Presentation is loading. Please wait.

Presentation is loading. Please wait.

Cloud Database Platforms for the SQL DBA

Similar presentations


Presentation on theme: "Cloud Database Platforms for the SQL DBA"— Presentation transcript:

1 Cloud Database Platforms for the SQL DBA
Warner Chaves Cloud Database Platforms for the SQL DBA

2 Bio DBA and Consultant for 11 years.
Previously an L3 DBA at HP in Costa Rica, now a Principal Consultant at Pythian in Ottawa, Ontario. Microsoft Data Platform MVP. Personal blog: sqlturbo.com Company: Pythian.com © 2016 Pythian. Confidential

3 SQLSaturday Sponsors! Titanium & Global Partner Gold Silver Bronze
Without the generosity of these sponsors, this event would not be possible! Please, stop by the vendor booths and thank them.

4 Agenda Amazon Google Microsoft
Goal: high-level overview of the main Database as a Service offerings in the major Public Cloud providers. Database Platforms as a Service options from: Amazon Google Microsoft © 2016 Pythian. Confidential

5 Scope Public Cloud providers.
Focus on Platform as a Service offerings. Focus on repositories or analytics services, not services that work on data in transit. Not an exhaustive list. High-level overviews. © 2016 Pythian. Confidential

6 Database as a service (PaaS)
No long term contracts (I only need my CC). I have nothing to do with the OS or below. I can scale up or down on demand. I can create and destroy on demand. Core operational tasks are all provider-managed. © 2016 Pythian. Confidential

7 The Corollas You know what you’re getting, what to expect.

8 Amazon RDS Automation layer built by Amazon on top of: MySQL, Oracle, PostgreSQL, SQL Server and Amazon’s Aurora. Very easy (if not the easiest) way to transition into DbaaS. Hourly rate dependent on license, compute and storage. © 2016 Pythian. Confidential

9 Amazon RDS Great for: lift and shift cloud migrations.
Watch out for: specific versions supported on RDS and specific technology limits. © 2016 Pythian. Confidential

10 Azure SQL Database Cloud-first SQL Server fork.
Lots of investment on elastic scale, horizontal scaling and helping the SaaS provider model. Fixed hourly-rate based on service tier per individual database or per database pool. Full instance support on the way. © 2016 Pythian. Confidential

11 Azure SQL Database Great for: leveraging existing SQL Server investment and T-SQL skills, SaaS model. Watch out for: properly choosing the service tier without proper SQL tuning. © 2016 Pythian. Confidential

12 Formula 1 Fit for purpose, without all the amenities.

13 Amazon DynamoDb NoSQL wide column table store.
Denormalized, flexible schema and complex types. Built for scale-out growth. Exposes a stream based API as well. Cost is based on storage, IO rate and streams read rate. © 2016 Pythian. Confidential

14 Amazon DynamoDb Great for: scale out of high ingest applications with known query patterns. Watch out for: schema impact on your application query patterns, indexing limits. © 2016 Pythian. Confidential

15 Azure CosmosDb NoSQL multi-model database.
Built-in support for partitioned collections and geo-replication. No schema restrictions on the JSON. SLAs defined on: availability, throughput, latency and consistency. Automatic indexing done by the system. Billing based on individual compute (RUs) plus collection storage. © 2016 Pythian. Confidential

16 Azure CosmosDb Great for: horizontal scaling with geo-replication built-in. Watch out for: storage consumption of automatic index policies, collection limits. © 2016 Pythian. Confidential

17 Google Spanner Fully managed “NewSQL” distributed relational database service Supports ACID transactions, joins, indexes. No support for FKs, uses ‘Interleaved tables’ instead. ANSI SQL Compatibility. Multi-region not yet implemented. Costing is based per node, per hour + storage + network egress charges between regions and internet facing.

18 Google Spanner Great for: Watch out for:
Relational database based applications that can no longer scale vertically. Watch out for: Not a pure relational engine, absent RI constructs such as foreign keys so migration is not direct. Indexes have to be called out explicitly.

19 The 18 Wheelers Heavy load of structured data.

20 Amazon Redshift Amazon’s modified PostgreSQL with columnar storage.
Relational MPP Data Warehouse with SQL and Python. Scales per node (both compute + storage) and is payed per node per hour. Lots of control on specific node configuration (cores, memory, SDD or HDD). © 2016 Pythian. Confidential

21 Amazon Redshift Great for: warehousing solution for all your AWS services. Watch out for: node count and configuration, copy issues due to S3 consistency, regular maintenance. © 2016 Pythian. Confidential

22 Azure SQL Data Warehouse
Relational, columnar, MPP SQL Server based service. Scales compute and storage independently. Allows pausing of the compute completely. No control of hardware, it’s 100% PaaS. Compute is based on a DWU unit. © 2016 Pythian. Confidential

23 Azure SQL Data Warehouse
Great for: warehousing solution for your Azure services, pause-friendly workloads. Watch out for: not all T-SQL data types supported yet. © 2016 Pythian. Confidential

24 Google Bigquery Can have tables, virtual tables, external tables.
Easy to use SQL querying interface. Streaming inserts directly into Bigquery supported as well. Hadoop can be attached but requires temporary data copy. Billed per storage, streaming rate and data read by queries ($5 per TB on base compute) © 2016 Pythian. Confidential

25 Google Bigquery Great for: one stop shop for streaming, relational, file based with no management. Watch out for: high compute queries cost, DML limits. © 2016 Pythian. Confidential

26 Demo

27 The Container Ships Can carry everything, in any shape or form.

28 Hadoop as a Service Azure HDInsight Amazon EMR Google Cloud DataProc
They all follow a similar pattern: pick machine models, deploy cluster, hook up to storage service. © 2016 Pythian. Confidential

29 Azure Data Lake Separated into a storage and an analytics service.
Storage has no limit on size of account or file. Can mix tables, files and external tables through U-SQL (mix of SQL and C#) or attach a Hadoop cluster. Analytics are done per job and can be scaled to increase compute (AU units). Billed by storage, AUs and jobs. © 2016 Pythian. Confidential

30 Azure Data Lake Great for: leveraging T-SQL and .NET skills for easy PaaS analytics. Watch out for: default limits of 50 AUs and 3 concurrent jobs. © 2016 Pythian. Confidential

31 Amazon Athena Serverless service
Interactive SQL Querying and metadata store Pay $5 per TB read from S3 (text, json, orc, parquet) Presto as the compute engine (open source in-memory distributed engine from Facebook), Hive for DDL. No mechanism to increase compute power (use compression, columnar formats or partitioning to go faster) © 2016 Pythian. Confidential

32 Amazon Athena Great for: maximum flexibility, quickly putting schema and querying files. Watch out for: cost of querying large uncompressed, not partitioned data sets. © 2016 Pythian. Confidential

33 Demo

34 Recap DBaaS platforms can improve SDLC velocity, TTM and improve ROI.
What we saw today: The Corolla’s: RDBMS at cloud scale Formula 1: NoSQL in the cloud 18 Wheelers: Data Warehousing in the cloud Container Ships: Big data in the cloud A modern data architecture and strategy will take advantage of many (all) of them.

35 Questions?

36 Thank You!!


Download ppt "Cloud Database Platforms for the SQL DBA"

Similar presentations


Ads by Google