Distributed Query Processing

Distributed Query Processing

Agenda

• Recap of query optimization

• Transformation rules for P&D systems

• Memoization

• Queries in heterogenous systems

• Query evaluation strategies

• Eddies

• Open-ended and stream-based queries

Introduction

• Alternative ways of evaluating a given query– Equivalent expressions– Different algorithms for each operation (Chapter 13)

• Cost difference between a good and a bad way of evaluating a query can be enormous– Example: performing a r X s followed by a selection r.A = s.B is

much slower than performing a join on the same condition

• Need to estimate the cost of operations– Depends critically on statistical information about relations which

the database must maintain– Need to estimate statistics for intermediate results to compute cost

of complex expressions

Introduction (Cont.)

Relations generated by two equivalent expressions have the same set of attributes and contain the same set of tuples, although their attributes may be ordered differently.

Introduction (Cont.)

• Generation of query-evaluation plans for an expression involves several steps:

1. Generating logically equivalent expressions

• Use equivalence rules to transform an expression into an equivalent one.

2. Annotating resultant expressions to get alternative query plans

3. Choosing the cheapest plan based on estimated cost

• The overall process is called cost based optimization.

Equivalence Rules1. Conjunctive selection operations can be deconstructed

into a sequence of individual selections.

2. Selection operations are commutative.

3. Only the last in a sequence of projection operations is needed, the others can be omitted.

4. Selections can be combined with Cartesian products and theta joins.

a. (E1 X E2) = E1 E2

b. 1(E1 2 E2) = E1 1 2 E2

))(())((1221

EE

))(()(2121

EE

)())))((((121EE ttntt

Equivalence Rules (Cont.)

5. Theta-join operations (and natural joins) are commutative.E1 E2 = E2 E1

6. (a) Natural join operations are associative:

(E1 E2) E3 = E1 (E2 E3)

(b) Theta joins are associative in the following manner:

(E1 1 E2) 2 3 E3 = E1 2 3 (E2 2 E3) where 2 involves attributes from only E2 and E3.

Pictorial Depiction of Equivalence Rules

Equivalence Rules (Cont.)

7. The selection operation distributes over the theta join operation under the following two conditions:(a) When all the attributes in 0 involve only the attributes of one of the expressions (E1) being joined.

0E1 E2) = (0(E1)) E2

(b) When 1 involves only the attributes of E1 and 2 involves only the attributes of E2.

1 E1 E2) = (1(E1)) ( (E2))

Equivalence Rules (Cont.)

8. The projections operation distributes over the theta join operation as follows:

(a) if involves only attributes from L1 L2:

(b) Consider a join E1 E2. – Let L1 and L2 be sets of attributes from E1 and E2,

respectively. – Let L3 be attributes of E1 that are involved in join

condition , but are not in L1 L2, and– let L4 be attributes of E2 that are involved in join

condition , but are not in L1 L2.

))(())(()( 2......12.......1 2121EEEE LLLL

)))(())((().....( 2......121 42312121EEEE LLLLLLLL

Equivalence Rules (Cont.)9. The set operations union and intersection are commutative

E1 E2 = E2 E1 E1 E2 = E2 E1

(set difference is not commutative).

10. Set union and intersection are associative.

(E1 E2) E3 = E1 (E2 E3)

(E1 E2) E3 = E1 (E2 E3)

11. The selection operation distributes over , and –.

(E1 – E2) = (E1) – (E2)

and similarly for and in place of –

Also: (E1 – E2) = (E1) – E2

and similarly for in place of –, but not for 12. The projection operation distributes over union

L(E1 E2) = (L(E1)) (L(E2))

Multiple Transformations (Cont.)

Optimizer strategies

• Heuristic

– Apply the transformation rules in a specific order such that the cost converges to a minimum

• Cost based

– Simulated annealing

– Randomized generation of candidate QEP

– Problem, how to guarantee randomness

Memoization Techniques

• How to generate alternative Query Evaluation Plans?

