PowerPoint Presentation
Velkommen og Microsoft har en touchskrm :-D
Erling SkaaleTSPMicrosoft
[email protected]
Todays menuIn-memory tablesWho and whenWhat and howUpdateable
columnstoreClusteredHybrid and the Cloud
Drive Real-Time Business with Real-Time Insights
Over 100x query speed and significant data compression
withIn-Memory ColumnStore
Up to 30x faster transaction processing with In-Memory OLTP
Greater performancewith In-Memory Analysis ServicesBillions of
rows per second with PowerPivot In-Memory for Excel4Faster
InsightsIN-MEMORY ANALYTICS Faster QueriesIN-MEMORY DWFaster
TransactionsIN-MEMORY OLTP
Now lets take a closer look at how we can impact your business
with our in-memory technology. We are the only provider to date
that can speed transactions as well as queries and insights with
in-memory technology optimized for each workload: OLTP, data
warehousing, and analytics.
With our new in-memory OLTP engine in SQL Server 2014, we have
customers that have seen up to 30x faster transaction processing. I
am not talking about query speed, but actual transaction write
speed, up to 30x faster. I know many of you might be thinking, well
Oracle and other database vendors are talking 100x. What they are
talking about is query speed, not transactional speed. We are the
only vendor that delivers an in-memory engine designed for OLTP
transaction performance increase.
Theres also built-in In-Memory columnstore for data warehousing
workloads to speed queries. We were already benchmarking over 100x
performance gains with many customers in the SQL Server 2012
release of the in-memory columnstore. With SQL Server 2014, the
in-memory columnstore gets even better, we will talk about that in
just a few minutes. Again, we can also increase query speed by over
100x.
Finally we offer business users the ability to analyze data and
data models much faster with built-in in-memory capabilities for
Excel through PowerPivot, and Analysis Services. The benefit is
that you can analyze billions of rows of data per second in Excel.
Meaning a business user can analyze data of nearly any size with
the tools they are most familiar with.
This is what we mean when we say driving real-time business with
real-time insights. We can significantly speed your transaction
business tied to your revenue stream. We can massively speed the
process to analyze both real-time transaction data, along with
historical and third party data from IT or business users. This is
why we are already seeing in-memory technologies transforming the
way businesses run.
2012 Microsoft Corporation. All rights reserved. Microsoft,
Windows, 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.4/7/20144
Key Benefits Optimized table structuresNo locks or latches with
100% data durabilityUp to 30x transactional performance
gainsIn-Memory for Increased Throughput & Speed16x faster
transactionsTo describe Hekaton in two words, its wicked fast.Rick
Kutschera, Bwin
BeforeAfter
Greater throughput with no locks or latches OLTPStored
Procedures30x faster transactionsNatively compile stored procedures
in-memoryApp
535x faster transactions
Now lets talk about how we increase both transactional speed and
throughput by removing contention in the database. Many of you
might be thinking, I could pin tables to memory in previous
versions of SQL Server, how is they any different or better? The
speed gains you have been hearing me talk about from SBI liquidity,
Ferranti are all comparing to previous versions of SQL Server
paging tables to memory. So why the massive speed gains? They key
reason is the table structures are now optimized to run in-memory,
there are no more paging of tables to memoryperiod. And there are
no more locks and latches which removes contention in the database.
This is how we can achieve transactional performance increases up
to 30x.
In addition to speed, we can also improve throughput because our
engineering team came up with an algorithm to remove locks and
latches without compromising durability. This means massive
reduction of contention in the database, which leads to increased
throughput as well as speed.
Bwin is an ISV in the online-gaming industry and for them,
transactions equates to revenue. With our unique in-memory OLTP
design point of optimized tables structures and no locks and
latches, they were able to improve transaction speed by 16x and
increase player capacity by 20x on the same hardware. Because
contention is significantly reduced, they were also able to cut
player response times from 50 milliseconds to 3 milliseconds. In
terms of business value that SQL Server 2014s in-memory OLTP
technology provided Bwin, this meant increased revenue,
significantly improved customer experience, and a greater number of
customers on the same infrastructure! This is why we feel in-memory
technology is transformational, its because of the significant
impact it can have on your business.
