Seubring, Wesley (2021) Data Locality Aware Scheduling on a Serverless Edge Platform. Master's Thesis / Essay, Computing Science.
|
Text
WMCS901-30_2021_Wesley_Seubring.pdf Download (4MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (124kB) |
Abstract
Serverless edge computing is a computing architecture that combines on-demand execution and scaling from serverless with the heterogeneous devices from an edge network, extending the platform outside the cloud. Serverless can be a general approach to efficiently utilise the available resources on an edge network without the complexity of orchestration. Serverless platforms abstract away from orchestration, providing (limited to) no control on the execution location of functions. This limitation brings concerns on privacy and bandwidth usage on a serverless edge system. This work provides an overview of the serverless landscape and proposes a data locality aware scheduler for serverless edge platforms that restricts function execution to nodes within the network of the data. To validate the proposed scheduler and show the performance impact compared to the native Kubernetes scheduler, the scheduler is simulated in a data-intensive application. The results show that moving the execution close to the data reduces bandwidth (cost) and time spent on data movement. In the data-intensive application, the system can provide similar throughput and utilise fewer resources. The scheduler limits the number of resources suitable for scheduling significantly, the limitation causes increased standard deviation of latency over the functions.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Lazovik, A. and Hadadian Nejad Yousefi, M. |
| Degree programme: | Computing Science |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 24 Sep 2021 08:37 |
| Last Modified: | 24 Sep 2021 08:37 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/26126 |
Actions (login required)
 |
View Item |