– Early generation systems centred around a tree representation of the plan

– Hardwired tree rewriting rules are deployed to enumerate part of the space of possible QEP

– For each alternative the total cost is determined

– The best (alternatives) are retained for execution

– Problems: very large space to explore, duplicate plans, local maxima, expensive query cost evaluation.

– SQL Server optimizer contains about 300 rules to be deployed.

Memoization Techniques

• How to generate alternative Query Evaluation Plans?

– Keep a memo of partial QEPs and their cost.

– Use the heuristic rules to generate alternatives to built more complex QEPs

– r1 r2 r3 r4

r1 r2 r2 r3 r3 r4 r1 r4

xLevel 1 plans

r3 r3Level 2 plans

Level n plans r4

r2 r1

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:

– The cost of a data transmission over the network.

– The potential gain in performance from having several sites process parts of the query in parallel.

Transformation rules for distributed systems

• Primary horizontally fragmented table:

– Rule 9: The union is commutative E1 E2 = E2 E1

– Rule 10: Set union is associative. (E1 E2) E3 = E1 (E2 E3)

– Rule 12: The projection operation distributes over union

L(E1 E2) = (L(E1)) (L(E2))

• Derived horizontally fragmented table:

– The join through foreign-key dependency is already reflected in the fragmentation criteria

Transformation rules for distributed systems

Vertical fragmented tables:

– Rules: Hint look at projection rules

Optimization in Par & Distr

• Cost model is changed!!!

– Network transport is a dominant cost factor

• The facilities for query processing are not homogenous distributed

– Light-resource systems form a bottleneck

– Need for dynamic load scheduling

Simple Distributed 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 S1

• depositor at S2

• branch at S3

• For a query issued at site SI, the system needs to produce the result at site SI

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 S2 and compute temp1 = account depositor at S2. Ship temp1 from S2 to S3, and compute temp2 = temp1 branch at S3. Ship the result temp2 to SI.

• Devise similar strategies, exchanging the roles S1, S2, S3

• Must consider following factors:– amount of data being shipped – cost of transmitting a data block between sites– relative processing speed at each site

Semijoin Strategy

• Let r1 be a relation with schema R1 stores at site S1

Let r2 be a relation with schema R2 stores at site S2

• Evaluate the expression r1 r2 and obtain the result at S1.

1. Compute temp1 R1 R2 (r1) at S1.

2. Ship temp1 from S1 to S2.

3. Compute temp2 r2 temp1 at S2

4. Ship temp2 from S2 to S1.

5. Compute r1 temp2 at S1. This is the same as r1 r2.

Formal Definition

• The semijoin of r1 with r2, is denoted by:

r1 r2

• it is defined by:

R1 (r1 r2)

• Thus, r1 r2 selects those tuples of r1 that contributed to r1 r2.

• In step 3 above, temp2=r2 r1.

• For joins of several relations, the above strategy can be extended to a series of semijoin steps.

Join Strategies that Exploit Parallelism

• Consider r1 r2 r3 r4 where relation ri is stored at site Si. The

result must be presented at site S1.

• r1 is shipped to S2 and r1 r2 is computed at S2: simultaneously r3 is

shipped to S4 and r3 r4 is computed at S4

• S2 ships tuples of (r1 r2) to S1 as they produced;

S4 ships tuples of (r3 r4) to S1

• Once tuples of (r1 r2) and (r3 r4) arrive at S1 (r1 r2) (r3

r4) is computed in parallel with the computation of (r1 r2) at S2 and

the computation of (r3 r4) at S4.

Query plan generation

• Apers-Aho-Hopcroft

– Hill-climber, repeatedly split the multi-join query in fragments and optimize its subqueries independently

• Apply centralized algorithms and rely on cost-model to avoid expensive query execution plans.

Query evaluators

Query evaluation strategy

• Pipe-line query evaluation strategy

– Called Volcano query processing model

– Standard in commercial systems and MySQL

• Basic algorithm:

– Demand-driven evaluation of query tree.

