Scherb, Christopher Max Constantin. Resolution strategies for serverless computing in information centric networking. 2020, Doctoral Thesis, University of Basel, Faculty of Science.
|
PDF
Available under License CC BY-NC-ND (Attribution-NonCommercial-NoDerivatives). 46Mb |
Official URL: http://edoc.unibas.ch/diss/DissB_13664
Downloads: Statistics Overview
Abstract
Named Function Networking (NFN) offers to compute and deliver results of computations in the context of Information Centric Networking (ICN). While ICN offers data delivery without specifying the location where these data are stored, NFN offers the production of results without specifying where the actual computation is executed. In NFN, computation workflows are encoded in (ICN style) Interest Messages using the lambda calculus and based on these workflows, the network will distribute computations and find execution locations. Depending on the use case of the actual network, the decision where to execute a compuation can be different: A resolution strategy running on each node decides if a computation should be forwarded, split into subcomputations or executed locally.
This work focuses on the design of resolution strategies for selected scenarios and the online derivation of "execution plans" based on network status and history. Starting with a simple resolution strategy suitable for data centers, we focus on improving load distribution within the data center or even between multiple data centers. We have designed resolution strategies that consider the size of input data and the load on nodes, leading to priced execution plans from which one can select the ones with the least costs.
Moreover, we use these plans to create execution templates: Templates can be used to create a resolution strategy by simulating the execution using the planning system, tailored to the specific use case at hand.
Finally we designed a resolution strategy for edge computing which is able to handle mobile scenarios typical for vehicular networking. This “mobile edge computing resolution strategy” handles the problem of frequent handovers to a sequence of road-side units without creating additional overhead for the non-mobile use case.
All these resolution strategies were evaluated using a simulation system and were compared to the state of the art behavior of data center execution environments and/or cloud configurations. In the case of the vehicular networking strategy, we enhanced existing road-side units and implemented our NFN-based system and plan derivation such that we were able to run and validate our solution in real world tests for mobile edge computing.
This work focuses on the design of resolution strategies for selected scenarios and the online derivation of "execution plans" based on network status and history. Starting with a simple resolution strategy suitable for data centers, we focus on improving load distribution within the data center or even between multiple data centers. We have designed resolution strategies that consider the size of input data and the load on nodes, leading to priced execution plans from which one can select the ones with the least costs.
Moreover, we use these plans to create execution templates: Templates can be used to create a resolution strategy by simulating the execution using the planning system, tailored to the specific use case at hand.
Finally we designed a resolution strategy for edge computing which is able to handle mobile scenarios typical for vehicular networking. This “mobile edge computing resolution strategy” handles the problem of frequent handovers to a sequence of road-side units without creating additional overhead for the non-mobile use case.
All these resolution strategies were evaluated using a simulation system and were compared to the state of the art behavior of data center execution environments and/or cloud configurations. In the case of the vehicular networking strategy, we enhanced existing road-side units and implemented our NFN-based system and plan derivation such that we were able to run and validate our solution in real world tests for mobile edge computing.
Advisors: | Tschudin, Christian and Ott, Jörg |
---|---|
Faculties and Departments: | 05 Faculty of Science > Departement Mathematik und Informatik > Informatik > Computer Networks (Tschudin) |
Item Type: | Thesis |
Thesis Subtype: | Doctoral Thesis |
Thesis no: | 13664 |
Thesis status: | Complete |
Number of Pages: | 1 Online-Ressource (xxix, 196 Seiten) |
Language: | English |
Identification Number: |
|
edoc DOI: | |
Last Modified: | 02 Sep 2020 12:30 |
Deposited On: | 02 Sep 2020 12:30 |
Repository Staff Only: item control page