Upload
johan-ahlen
View
128
Download
5
Embed Size (px)
Citation preview
• Johan Åhlén
• SolidQ Nordisk CTO
• Ordförande SQL Server användarförening i Sverige
• Ansvarig för prestanda-SM på SQL Server
• SQL Server MVP sedan 2011
• Inblandad i ett par böcker
Blog: www.johanahlen.info
Presentatör
5 myter om In-Memory OLTP• ”All data försvinner om man drar ut sladden”• “Hekaton är en NoSQL databas”• “Hekaton är en förbättrad DBCC PINTABLE”• “Inga anpassningar behövs av befintlig kod”• “Alla applikationer går 100 gånger
snabbare”
Mission critical and cloud performance Cloud-ready Self-service BIPerformance
and productivityMission critical
Modern DB platform
The evolution of the Microsoft data platform
SQL Server2000
SQL Server2005
SQL Server2008
SQL Server2008 R2
SQL Server2012
XML ● HA
Management studio ● Mirroring
Compression ● Policy-based mgmt. ● Resource Governor
PowerPivot (In-Memory) ● SharePoint integration ● Master data services
AlwaysOn ● ColumnStore Indexes ● Data quality services ● Power View ● Cloud
SQL Server2014In-memory OLTP ● Clustered ColumnStore Indexes ● Performance and scale ● Hybrid cloud optimized
Key trends impacting in-memory design CPU Performance Flattening OutDecreasing Cost of Memory$/GB of Memory
1
10
100
1,000
10,000
100,000
1,000,000
US$
/GB
1970197519801985199019952000200520100
1
10
100
1,000
10,000
100,000
1,000,000
10,000,000
Perf/Clock (ILP)
Power
Clock Speed (MHz)
Transistors (000)
Write speed
of 200 million rowsin 15 minutes
Real-time data access
In-memory built-inKey Benefits Leverage familiar tools
No costly add-ons
Works seamlessly with existing SQL Server features
Use the same tools across services New! In-Memory
Relational Data Services
Microsoft Azure Infrastructure Services
Disk-based Relational Data Services
Familiar Dev & Management Tools
The entire DB doesn’t need to be In-Memory
10xFaster performance
In-memory flexibilityKey Benefits Minimize capex as data volumes grow
Access In-Memory and on-disk with a single query
Don’t need to rewrite entire app with
scalability and reduced
operating costs
Exponential
growth
On-Disk
ApplicationWarm
and hot data
Cold Data Single Query
Key Benefits Optimized table structures
No locks or latches with 100% data durability
Up to 30x transactional performance gains
In-memory for increased throughput & speed
16x faster transactions
“To describe Hekaton in two words, it’s wicked fast.”
Rick Kutschera, Bwin
Before After
Greater throughput with no locks or latches
OLTPStored Procedures
30x faster transactionsNatively compile stored procedures in-memory
App
35x faster transactions
Jämförelse
14
In-Memory OLTP
Hög skrivhastighet
Hög läshastighet
Kan skrivas till disk
Okomprimerad
Måste rymmas i minnet
ColumnStore Indexes
Låg skrivhastighet
Hög läshastighet
Skrivs alltid till disk
Komprimerad
Behöver inte rymmas i minnet
Saker som supporteras av In-Memory OLTP ”version 1.0”
15
Lagring på disk (om du vill)
AlwaysOn
Resource Governor
SSIS
Service Broker
PowerShell
DACPAC och BACPAC