– Operators exchange data in units such as records

– Each operator supports the following interfaces:– open, next, close

• open() at top of tree results in cascade of opens down the tree.

• An operator getting a next() call may recursively make next() calls from within to produce its next answer.

• close() at top of tree results in cascade of close down the tree

Query evaluation strategy

• Pipe-line query evaluation strategy

– Evaluation:

• Oriented towards OLTP applications– Granule size of data interchange

• Items produced one at a time

• No temporary files– Choice of intermediate buffer size allocations

• Query executed as one process

• Generic interface, sufficient to add the iterator primitives for the new containers.

• CPU intensive

• Amenable to parallelization

Query evaluation strategy

• Materialized evaluation strategy

– Used in MonetDB

– Basic algorithm:

• for each relational operator produce the complete intermediate result using materialized operands

– Evaluation:

• Oriented towards decision support queries

• Limited internal administration and dependencies

• Basis for multi-query optimization strategy

• Memory intensive

• Amendable for distributed/parallel processing

Eddies: Continuously Adaptive Query processing

R. Avnur, J.M. Hellerstein

UCB

ACM Sigmod 2000

Problem Statement

• Context: large federated and shared-nothing databases

• Problem: assumptions made at query optimization rarely hold during execution

• Hypothesis: do away with traditional optimizers, solve it thru adaptation

• Focus: scheduling in a tuple-based pipeline query execution model

Problem Statement Refinement

• Large scale systems are unpredictable, because

– Hardware and workload complexity,

• bursty servers & networks, heterogenity, hardware characteristics

– Data complexity,

• Federated database often come without proper statistical summaries

– User Interface Complexity

• Online aggregation may involve user ‘control’

Research Laboratory setting

• Telegraph, a system designed to query all data available online

• River, a low level distributed record management system for shared-nothing databases

• Eddies, a scheduler for dispatching work over operators in a query graph

The Idea

• Relational algebra operators consume a stream from multiple sources to produce a new stream

• A priori you don’t now how selective- how fast- tuples are consumed/produced

• You have to adapt continuously and learn this information on the fly

• Adapt the order of processing based on these lessons

The Idea

JOIN JOIN

JOIN

next

next next

next

next next

The Idea

• Standard method: derive a spanning tree over the query graph

• Pre-optimize a query plan to determine operator pairs and their algorithm, e.g. to exploit access paths

• Re-optimization a query pipeline on the fly requires careful state management, coupled with

– Synchronization barriers

• Operators have widely differing arrival rates for their operands– This limits concurrency, e.g. merge-join algorithm

– Moments of symmetry

• Algorithm provides option to exchange the role of the operands without too much complications

– E.g switching the role of R and S in a nested-loop join

Nested-loop

R

s

Join and sorting

• Index-joins are asymmetric, you can not easily change their role

– Combine index-join + operands as a unit in the process

• Sorting requires look-ahead

– Merge-joins are combined into unit

• Ripple joins

– Break the space into smaller pieces and solve the join operation for each piece individually

– The piece crossings are moments of symmetry

The Idea

Tuple buffer JOIN

JOIN JOIN

Eddienext next next next

next next next

next

Rivers and Eddies

Eddies are tuple routers that distribute arriving tuples to interested operators

– What are efficient scheduling policies?

• Fixed strategy? Random ? Learning?

Static Eddies• Delivery of tuples to operators can be hardwired in the Eddie to reflect

a traditional query execution plan

Naïve Eddie

• Operators are delivered tuples based on a priority queue

• Intermediate results get highest priority to avoid buffer congestion

Observations for selections

• Extended priority queue for the operators– Receiving a tuple leads to a credit increment

– Returning a tuple leads to a credit decrement

– Priority is determined by “weighted lottery”

• Naïve Eddies exhibit back pressure in the tuple flow; production is limited by the rate of consumption at the output

• Lottery Eddies approach the cost of optimal ordering, without a need to a priory determine the order

