New Design Patterns in Microservice Solutions
Download as pptx, pdf0 likes26 views
Mimik Technology's document explores hybrid edge cloud (HEC) architecture, where smart devices function as clients and cloud servers, optimizing microservice management. It details various stages of microservices evolution at the edge, introducing patterns like shadowing, edge caching, prepping, and the role of edge cloud brokers in resource allocation. The conclusion emphasizes the synergy between microservices and edge technologies, advocating for new approaches to overcome traditional cloud limitations.
New Design Patterns in Microservice Solutions
- 1. mimik Technology, Inc. copyright New Design Patterns in Microservice Solutions or Exploring the Influence of Hydrid Edge Cloud on Microservice Architecture
- 2. Michel Burger: Chief Technology Officer of mimik Highlights: • CTO at mimik • Chief Architect for the service providers and IOT at Apigee • Chief Architect & Distinguished Engineer for Vodafone Group R&D Technology • Head of Architecture for Vodafone Internet Services • TM Forum Fellow • CTO of the Communications Sector at Microsoft • CTO at Embrace Networks • Director of innovation at Sapient • Chief Architect PBX division at Nortel Networks • Holds many patents mimik Technology, Inc. copyright 2011, Vodafone distinguished engineer 2015, Distinguished Fellow TM Forum
- 3. mimik Technology Inc. copyright What is Hybrid Edge Cloud?
- 4. What is Hybrid Edge Cloud (HEC)? mimik Technology Inc. copyright 1. When smart devices can act as both clients and cloud servers. 2. When smart devices understand their resources and can expose them to other devices. 3. When smart devices can collaborate with other devices including servers in data centers, edge data centers, or edge servers. 4. Where microservices are hosted as close as possible to the workflow or saga. HEC is an architectural approach where smart devices CAN act as IT resources running microservices
- 5. Cloud data centers x thousands Edge data centers x 100’s thousands On-Premise edge servers x millions Billions of smart devices, 10s of billions of apps/services Most edge cloud approaches today use dedicated hardware for cloud server function. The cloud hierarchy mimik Technology Inc. copyright dedicated cloud server approach Hybrid Edge Cloud (HEC)
- 6. mimik Technology Inc. copyright Microservice architecture evolution
- 7. mimik Technology Inc. copyright Microservices Almost all activities have to go through servers in data centers application μS1 API GATEWAY μS2 μS3 microservices can interact via API (aka cloud native architecture) device service mesh
- 8. mimik Technology Inc. copyright Stage 1: Microservices at the edge application μS1 API GATEWAY μS2 μS3 API GATEWAY μSa μSb μSc Communication with the cloud is done only if necessary device service mesh
- 9. mimik Technology Inc. copyright Stage 2: App to App communication μS11 API GATEWAY μS21 μS31 μS12 API GATEWAY μS22 μS32 application 1 API GATEWAY μSa1 μSb1 μSc1 device application 2 API GATEWAY μSa2 μSb2 μSc2 Minimize the need for back-end integration service mesh service mesh
- 10. mimik Technology Inc. copyright Stage 3: Ad hoc edge service mesh application API GATEWAY μSa μSb μSc device 1 application API GATEWAY μSa μSb μSc device 2 μS1 API GATEWAY μS2 μS3 No need for a cloud round trip for the devices to communicate ad hoc edge service mesh service mesh
- 11. mimik Technology Inc. copyright Stage 4: Service mesh aware network mimik Technology Inc. copyright application API GATEWAY μSa μSb μSc device 1 application API GATEWAY μSa μSb μSc device 2 Service mesh context can be passed to the network for its optimization (5G slice, SON) ad hoc edge service mesh network network element (eg. RAN) Network stack
- 12. mimik Technology Inc. copyright Edge Patterns
- 13. Shadowing Definition Without shadowing An edge microservice implementing the shadow pattern will execute tasks that were performed by a cloud microservice and will send the results of the execution to the cloud. The API implemented by the edge microservice will be a subset of the the API implemented by the cloud microservice. API Consumer Cloud microservices 1:API call 2:API call result With shadowing API Consumer Cloud microservices Edge microservice Device 1:API call 2:API call result 3:Send Results Device - Improve latency - Reduce cloud cost - No impact on the API consumer mimik Technology Inc. copyright
- 14. Edge caching Definition An edge microservice implementing the edge caching pattern will be able to answer API call requests (GET) on behalf of the cloud microservice either by calling the cloud microservice itself or by preempting the call to the cloud microservice (preemptive edge caching) or receiving a call from the cloud microservice (reverse preemptive edge caching), or by calling edge microservices in a specified cluster (network, proximity, account) for the data (cluster edge caching). Basic edge caching Preemptive edge caching Cluster edge caching API Consumer Cloud microservices edge microservice 1:API call (GET) 4(or 2):API call result 3:API call result Device 2:API call (GET) If the data is not is the edge microservice Data is fetch based on the edge microservice logic(schedule based, policy or even AI based) or based on the logic of the cloud microservice API Consumer Cloud microservices edge microservice 3:API call (GET) 4:API call result 2:API call result Device 1:API call (GET) API Consumer Cloud microservices edge microservice 1:API call (GET) 6(or 4 or 2 ):API call result 5:API call result Device 4:API call (GET) edge microservice edgeEngine Device in the same cluster If the data is not in the defined cluster - improve latency - cost reduction - make the cloud work for the edge mimik Technology Inc. copyright
