SQL Server Columnstore Performance Tuning

Eric N HansonPrincipal Program ManagerMicrosoft Corporation

DBI409

Session Objectives and Takeaways

Session Objective(s): Learn to make query performance go from good to outstanding

Batch mode is crucial to speedupsGood application design can work around limits

Why pay attention?

Your customer with well-tuned xVelocity columnstore application

Your customer with poorly-tuned xVelocity columnstore application

demo

Outer Join Performance Limit

Fundamentals of xVelocity Columnstore Performance

Each column stored separately, compressedBatch mode query executionSegment elimination

Columnstore physical structure

• Segment (a.k.a. row group) = 1 million rows

• Table partition has 1 or more segments

• Each column segment is a BLOB

• Dictionaries are BLOBs• Catalog has one row per

column segment with min/max values

segment 1

segment Ndictionaries

catalog

…

Batch vs. Row Mode

Batch mode processing

Process ~1000 rows at a timeBatch stored in vector form

Vector operators implementedFilter, hash join, hash aggregation

Greatly reduced CPU time (7 to 40X)

8

bit

map o

f qu

alif

yin

g

row

s

Column vectors

Batch object

#1 Takeaway!

Make sure most of the work of the query happens in batch mode.

Segment Elimination

column_id

segment_id

min_data_id max_data_id

1 1 20120101 20120131

1 2 20120115 20120215

1 3 20120201 20120228

• Segment (rowgroup)= 1 million row chunk• Min, Max kept for each column in a segment• Scans can skip segments based on this

select Date, count(*) from dbo.Purchase where Date >= 20120201 group by Date

skipped

DOs

Use star schemaPut columnstores on large tables onlyInclude every column of table in columnstore indexUse integer surrogate keys for joinsStructure your queries as star joins with grouping and aggregation as much as possible

DON’T

Join on string columnsFilter on string columns of columnstore tableJoin pairs of very large tables if you don’t have toUse NOT IN <subquery> on columnstore tableUse UNION ALL to combine columnstore tables with other tables

Outer Join Workaround

Outer join prevents batch processingRewrite query to do most work in batch mode

DEMO

IN and EXISTs Issue & Workaround

Using IN and EXISTS with subqueries can prevent batch mode executionIN ( <list of constants> ) typically works fine DEMO

Union All Issue & Workaround

UNION ALL often prevents batch modeWorkaround

Push GROUP BY and aggregation over UNION ALLDo final GROUP BY and aggregation of resultsCalled “local-global aggregation”

DEMO

Scalar Aggregates Issue & Workaround

Aggregate without group by doesn’t get batch processingselect count(*) from dbo.Purchase 1.2 secondsWorkaroundwith CountByDate (Date, c) 0.093 secondsas ( select Date, count(*) from dbo.Purchase group by Date ) select sum(c) from CountByDate;

Multiple DISTINCT aggregates issue & workaround

Table Spool if query has 2 or more DISTINCT aggregatesSpool takes time to writeSpool write is single threadedReads of spool in plan are single threaded, in row modeWorkaround:

form each DISTINCT agg in separate subqueryJoin results on grouping keys

DEMO

SQL Server 2012runs queries with 1 DISTINCT aggand 1 or more non-distinct aggs in batch mode without any spool!

Physical DB Design, Loading, and Index Management

Adding Data Using Partition Switching

Columnstores must be partition-aligned Partition switching fully supportedTo add data daily

Partition by dayEvery day

Split last partitionCreate staging table and columnstore index itSwitch it in

Avoids costly drop/rebuild

Trickle Loading with Columnstore Indexes

Master table (columnstore)Delta table (rowstore)Query using UNION ALL local-global aggregation workaroundAdd Delta to Master nightly DEMO

Avoid Nonclustered B-trees

Covering B-trees no longer neededExtra B-trees can cause optimizer to choose poor planSave spaceReduce ETL time

Eliminating Unsupported Data Types

Omit column, orModify column type to supported type

Reduce precision of numerics to 18 digits or lessConvert guid’s to intsReduce precision of datetimeoffset to 2 or lessConvert hierarchyid to int or string

Achieving Fast Columnstore Index Builds

One thread per segmentLow memory throttles parallelismConsider

high min server memory setting Set REQUEST_MAX_MEMORY_GRANT_PERCENT to 50 or moreAdd memoryOmit columnsNormalize fact columns to dimensionsVertically partition

sys.dm_exec_query_memory_grants shows DOP

Not enough memory to build a columnstore index?

