Bay Area In-Memory Computing Meetup, Aug. 23, 2018

1. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.com
©2018 Heimdall Data, inc. www.heimdalldata.com
SQL Database Proxy
Roland Lee

2. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 2
Agenda
1) Database Proxy Introduction
2) Demo

3. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 3
Web scale Challenges
More users and applications implies
• Higher data volumes
• More tables, changes to schema
• Higher latency responses

4. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.com
Feature ProxySQL
Automated Failover
SQL Read/Write Splitting
Database Vendor Neutral
Automated Cache
invalidation
Reduces network latency
Database Proxy Vendors
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓

5. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 5
Heimdall Database Proxy
• No code changes
• Database vendor neutral

6. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 6
Heimdall Data Software Options
Application
Heimdall Data
Proxy
Vendor Database
Driver
Application Server
Runs as an
agent
Application
Heimdall Data
JDBC
Vendor JDBC
Driver
JDBC driver, .jar
file
Application Server
Any JDBC data source

7. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 7
Heimdall Distributed Database Proxy
Removes network
latency
Auto-caching
Auto-invalidation
Distributed, client
SQL Optimization

8. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 8
Heimdall Data Use Cases
Use case Customer Benefit
SQL Results Caching
• Auto-caching / Auto-invalidate
• NO code changes
Automated failover
• Faster failover for MySQL & SQL Server AlwaysOn
• PGPool-II Replacement
Horizontal Scale-out
• Better use of read replicas
• Lag detection to guarantee fresh data
• NO code changes
Auditing for Privacy
Compliance
• Logs data access: Who, what, when
• GDPR, SOX, PCI, HIPPA

9. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 9
Application Server
Application
Heimdall
Data
Heimdall Distributed Deployment
Application Servers
Application Server
Application
Heimdall
Data
Application Server
Application
Heimdall
Data
Application Server
Heimdall
Central
Console
• You choose the storage
• Heimdall auto-caches
• Heimdall auto-invalidates
Automated
DB Failover
Elasticache
SQL Analytics
Audit Logging

10. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.com
L2 Cache
Heimdall
DB Proxy
Removes network
latency
Application Server
Local
Cache
Application
SQL SQL
Application Server - SQL Caching
Elasticache

11. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 11
How it Works
1) Heimdall proxy intercepts and forwards SQL
traffic between app & database.
2) Heimdall caches SQL results only if there is a
performance benefit.
3) Heimdall acts as a look-aside cache
4) Not a write-through or read-through cache
5) Heimdall insures data consistency.

12. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 12
What to cache?
Uses real-time analysis and statistics on:
– Query frequency and variability
– Relative performance of Cache vs. Database
Provides:
– Auto-cache only if there is a performance benefit
– Cache recommendations and benefits

13. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 13
Heimdall Query Analytics
Very cacheable. 700 s per query

14. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 14
Heimdall Cache Rules Analytics
After: Cached response in 70 s per query. Reduced database load.
Before: Query response time 700 s per query. Significant repetitive database load.

15. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.com
©2018 Heimdall Data, inc. www.heimdalldata.com
Use Cases

16. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.comCopyright © 2018 Heimdall Data, Inc. www.heimdalldata.com 16
Use case #1: Horizontal Scale-out
1. Automate use of read replicas (read/write splits)
1. Replication lag detection to ensure data freshness
2. Scale out the database with no application changes!
Write
Application Server
Application Heimdall
Data
Read 1
Read 2
Read 3
Application Server

17. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.com
Data
Consumers
Heimdall Data
SQL Traffic Manager
OLTP
Analytics
Data Federation
Intelligent SQL Routing
Amazon Redshift

18. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.com
Heimdall
DB Proxy
Application Server
Application
SQL
Heimdall
Central
Console
Use case #3: End-to-End SQL Analytics
Identifies performance bottleneck
• Application?
• Database?
• Network?

19. Click to edit Master title style
Copyright © 2016 Heimdall Data, Inc.| www.heimdalldata.com
Thank You