2012 Microsoft Corporation. All rights reserved. Microsoft,
Windows, 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.4/7/20145
Business Transformation with In-MemoryBusiness ChallengeUtility
ISV wanted to increase the processing power of its MECOMS solution
to handle a smart-meterdata floodKey Benefits Optimized utility
rates based on consumptionNew smart-appliances tie to smart meter
for cost effective energy utilizationIncreased database write speed
100%
BeforeWith In-Memory OLTP & Windows Azure HDInsight
5 million measurements per year500 million measurements per
dayMechanical metersMechanical & smart-meters
HDInsight200M rowsin 15 minNon-relational dataSQL Server
2014In-Memoryrelational data
Now lets show you how in-memory can be transformational to a
business. Ferranti computer systems designs software for utility
companies that is helping to revolutionize how electricity is
consumed and sold. In-memory technology is key to this
transformation.
So why do I say revolutionize how electricity is consumed and
sold? Imagine a world where your home appliancesdishwashers,
washing machines, dryersautomatically knew when electricity was
cheapest in your city and would turn on only during those hours.
Imagine a world with no more power outages because electricity was
so efficiently used. Thats great for consumers, and its also good
for the environment.
This is what Ferranti software is working with utility companies
to create. Utility companies are using smart meter data to help
improve efficiency and control consumption. The world of mechanical
meters to measure consumption is going away in favor of smart
meters that continually feed power consumption and quality data
back to the utility companies. But with these smart meters you have
a new problem, massive amounts of data being created that needs to
be quickly harnessed to set rates. This is where in-memory comes in
to playMicrosofts in-memory OLTP solution not only speeds
transactions but also removes contention to massively increase in
write speeds. In this case, Ferranti needed to quickly write more
than 200 million rows of relational data to a database, and with
SQL Server 2014 in-memory OLTP technology removing database
contention, they were able to do this in less than 15 minutes.
And because our in-memory technology was built-in, they could
use additional Microsoft data platform services in parallel. Like
Windows Azure HDInsight service to tackle non-relational big data
to help them compute utility rates. This new optimized electricity
consumption model based on smart meters is just the first step. It
has opened the door to smart appliances that communicate with smart
meters to determine the most cost effective times to run.
Significantly increasing write speeds and removing contention in
the database were key in helping Ferranti solve this big data
problem.Server & Tools Business 2012 Microsoft Corporation. All
rights reserved. Microsoft, Windows, 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.4/7/20146
SQL Server IntegrationSame manageability, administration &
development experienceIntegrated queries &
transactionsIntegrated HA and backup/restoreMain-Memory
OptimizedOptimized for in-memory dataIndexes (hash and range) exist
only in memoryNo buffer poolStream-based storage for durability
High ConcurrencyMulti-version optimistic concurrency control
with full ACID supportCore engine uses lock-free algorithmsNo lock
manager, latches or spinlocksT-SQL Compiled to Machine CodeT-SQL
compiled to machine code via C code generator and VCInvoking a
procedure is just a DLL entry-pointAggressive optimizations @
compile-timeSteadily declining memory price, NVRAMMany-core
processorsStalling CPU clock rateTCOHardware TrendsBusinessHybrid
Engine and Integrated ExperienceHigh Performance Data
OperationsFrictionless Scale-UpEfficient, Business-Logic
ProcessingBenefitsIn-Memory OLTP Tech PillarsDriversIn-Memory OLTP
arkitektur
7
Main-Memory OptimizedOptimized for in-memory dataIndexes (hash
and range) exist only in memoryNo buffer poolStream-based storage
for durabilitySteadily declining memory price, NVRAMHardware
trendsTable ConstructsFixed schema no ALTER TABLE, must
drop/recreate/reloadNo LOB datatypes; row size limited 8060No
constraints support (PK only)No Identity or Calculated columns, CLR
etcData and table size considerationsSize of tables = (row size * #
of rows) Size of hash index = (bucket_count * 8-bytes) Max size
SCHEMA_AND_DATA = 512 GBIO for Durability SCHEMA_ONLY vs.
SCHEMA_AND_DATAMemory Optimized Filegroup Data and Delta files
Transaction Log Database Recovery
Memory Optimized Tables Design Considerations8High performance
data operationsBenefitsIn-Memory OLTP Tech PillarsDrivers
8
Memory optimeret table vs normal
TechReady 16 2013 Microsoft Corporation. All rights reserved.
Microsoft, Windows, 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.4/7/20149
Indexing Memory-Optimized Tables
Memory Optimized tables support two types of indexes:
Non-Clustered Hash Index: This is amemory optimizedindex, it
does not support inequality operators as well as sort-order
matching operations.
