MicroBlend is a framework designed to automatically integrate IaaS and FaaS services, focusing on both cost efficiency and performance optimization.
- Compiler: Transforms code to be FaaS-compatible.
- Controller: Manages autoscaling and allocates services.
- Traces: Utilizes real-world traces (WITS, WIKI) for evaluation purposes.
- Workload: A benchmark tool used for evaluation, written in Python.
- Loadcat: Essential for controlling nginx, which acts as a load balancer. You can access the Loadcat Repository here.
- Install Loadcat: Once installed, initialize several servers and add them to Loadcat.
- Deploy the Workload: Ensure the workload is appropriately distributed across the servers. (Require background on Docker)
- Once the workload is deployed, the one of microservices will initiate Prometheus, which will be responsible for collecting
runq_latencymetrics for each microservice. This data can be subsequently used to feed into a training model, enabling the selection of microservices that are best suited to meet the Service Level Objectives (SLO). During the provisioning phase, a compiler is utilized to transform the microservice into a Lambda function. Additionally, compiler modifies the orchestrator function, transitioning its calls from a VM-based function to Lambda. And controller.py would reroute requests to run hybrid code through Loadcat.
- Run the Controller: Further details can be found in
Controller/controller.py.