Eddies Continuously Adaptive Query processing R Avnur J
- Slides: 59
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 sharednothing 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 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 JOIN Tuple buffer JOIN next Eddie next
Rivers and 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)
Database streams: You only get one chance to look Prof. Dr. Martin Kersten CWI Amsterdam March 2003
The tranquil database scene • Traditional DBMS – data stored in finite, persistent data sets, SQL-based applications to manage and access it Data entry application OLTP-web application RDBMS ‘Ad-hoc’ reporting
The tranquil database scene • The user community grows and MANY wants up-to-thesecond (aggregate) information from the database Data entry application OLTP-web application ‘Ad-hoc’ reporting RDBMS Informed reporting
The tranquil database scene • Database entry is taken over by a remote device which issues a high-volume of update transactions Data entry application Dataentry application OLTP-web application ‘Ad-hoc’ reporting RDBMS Informed reporting
The tranquil database scene • Database entry is taken over by MANY remote devices which issues a high-volume of update transactions Dataentry application OLTP-web application ‘Adhoc’ reporting RDBMS Informed reporting
The tranquil database scene • Database solutions can not carry the weight Dataentry application OLTP-web application ‘Adhoc’ reporting RDBMS Informed reporting
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 • • QUERYING middelware on XML exchange Monitoring Monitoring WEB SERVICES STREAM UPDATE
Continuous queries • Continous query – the user observes the changes made to the database through a query – Query registration once – Continously up-to-date answers. Continuous queries 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 Dataentry application DSMS
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 ? Dataentry application DSMS Informed reporting
DBMS versus DSMS • Persistent relations • Transient streams • Transaction oriented • Query orientation • One-time queries • Continuous queries • Precise query answering • Best-effort query answering • Access plan determines physical database design • 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 DSMS • Organizing hundreds of ad-hoc queries DSMS • Sensor-network based querying 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 • Q 1 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 With low stream flow it could be handled with a DBMS trigger, Otherwise sample the stream to get an approximate answer
A stream application • Q 2 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 • Q 3 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, traffic From Load as l 1 Where (Select count(*) From Load as l 2 Where l 2. traffic <l 1. traffic) > (Select 0. 95*count(*) From Load) Order By Traffic This query contains ‘blocking’ operators
STREAM architecture TPc DSMS Answer Store Answer TPb Scratch Area Trash
• Q 1 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 answer store area simply needs an integer
• Q 2 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) The scratch area should maintain part of the two streams to implement the join. Or a complete list of saddr and daddr.
Joining two tables Rel. A Nested loop join Rel. B
Joining two tables Rel. A Nested loop join Rel. B
Joining two stream PTa ……. . Nested loop join PTb ……. . An unbounded store would be required
Joining two stream PTa ……. . merge join PTb ……. . If the streams are ordered a simple merge join is possible With limited resource requirements
Joining two stream window PTa ……. . histogram Join synopsis histogram PTb ……. . A statistical summary could provide an approximate answer
• Q 3 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, traffic From Load as l 1 Where (Select count(*) From Load as l 2 Where l 2. traffic <l 1. 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 • [De. Witt] Consider a financial feed where thousands of clients can register arbitrary complex continues queries. – XML stream querying XML DSMS
Finance • Q 5 Notify me whenever the price of KPN stock drops below 6 euro Select notify. User(name, price) From ticker t 1 Where t 1. name = “KPN” and t 1. price < 6
Finance • Q 5 Notify me whenever the price of KPN stock drops by 5% over the last hour Select notify. User(name, price) From ticker t 1, t 2 Where t 1. name = “KPN” and t 2. name= t 1. name and getminutes(t 1. timestamp-t 2. timestamp) <60 and t 1. price < 0. 95 * t 2. price
