Frameworks such as Apache Storm, Flink, Heron and Spark were developed to run on clusters or cloud. These such kinds of infrastructures do not have memory, CPU and bandwidth limitations. In contrast, computing resources at the network edge are constrained regarding their capabilities. I am aware of the Apache Edgent and Nifi frameworks. However, they were conceived to run locally on a single computing resource. If you want to run them in a distributed infrastructure, you might create your own stack of components (broker + framework).
https://www.ciena.com/insights/articles/How-the-explosion-of-gaming-and-esports-is-driving-the-need-for-network-evolution.html?utm_campaign=X863118&utm_source=Google&utm_term=&utm_medium=Text&lang=EN&si=AdaptiveNetwork
Himadri Nath Saha, it seems to be a fascinating motivation scenario. The investments and the continuous growth of streaming games have imposed new requirements of response time (i.e., ultra-low latency). I am looking for DSP systems where latency-sensitive applications can be easily deployed on geo-distributed constrained resources avoiding creating a multi-tier of components.
- Badges
- Science topic