• Lottery Eddies outperform heuristics

– Hash-use first, or Index-use first, Naive

Observations

• The dynamics during a run can be controlled by a learning scheme

– Split the processing in steps (‘windows’) to re-adjust the weight during tuple delivery

• Initial delays can not be handled efficiently

• Research challenges:

– Better learning algorithms to adjust flow

– Aggressive adjustments

– Remove pre-optimization

– Balance ‘hostile’ parallel environment

– Deploy eddies to control degree of partitioning (and replication)

The tranquil database scene

• Traditional DBMS – data stored in finite, persistent data sets, SQL-based applications to manage and access it

OLTP-webapplication

‘Ad-hoc’reporting

RDBMS

Data entryapplication

The tranquil database scene

• The user community grows and MANY wants up-to-the-second (aggregate) information from the database

OLTP-webapplication

‘Ad-hoc’reporting

RDBMS

Data entryapplication

Informedreporting

The tranquil database scene

• Database entry is taken over by a remote device which issues a high-volume of update transactions

OLTP-webapplication

‘Ad-hoc’reporting

RDBMS

Dataentryapplication

Data entryapplication

Informedreporting

The tranquil database scene

• Database entry is taken over by MANY remote devices which issues a high-volume of update transactions

OLTP-webapplication

‘Adhoc’reporting

RDBMS

Dataentryapplication

Dataentryapplication

Informedreporting

The tranquil database scene

• Database solutions can not carry the weight

OLTP-webapplication

‘Adhoc’reporting

RDBMS

Dataentryapplication

Dataentryapplication

Informedreporting

Application domains

• Personalized financial tickers

• Personalized information delivery

• Personalized environment control

• Business to business middelware

• Web-services application based on XML exchange

• Monitoring the real-world environment (pollution, traffic)

• Monitoring the data flow in an ISP

• Monitoring web-traffic behaviour

• Monitoring the load on a telecom switch

• Monitoring external news-feeds

Application domains

• Personalized financial tickers

• Personalized information delivery

• Personalized environment control

• Business to business middelware

• Web-services application based on XML exchange

• Monitoring the real-world environment (pollution, traffic)

• Monitoring the data flow in an ISP

• Monitoring web-traffic behaviour

• Monitoring the load on a telecom switch

• Monitoring external news-feeds

Application domains

• Personalized

• Personalized

• Personalized

• middelware

• on XML exchange

• Monitoring

• Monitoring

• Monitoring

• Monitoring

• Monitoring

QUERYING

STREAM UPDATE

WEB SERVICES

Continuous queries

• Continous query – the user observes the changes made to the database through a query

– Query registration once

– Continously up-to-date answers.Continuousqueries

RDBMS

Data Streams

• Data streams

– The database is in constant bulk load mode

– The update rate is often non-uniform

– The entries are time-stamped

– The source could be web-service, sensor, wrapped source

DSMS

Dataentryapplication

DSMS

Data Stream Management Systems (DSMS) support high volume update streams and real-time response to ad-hoc complex queries.

What can be salvaged from the DBMS core technology ?What should be re-designed from scratch ?

DSMS

Dataentryapplication

Informedreporting

DBMS versus DSMS

• Persistent relations

• Transaction oriented

• One-time queries

• Precise query answering

• Access plan determines physical database design

• Transient streams

• Query orientation

• Continuous queries

• Best-effort query answering

• Unpredictable data characteristics

Old technology to rescue?

• Many stream based applications are low-volume with simple queries

– Thus we can live with automatic query ‘refresh’

• Triggers are available for notification of changes

– They are hooked up to simple changes to the datastore

– There is no technology to merge/optimize trigger groups

Outline of remainder

• Query processing over multiple streams

• Organizing hundreds of ad-hoc queries

• Sensor-network based querying

DSMS

DSMS

DSMS

A stream application

• [Widom] Consider a network traffic system for an ISP

• with customer link and backbone link and two streams

• keeping track of the IP traffic

A stream application

• [Widom] Consider a network traffic system for an ISP

• with customer link and backbone link and two streams

• keeping track of the IP traffic

TPc(saddr, daddr, id, length, timestamp)

