor, which is unaware 4 Hard to perform query optimization l Replication is under the control of the vendor l 4 Hard to ensure the latest version of data are read l Data held by another organization are risked in terms of security and legal liability Many issues are still studied. Database System Concepts - 6 th Edition 19. 52