Same ideas apply as for achieving fast buildsRun out of memory during execution, not startup?

Errors 701 or 802 indicate thisWorkaround: reduce DOP explicitly, e.g.create columnstore index <name> on <table>(<columns>) with(maxdop = 1);

Maximizing the Benefits

of Segment Elimination

Verifying Segment Elimination

You may want to check if segments are eliminatedWays to get segment elimination confirmation:

Xevent sqlserver.column_store_segment_eliminate Text output messages

Trace flag 646 outputs segment elimination messagesTrace flag 3605 directs messages to error log file

DEMO

Ensuring segment elimination by date

Use clustered B-tree on date Columnstore inherits order

Or, partition by dateOrdering by load date, ship date, order date etc. can all work

Dates are naturally correlated

Partitioning and Segment Elimination -- Examplepartition 1

Date = 20120301segment 1 min(Date) = 20120301, max(Date) = 20120301segment 2 “ segment 3 “segment 4 “segment 5 “

partition 2

Date = 20120302segment 1 min(Date) = 20120302, max(Date) = 20120302segment 2 “ segment 3 “ segment 4 “ segment 5 “

Select …From FactWhere Date = 20120301

Multi-dimensional segment eliminationPre-sort or partition on a keydrawn from >1 dimensionE.g. YYYYMMRRCan be persisted computed column not used in columnstore

select f.region_id, avg(f.duration)from fact_CDR f where f.region_id = 1 and f.date_id between 20120101 and 20120131 group by f.region_id

Additional Tuning Considerations

String performance issuesString filters don’t get pushed to storage engine

more batches to processdefeats segment elimination

Joining on string columns is slowFactor strings out to dimensions

Date LicenseNum Measure

20120301 XYZ123 100

20120302 ABC777 200

Date LicenseId Measure

20120301 1 100

20120302 2 200

LicenseId LicenseNum

1 XYZ123

2 ABC777

Forcing use or non-use of Columnstores

Query hintOPTION(IGNORE_NONCLUSTERED_COLUMNSTORE_INDEX)

Index hint… FROM F WITH(index=MyColumnStore) …… FROM F WITH(index=MyClusteredBtree) …

Summary

Key’s to fast processingBatch modeSegment elimination

Make sure most work is done in batch mode!Can work around read-only property by

PartitioningUsing Delta-file for trickle load

Future work will reduce need for tuning

Related Content

Columnstore Tuning Guide (your reading homework )

http://social.technet.microsoft.com/wiki/contents/articles/sql-server-columnstore-performance-tuning.aspx

Columnstore FAQ http://social.technet.microsoft.com/wiki/contents/articles/sql-server-columnstore-index-faq.aspx

Introductory Columnstores Talk Video TechEd 2011 session DBI312

Customer Case StudiesSee list in FAQ

Track Resources

@sqlserver@ms_teched

mvaMicrosoft Virtual Academy

SQL Server 2012 Eval Copy

Get Certified!

Hands-On Labs

Resources

Connect. Share. Discuss.

http://northamerica.msteched.com

Learning

Microsoft Certification & Training Resources

www.microsoft.com/learning

TechNet

Resources for IT Professionals

http://microsoft.com/technet

Resources for Developers

http://microsoft.com/msdn

Complete an evaluation on CommNet and enter to win!

Please Complete an Evaluation Your feedback is important!

Multipleways to Evaluate Sessions

Scan the Tagto evaluate thissession now on myTechEd Mobile

© 2012 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to

be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS

PRESENTATION.