TPb(saddr, daddr, id, length, timestamp)

PTc

PTb

DSMS

A stream application

• Q1 Compute the load on the backbone link averaged over one minute period and notify the operator when the load exceeds a threshold T

Select notifyoperator(sum(length))From PTbGroup By getminute(timestamp)Having sum(length) >T

With low stream flow it could be handled with a DBMS trigger,Otherwise sample the stream to get an approximate answer

A stream application

• Q2 Find the fraction of traffic on the backbone link coming from the customer network to check cause of congestion.

( Select count(*) From PTc as C, PTb as B Where C.saddr = B.saddr and C.daddr=B.daddr and C.id=B.id ) /( Select count(*) From PTb)

Both streams might require an unbounded resource to perform the join, which could be avoided with an approximate answer and synopsis

A stream application

• Q3 Monitor the 5% source-to-destination pairs in terms of traffic on the backbone.

With Load As (Select saddr, daddr,sum(length) as traffic From PTb Group By saddr,daddr)

Select saddr, daddr, trafficFrom Load as l1Where (Select count(*) From Load as l2

Where l2.traffic <l1.traffic) > (Select 0.95*count(*) From Load)

Order By Traffic This query contains ‘blocking’ operators

STREAM architecture

DSMSTPc

TPb

Answer AnswerStore

ScratchArea

Trash

The answer store area simply needs an integer

• Q1 Compute the load on the backbone link averaged over one minute period and notify the operator when the load exceeds a threshold T

Select notifyoperator(sum(length))

From PTb

Group By getminute(timestamp)

Having sum(length) >T

The scratch area should maintain part of the two streams to implement the join. Or a complete list of saddr and daddr.

• Q2 Find the fraction of traffic on the backbone link coming from the customer network to check cause of congestion.

( Select count(*)

From PTc as C, PTb as B

Where C.saddr = B.saddr and C.daddr=B.daddr

and C.id=B.id ) /

( Select count(*) From PTb)

Joining two tables

RelA

RelB

Nested loop join

Joining two tables

RelA

RelB

Nested loop join

Joining two stream

……..

PTb

Nested loop join

PTa

……..

An unbounded store would be required

Joining two stream

……..

PTb

merge join

PTa

……..

If the streams are ordered a simple merge join is possibleWith limited resource requirements

Joining two stream

……..

PTb

Join synopsis

PTa

……..

A statistical summary could provide an approximate answer

histogram

histogram

window

• Q3 Monitor the 5% source-to-destination pairs in terms of traffic on the backbone.

With Load As (Select saddr, daddr,sum(length) as traffic From PTb Group By saddr,daddr)

Select saddr, daddr, trafficFrom Load as l1Where (Select count(*) From Load as l2

Where l2.traffic <l1.traffic) > (Select 0.95*count(*) From Load)

Order By Traffic The scratch area should maintain part of the two streams to implement the join.

Finance

• [DeWitt] Consider a financial feed where thousands of clients can register arbitrary complex continues queries.– XML stream querying

DSMSXML

Finance

• Q5 Notify me whenever the price of KPN stock drops below 6 euro

Select notifyUser(name, price)

From ticker t1

Where t1.name = “KPN” and t1.price < 6

Finance

• Q5 Notify me whenever the price of KPN stock drops by 5% over the last hour

Select notifyUser(name, price)

From ticker t1,t2

Where t1.name = “KPN” and t2.name= t1.name

and getminutes(t1.timestamp-t2.timestamp) <60

and t1.price < 0.95 * t2.price

Finance

• Q6 Notify me whenever the price of KPN stock drops by 5% over the last hour and T-mobile remains constant

Select notifyUser(name, price)

From ticker t1,t2, t3,t4

Where t1.name = “KPN” and t2.name= t1.name

and getminutes(t1.timestamp-t2.timestamp) <60

and t1.price < 0.95 * t2.price

and t1.timestamp=t3.timestamp and t2.timestamp=t4.timestamp

and t3.name = “T-Mobile” and t4.name= t3.name

and getminutes(t3.timestamp-t4.timestamp) <60

and t3.price = t4.price

Query signatures

• Traditional SQL applications already use the notion of parameterised queries, I.e. some constants are replaced by a program variable.

