Microservices: Another Way
•Download as PPTX, PDF•
0 likes•583 views
What Microsoft Azure Service Fabric give to Java developers to build applications with Micro-services architecture more easily
Report
Share
Microservices: Another Way
- 2. INTRODUCTION 13 Distributed & Highload Cloud Solutions WorkExperience Languages Focus
- 3. AGENDA Is Monolith a real problem? Why Micro-Services? How we gonna do Micro-Services? Why Service Fabric? What is Another Way?
- 4. IS MONOLITH A REAL PROBLEM?
- 5. MONOLITH Azure AWS Desktop Browsers 80% backend logic 20% backend logic Monolith
- 6. MONOLITH - SCALABILITY 80% backend logic Azure Redis: Cache Redis: WebSockets Backplane SQL Store Monolith Instance 1 Monolith Instance 2 Monolith Instance 3 Monolith Instance N Stateless & Scalable Redis SQL – limited scale – 10% load
- 7. PROBLEM - TECHNICAL 1 Language / Platform limitation Module cannot be scaled separately DB Schema sharing
- 8. PROBLEM - BUSINESS Acquiring several companies Expecting growth ~ x5 in a year Targeting for Hybrid solution
- 10. SCALING & HARDWARE ~ 20 Modules 3 – High Load 1 – requires Memory GPU Module 1 GPU Module 2 SSD Module 1 SSD Module 2 Node with SSDNode with GPU
- 11. LANGUAGE & TOOLS Use appropriate language & tools for component goals Languages: Java .NET NodeJS Ruby C++ Python Storages & tools: SqlServer MySql Redis ELK Kafka Apache Solr Apache Spark
- 12. HYBRID (CLOUD + ON PREMISES) Cloud Service 1 Service 2 Data Service 1 Data Service 2 Service 3 Regular Tenants Classified Tenants Tenant DC
- 13. HOW WE GONNA DO MICRO-SERVICES?
- 15. REQUIREMENTS - DEVOPS Script minimization Upgrade in single command High Availability & Self-Healing Service Discovery Hardware Constraints
- 16. DEPLOYMENT – SERVICE PER VM
- 19. WHY WE CHOSE SERVICE FABRIC?
- 20. WHAT IS Service Orchestrating Tool
- 21. SERVICE FABRIC vs CONTAINERS ORCHESTRATION Service Fabric Containers Orchestrating Services Containers Ops Cost Low Medium App Infrastructure One SDK Custom Scaling Instance + Partitioning Instance Service Isolation No, Container, Kernel Container
- 22. Service Fabric Cluster Node 1 Node 5 Node 3Node 4 Node 2Resource Manager Cluster Manager Naming Service Image Store Health Manager CLUSTER
- 23. Service Fabric Cluster Node 1 Node 5 Node 3Node 4 Node 2 Manage Admin UI PS / CLI Application Web Service Worker Service Deploy Web Service Web Service Web Service Worker Service Worker Service Worker Service Worker Service Worker Service MANAGEMENT
- 24. PLATFORMS
- 25. SDK
- 26. PROGRAMMING MODELS Reliable Services Stateful Stateless Actors SDK No SDK Containers Guest Executable
- 27. RELIABLE SERVICES Cluster State Stateless Stateful State
- 30. GUEST EXECUTABLE
- 31. WHAT IS ANOTHER WAY?
- 32. ORCHESTRATING / SCALING / SELF-HEALING Usually Another Way Resource Manager Failover Manager Health Manager
- 33. SERVICE DISCOVERY Usually Another Way Eureka Naming Service
- 34. CONFIGURATION Usually Another Way Image Store
- 35. LOAD BALANCING Ribbon Usually Another Way Naming Service Reverse Proxy
- 36. STATE Usually Another Way Reliable Collections
- 37. HEALTH / MONITORING Eureka Usually Another Way Health Manger
- 38. TESTABILITY Usually Another Way Fault Analysis Service
- 40. DEMO
- 41. SUMMARY
- 42. WHEN TO USE
- 43. WHEN TO USE
- 44. WHEN TO USE
- 45. WHEN TO USE
- 46. WHEN TO USE
- 47. WHEN TO USE You want to try Another Way
- 48. LIMITATIONS Hyper-V Virtualization – Q3 201 Full Linux support – Q4 2017 Stateful services on Linux – Q4 2017 Reverse Proxy on Linux – Q4 2017 Docker Compose support – Q4 2017 Heterogeneous Cluster – Q1 2018
Editor's Notes
- Hybrid Application – Cloud + OnPrem Application must be hosted on Azure Lots of Programming Languages Minimize SysOps / DevOps workload Simplify migration from Monolith Heterogeneous Cluster (Future) – Windows + Linux