Scalability of a Node.js based microservice architecture on Heroku
Tapaninen, Juuso (2018)
2018
Tietotekniikka
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2018-06-06
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201805241840
https://urn.fi/URN:NBN:fi:tty-201805241840
Tiivistelmä
Microservices are a method for creating distributed services. Instead of monolithic applications, where all of the functionality runs inside the same process, every microservice specializes in a specific task. This allows for more fine-grained scaling and utilization of the individual services, while also making the microservices easier to reason about.
Push notifications can cause unexpectedly high loads for services, especially when they are being sent to all users. With enough users, this load can become overwhelming. Most services have three options to meet this increased demand: scale the service either horizontally or vertically, improve the performance of the service or send the notifications in batches. In our service, we chose to implement the batched sending of notifications. This caused issues in the amount of time it took to send all notifications. Instead of a short peak in traffic, the service had to manage consistently high load for a long period of time.
This thesis is part literary study, where we research microservices in more detail and go through the more common architectural patterns associated with them. We explore a production service that had issues with meeting the demand during high load caused by push notifications. To understand the production environment and its restrictions, we also explain the runtime, Node.js, and the cloud provider, Heroku, that were used. We go through the clustering implementation details that allowed our API gateway to scale vertically more effectively.
Based on our performance evaluation of an example Node.js application and our production environment, clustering is an easy and effective way to enable vertical scaling for Node.js applications. However, even with better hardware, there still exists a breaking point where the service can not manage any more traffic and autoscaling is not good enough to meet the demand. A service requires constant monitoring and performance improvements from the development team to be able to meet high demand.
Push notifications can cause unexpectedly high loads for services, especially when they are being sent to all users. With enough users, this load can become overwhelming. Most services have three options to meet this increased demand: scale the service either horizontally or vertically, improve the performance of the service or send the notifications in batches. In our service, we chose to implement the batched sending of notifications. This caused issues in the amount of time it took to send all notifications. Instead of a short peak in traffic, the service had to manage consistently high load for a long period of time.
This thesis is part literary study, where we research microservices in more detail and go through the more common architectural patterns associated with them. We explore a production service that had issues with meeting the demand during high load caused by push notifications. To understand the production environment and its restrictions, we also explain the runtime, Node.js, and the cloud provider, Heroku, that were used. We go through the clustering implementation details that allowed our API gateway to scale vertically more effectively.
Based on our performance evaluation of an example Node.js application and our production environment, clustering is an easy and effective way to enable vertical scaling for Node.js applications. However, even with better hardware, there still exists a breaking point where the service can not manage any more traffic and autoscaling is not good enough to meet the demand. A service requires constant monitoring and performance improvements from the development team to be able to meet high demand.