Finance • Q 6 Notify me whenever the price of KPN stock drops by 5% over the last hour and T-mobile remains constant Select notify. User(name, price) From ticker t 1, t 2, t 3, t 4 Where t 1. name = “KPN” and t 2. name= t 1. name and getminutes(t 1. timestamp-t 2. timestamp) <60 and t 1. price < 0. 95 * t 2. price and t 1. timestamp=t 3. timestamp and t 2. timestamp=t 4. timestamp and t 3. name = “T-Mobile” and t 4. name= t 3. name and getminutes(t 3. timestamp-t 4. timestamp) <60 and t 3. price = t 4. 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 notify. User(name, price) From ticker t 1 Where t 1. name = “KPN” and t 1. price < 6 Client Name Threshold 192. 871. 12. 1 KPN 6 192. 777. 021 12 ING Price
Finance • Queries can be organized in groups using a signature and evalution can be replaced by single multi-user request. Select notify. User(c. client, t 1. name, t 1. price) From ticker t 1, clients c Where t 1. name = c. name and t 1. price < c. price Client Name Threshold 192. 871. 12. 1 KPN 6 192. 777. 021 12 ING Price
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 notify. User(name, price) From ticker t 1, t 2 Where t 1. name = “KPN” and t 2. name= t 1. name and getminutes(t 1. timestamp-t 2. timestamp) <60 and t 1. price < 0. 95 * t 2. price
Finance • Complex queries can be broken down into independent components Select notify. User(name, price) From ticker t 1, t 2, t 3, t 4 Where t 1. name = “KPN” and t 2. name= t 1. name and getminutes(t 1. timestamp-t 2. timestamp) <60 and t 1. price < 0. 95 * t 2. price and t 1. timestamp=t 3. timestamp and t 2. timestamp=t 4. timestamp and t 3. name = “T-Mobile” and t 4. name= t 3. name and getminutes(t 3. timestamp-t 4. timestamp) <60 and t 3. price = t 4. price
Finance • Intermediate results should be materialized. Can be integrated in tradition query evaluation schemes t 1. timestamp=t 3. timestamp and t 2. timestamp=t 4. 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 • Q 7 Give me the traffic density on the A 1 for the last hour Select avg(t. car) From traffic t Where t. segment in (Select segment From roads Where name = “A 1”) Group By gethour(t. timestamp)
Sensor networks • The sensors should organize themselves into a P 2 P 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 • Niagara. CQ: A Scalable Contious Query System for Internet Databases, J. Chen, D. J. de. Witt, 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
- What is the role of eddy in adaptive query processing
- Ruby pier
- Query tree and query graph
- Query tree and query graph
- Iterative query vs recursive query
- What does continuously monitoring internal marketing
- What is compounded continuously
- How to find apy
- Compound interest problem example
- What term describes a slideshow that plays continuously?
- Continuously compounded interest
- Distributed query processing
- Objectives of query processing
- Measures of query cost in dbms
- Sql server intelligent query processing
- Distributed query processing in dbms
- Steps in query processing
- Distributed query processing
- Parsing and translation in query processing
- Distributed query processing
- Steps in query processing
- Distributed query processing
- Steps in query processing
- Sketch techniques for approximate query processing
- Steps in query processing
- Which algorithm
- Algorithms for query processing and optimization
- Batch processing vs interactive processing
- Fractal image
- Bottom up processing example
- What is point processing in digital image processing
- Histogram processing in digital image processing
- Digital image processing
- Unsharp masking matlab
- Top down vs bottom up processing
- Parallel processing vs concurrent processing
- Top-down processing example
- Primary and secondary processing of food
- Gloria suarez
- Nonlinear image processing
- Wpf adaptive layout
- Embedded systems in automobiles
- Adaptive value of behavior
- Dash dynamic adaptive streaming over http
- Adaptive optics
- Inert organizational culture
- Phylogenetisches artkonzept
- Adaptive learning ppt
- Sas customer intelligence
- Adaptive cruise control
- Adaptive immunity
- Adaptive case management definition
- What is mechanical isolation biology
- Adaptive capacity
- Adaptive replacement cache
- How to calculate deseasonalized demand
- Sitecore adaptive personalisation
- Types of selection
- Adaptive cbc
- Adaptive leadership development actions