2. Data Expertise / Lynn Langit
Practicing Architect
• Cloud Deployments (Azure, AWS, Google)
Technical author / trainer
•
•
•
•
Google Cloud Developer Series
SQL Server 2012 Developer Series
Cloudera Certified Developer
2 books on SQL Server BI
Industry awards
•
•
•
Microsoft – MVP for SQL Server
Google – GDE for Cloud Platform
10Gen – Master for MongoDB
Former MSFT FTE
•
4 years
3. What and Why AWS?
Market leader
AWS
Amazon’s cloud
Large Set of
services
• Compute
• Data
• More
• In market longest
• Usually cheapest
• Most often used
in production
17. RDS vs. EC2 for SQL Server
Why RDS
costs more
• Provisioned IO –
performance guarantees
• Scheduled backups
• Point in time restores
• Scheduled maintenance
windows
• Full use of all SQL tools,
SSMS, Profiler, DTA, etc…
• Supports Availability Groups
(requires 2012 Enterprise)
• Cross-regional snapshots
24. New Services - AWS:Invent
Kinesis – real-time processing
of streaming Big Data (into
AppStream – deliver streaming
applications to clients from
AWS
CloudTrail – capture AWS API
calls
RDS addition – now supports
PostgreSQL
Workspaces – Virtual Desktops
for PC or Mac
24
30. Demo – AWS SDK
Add-in for Visual Studio and .NET
30
31. Cloud Database Services by Vendor
AWS
Google
Microsoft
Virtual Machines
EC2
GCE – Linux only
Azure VM
Cloud RDBMS
RDS - SQL Server,
MySQL, Oracle
Redshift - Postgres
mySQL > MariaDB
SQL Azure
NoSQL buckets
Key-Value stores
EBS
S3
Glacier
DynamoDB
Cloud Storage
HR Datastore on GAE
Azure Blobs
Azure Tables
Pipelines
Data Pipelines
Via APIs only
SSIS (on-premises)
Document
MongoDB on EC2
None
MongoDB on Windows
Azure
Hadoop MapReduce
or Dremel
MapReduce on EC2
using S3
Big Query
HDInsight (HDFS)
Other
Datasets
Streaming
Machine Learning
Kinesis
EBS volumes
w/datasets
Freebase
Translation API
Full-text search
Prediction API
StreamInsight
Azure Marketplace
34. Creative Financing
Regular Pricing
• Use what you need and no more, i.e. instance size, storage size…
• Watch for price drops – RDS price decrease this week
Smart EC2 Instance Usage
• Pause EC2 instances to reduce compute charges
• Delete EC2 instances to reduce storage charges
Vanity Pricing
• Set pricing alerts
• Use spot pricing
• Re-selling compute / storage
38. Tip: Use Pricing Calculators
Example – from RightScale ‘PlanForCloud’
38
39. Conclusions
EC2 for testing, training and production (IaaS)
S3 for archiving R/W
Glacier for archiving W fast & cheap, R slow & expensive
RDS for HA SQL Server
Redshift for Data Warehousing on demand
DynamoDB for fast NoSQL – on SSDs
Elastic Map Reduce for easy Hadoop MapReduce