delivering performant, reliable, and scalable apps with anypoint platform
TRANSCRIPT
6
Reliability
• Uptime
• Zero downtime upgrades
• Disaster Recovery
Data Service• Zero Message Loss
• Message Recovery
9
1) Take it seriously, plan for it.
• First class non-functional requirements
• Define early
• Be Quantitative
• Consider Growth Scenarios
11
3) Measure/Tune iteratively
• Measure
• Generate load
• Compare with baseline, SLA
• Identify hot spots
• Tune and profile
12
4) Remember the big picture
• Runtime Engine
• JVM
• Operating System
• File System
• Network
• Downstream Systems
14
• Asynchronous vs Synchronous
• Real-time vs Batch
• On-Premises vs Cloud
• Caching
Design and Deployment Options
16
Infrastructure 32-core Amazon EC2 RedHat Linux
Throughput 7600tps
Latency Transaction: 5msGateway: 2 ms
API Gateway performance
18
Scalability: Best Practices
• Scale Up, Scale Out
• Stateful vs Stateless
• Synchronous
• Asynchronous
21
Reliability: Best Practices
• Degrees of reliability- At least once- Once and only once {XA}
• Reliable transactions- JMS, JDBC, VM-Endpiont- Multi-resource transactions- Business vs Infrastructure Exceptions
23
Out of the Box Reliability
• Transactional Scope
• Retry Mechanisms
• Until Successful
• First Successful
• Persistent Queues• HA Clustering• In-Memory
Replication
24
On-Prem versus Cloud
• Anypoint Fabric• Redundant Platform
• Intelligent Healing
• Zero Downtime Updates
• Data Center Redundancy
On-Prem • In-memory data grid
• HA Clustering
CloudHub
26
Customer PaaS performance
6,000+ Mule VMs deployed to test 500+ Mule VMs deployed in production200+ Service consumers in production200+ Service providers in production600+ Service contracts in production9.5 million messages per day850,000 messages per hour peak
30
Mule 3.5 & 3.6: Performance Tuning Guides
Mule 3.6 https://www.mulesoft.com/lp/whitepaper/soa/performance-tuning-guide-mule-36xMule 3.5 https://www.mulesoft.com/lp/whitepaper/soa/performance-tuning-guide