- 15. Prepping Definition An edge microservice implementing the prepping pattern will be able to answer API call requests intended to the cloud microservice even if the connection with the cloud is not established. The microservice will stage the activities and send the information to the cloud microservice when the connection to the internet is reestablished. Without prepping With prepping API Consumer Cloud microservices Device If the communication with the cloud microservice is lost, an app (API consumer) will quickly stop and becomes useless API Consumer Cloud microservices Edge microservice 1:API call 2:API call result 3:Send Activities Device - better user experience while not being connected When the communication with the cloud microservice is lost, the application (API consumer) is still responding to the extent of what the edge microservice is can do without relationship with the cloud microservice When the communication with the cloud microservice is reestablished, the edge microservice send the activities in order to synch the cloud microservice with the edge microservice mimik Technology Inc. copyright
- 16. Edge cloud broker Definition An edge cloud broker provides the ability to select the most appropriate resource (edge of cloud) to fulfill an API request.. RAN edgeCloud Broker API call path configuration request connectivity ***if connected to a global service broker OR resource provider cluster information from edgeEngine information from the network (local/global) basic configuration & policies Step 1: Information gathering Step 2: API call + new information Step 4: Selection App/Data meta data information AND ***if running on top of network element ***if part of edge service mesh ***if running on top of network element Step 3: Processing Operations mimik Technology Inc. copyright
- 17. Management of microservices Orchestration mimik Technology Inc. copyright • A service controller manages communications between microservices and directs their transactions • The service controller communicates directly with each microservice and waits for their response • There is tight coupling between microservices and the controller is a single point of failure • Can experience excessive wait times and service availability and scalability issue Choreograph y • Microservices act independently and interact with loose coupling . • There is no single point of failure and transactions can happen much faster. • It requires a different approach to architecture design to minimize dependencies at the outset. • Faster and more agile software development and more consistent and efficient applications Hybrid • A centralized controller provides visibility to all microservices • Microservices can communicate independently or through the centralized controller (asynchronous or synchronous)
- 18. mimik Technology Inc. copyright Example implementations
- 19. Software Hardware ASIL-D services Containerized services (mixed criticality and type resource aware deployment orchestrator, legacy apps ) edge services (exposing REST and/or streaming API) Native app (recognition software that access GPU directly, UI,.... ) Containers environment (docker, CoreOS, etc. …) edgeEngine (standard, main-child, controller-worker) OS (RTOS, AAOS, Linux, etc. non/direct on top of hypervisor) Network (TSN, non TSN) Generic resources (central processing, generic control unit) Typed resources (display, smart sensor, smart actuator, zonal gateway, controllers) mimik Technology Inc. copyright Adhoc edge service mesh (passenger phone, traffic light, home…) Any add-on device expands the service mesh Zonal architecture
- 20. mReport Field Service App Smart Pump Simulation mIoT Incident Management Smart Pump Smart Pump App Smart Pump microservice mSharedModel mDeploy mIoT Sensor /Actuator Incident Machine Learning Incident Management Field Service Tablet Sensor /Actuator Sensor /Actuator - When a sensor detects a problem, the Smart Pump app (via mIoT and Smart Pump microservice) using an ML model in mSharedModel informs the back-end systems - The back-end systems deploy an Incident Management shadow microservice, a new ML model, a Tech Support Documentation cache microservice and dispatch a Field Agent - The Tech Support Documentation prefetches the relevant support documentation - When the Field Service Tablet arrives in proximity of the Smart Pump, they discover each other and the microservices/app in the tablet communicate with the different microservices in the smart pump to diagnose and fix the problem - Smart Pump Simulation may also receive/send inputs to the sensors Cloud or on-premise Local Sensors By-directional dynamic information exchange between sensor an other device with proper authorization Tech Support Documentation Technical Documentation New Model New Model Tech Support Info SDK SDK SDK mimik Technology Inc. copyright
- 21. mimik Technology Inc. copyright Conclusion
- 22. mimik Technology Inc. copyright • Microservices and edge are working well together • However, it is not just about moving components to the edge • Conventional Cloud orchestration limitations make room for versatile technologies • After mobile internet we are entering in a new internet: the cognitive internet • Try mimik edge engine (https://developer.mimik.com) • and also read IEEE article about hybrid edge cloud (https://mimik.com/hybrid-edge-cloud-a-pragmatic-approach-for- decentralized-cloud-computing)
Editor's Notes
- #6: .