building performant, reliable, and scalable integrations with mule esb
DESCRIPTION
How to build integrations with Mule ESB that meet your customers' needs around performance, scalability, and reliability. Presented with Rupesh Ramachandran at MuleSoft CONNECT 2014.TRANSCRIPT
1
2
Building Performant, Reliable, Scalable Integrations with Mule ESB
Ryan Hoegg, Integration Architect, ConfluexRupesh Ramachandran, Solutions Architect, MuleSoft
3
Service Level Agreement
First Class Requirements
Precision Matters
4
5
Reliability SLAs
AvailabilityUptimeTime to Recovery
Message LossMaximum GuaranteeDetectionRecovery
6
Scalability SLAs
CapacityPeakSustained(Degraded)
Message VolumeMessage Size
7
Performance SLAs
Response Time ThroughputConcurrencySustainability
8
Performance Tuning: Big picture
Mule ESBJVMOperating SystemFile SystemNetworkDownstream Systems
9
Performance: Best Practices
Asynchronous vs SynchronousReal-time vs BatchStateful vs Stateless (web scale)On-Premise vs iPaaS
10
Performance: Manage SLA’s
11
Performance: Case Study
Use Case: API GatewayXML to JSON transformsMixed Payload Sizes
12
Performance: Case Study
Test Case Mule ESB 3.5 EE as API Gateway
Infrastructure Amazon EC2 with 10GbE network
Throughput ~8000tps
Latency ~5ms
Scale Linear scale out, to 6 boxes*
13
Performance Characteristics: Mule ESB 3.5 EE
14
Making SLAs a Reality
PrioritizeModelMeasure
15
Tuning: Focusing on What Matters
ObserveIdentify hot spotGenerate loadCompare with SLA
16
Tuning: Improve Scalability
Scale Up, Scale OutSEDAStore and ForwardMessage Oriented Middleware
17
Case Study: Gaming Platform
Public Beta Launch“Code Complete”Players are unforgiving
18
Case Study: Gaming Platform
1. Catalog Services, Estimate Load
2. Prioritize
3. Isolate Mule and Instrument
4. Generate Load
5. Observe
6. Tune
19
Case Study: Gaming Platform
20
Case Study: Gaming Platform
21
Tuning: Improve Reliability
Reliable Acquisition PatternTransactionsRetryDelegate
22
Case Study: Retail
Business Critical Integration“Code Complete”Losing Purchase Orders
23
Case Study: Retail
1. Determine failure modes
2. Decide how to respond
3. Induce and observe
4. Apply reliability patterns
24
Case Study: Retail
25
Questions?Please visit Confluex and
MuleSoft experts in the Expo Hall
26