Non-Clustered Index: This is adisk basedindex, fully supports
all normal index operations.
Thebucket_countindex parameter on a Non-Clustered Hash Index
dictates the size of the Hash table allocated for the index. This
parameter needs to be set carefully, as it will affect the
performance of the Memory Optimized table. Higher bucket count
could lead to larger memory utilization and longer scans, lower
bucket count could lead to performance degredation on lookups and
inserts. Microsoft recommendsthe bucket_count should be twice the
maximum number of unique index keys.
Indexing Memory-Optimized TablesThere are a few stipulations for
creating indexes on Memory-Optimized tables:
Only 8 indexesare allowed on a Memory Optimized table
Indexes cannot be added to a Memory Optimized table, instead the
table has to be dropped and re-created with the new index.
Primary Key is a requirementfor Memory Optimized tables.
All indexes are covering, which means they include all columns
in a table.
Indexes reference the (hashed) row directly, rather than
referencing the Primary Key.
Statistics Maintenance on Memory-Optimized TablesStatistics
behaves slightly differently on In-Memory tables compared to
disk-based tables, there are a few things to remember here:
Statistics are not updated automatically, it is recommended to
setup a manual regular statistics update operation on your
Memory-Optimized table.
sp_updatestats always runs a statistics update, unlike
disk-based table weresp_updatestats only updates statistics if
there has been modifications since last run.
Statistics update must always be specified as FULLSCANupdate,
rather than SAMPLED.Index key statistics are created when the table
is empty, it is recommended to always update statistics after
ETL.
Natively Compiled stored procedures need to be recreated when
statistics are updated, this is because execution plans for
Natively Compiled stored procedures consider statistics only once,
when the SPs are created.
Native Compiled Stored Proc - Migration
TechReady 16 2013 Microsoft Corporation. All rights reserved.
Microsoft, Windows, 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.4/7/201413
Accessing Memory Optimized TablesNatively compiled stored
proceduresAccess only memory optimized tablesMaximum
performanceLimited T-SQL surface areaWhen to useOLTP-style
operationsOptimize performance critical business logic
Interpreted T-SQL accessAccess both memory- and disk-based
tables Less performantVirtually full T-SQL surface areaWhen to
useAd hoc queriesReporting-style queriesSpeeding up app
migration
TechReady 16 2013 Microsoft Corporation. All rights reserved.
Microsoft, Windows, 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.4/7/201414
Clustered Columnstore Index
Clustered Columnstore Index
Update
Clustered Columnstore Index
Cloud Backup (SQL Server 2014)Whats being deliveredAn agent that
manages and automates SQL Server backup policyMain benefitLarge
scale management and no need to manage backup policyContext-aware
e.g. workload/throttlingMinimal knobs control retention
periodManage whole instance, or particular DBsLeverage Backup to
AzureInherently off-siteGeo-redundantMinimal storage costsZero
hardware managementExample:EXEC smart_admin.sp_set_db_backup
@database_name='TestDB', @storage_url=, @retention_days=30,
@credential_name='MyCredential', @enable_backup=1
18
Hybrid Cloud SolutionsFast disaster recovery (low RTO)Easy to
deploy and manageCloud burstingGreater global reachBetter isolation
of internal assetsSimplified Cloud Backup Cloud Disaster Recovery
Extend On-Premises Apps
Reduce CAPEX and OPEX with Cloud DROn-Premises Network
Secondary ReplicaDomain ControllerVPN Tunnel (Windows Azure
Virtual Network)
Asynchronous Commit
Primary Replica
Secondary Replica
Synchronous Commit
Windows Azure Run backups Run BI reports Manual or automaticAt
an instance level with point in time restoreMeasures DB usage
patterns to set backup frequency
Talk track in slide 23Server & Tools Business 2012 Microsoft
Corporation. All rights reserved. Microsoft, Windows, 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.4/7/201419
An event can cause on-premises SQL Server to become
unavailableTemporarily (e.g. gateway failure)Permanently (e.g.
flooding)
A disaster recovery site is expensiveSite rent +
maintenanceHardwareOpsWhy Do We Need Cloud DR for SQL Server?
TechReady 16 2013 Microsoft Corporation. All rights reserved.
Microsoft, Windows, 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.4/7/201420
2014 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.