– Subsequent calls use the same query evaluation plan

• In a DSMS we should recognize such queries as quick as possible

– Organize similar queries into a group

– Decompose complex queries into smaller queries

– Manage the amount of intermediate store

Finance

• Queries can be organized in groups using a signature and evaluation can be replaced by single multi-user request.

Select notifyUser(name, price)

From ticker t1

Where t1.name = “KPN” and t1.price < 6

Client Name Threshold Price

192.871.12.1 KPN 6

192.777.021 ING 12

Finance

• Queries can be organized in groups using a signature and evalution can be replaced by single multi-user request.

Select notifyUser(c.client, t1.name, t1.price)

From ticker t1, clients c

Where t1.name = c.name and t1.price < c.price

Client Name Threshold Price

192.871.12.1 KPN 6

192.777.021 ING 12

Finance

• Timer-based queries call for a stream window with incremental evaluation

• Multiple requests can be organized by time-table and event detection methods provided by database triggers.

Select notifyUser(name, price)

From ticker t1,t2

Where t1.name = “KPN” and t2.name= t1.name

and getminutes(t1.timestamp-t2.timestamp) <60

and t1.price < 0.95 * t2.price

Finance

• Complex queries can be broken down into independent components

Select notifyUser(name, price)

From ticker t1,t2, t3,t4

Where t1.name = “KPN” and t2.name= t1.name

and getminutes(t1.timestamp-t2.timestamp) <60

and t1.price < 0.95 * t2.price

and t1.timestamp=t3.timestamp and t2.timestamp=t4.timestamp

and t3.name = “T-Mobile” and t4.name= t3.name

and getminutes(t3.timestamp-t4.timestamp) <60

and t3.price = t4.price

Finance

• Intermediate results should be materialized. Can be integrated in tradition query evaluation schemes

t1.timestamp=t3.timestamp and t2.timestamp=t4.timestamp

Sensor networks

• [Madden] Sensor networks are composed of thousands of small devices, interconnected through radio links. This network can be queried.

– Sensors have limited energy

– Sensors have limited reachability

– Sensors can be ‘crushed’

DSMS

Aggregate Queries Over Ad-Hoc Wireless Sensor Networks

Sensor networks

• Q7 Give me the traffic density on the A1 for the last hour

Select avg(t.car)

From traffic t

Where t.segment in (Select segment From roads

Where name = “A1”)

Group By gethour(t.timestamp)

Sensor networks

• The sensors should organize themselves into a P2P infrastructure

• An aggregate query is broadcasted through the network

• Each Mote calculates a partial answer and sent it to its peers

• Peers aggregate the information to produce the final answer.

• Problems

– The energy to broadcast some information is high

– Tuples and partial results may be dropped

Conclusions and outlook

• Data stream management technology require changes in our expectation of a DBMS functionality

– Queries not necessarily provide a precise answer

– Queries continue as long as we are interested in their approximate result

– The persistent store not necessarily contains a consistent and timeless view on the state of the database.

Conclusions and outlook

• Datastream management technology capitalizes upon proven DBMS technology

• DSMS provide a basis for ambient home settings, sensor networks, and globe spanning information systems

• It is realistic to expect that some of the properties to support efficient datastream management will become part of the major products

– Multi query optimization techniques should be added.

Literature

• NiagaraCQ: A Scalable Contious Query System for Internet Databases, J. Chen, D.J. deWitt, F. Tian, Y. Wang, Wisconsin Univ.

• Streaming Queries over Streaming Data , Sirish Chandrasekaran, Michael J. Franklin, Univ Berkeley

• Continous Queries over Data Streams, S.Babu, J. Widom, Stanford University