dc.contributor.author |
Tatarski, Michael |
|
dc.contributor.author |
Parzyjegla, Helge |
|
dc.contributor.author |
Danielis, Peter |
|
dc.contributor.author |
Mühl, Gero |
|
dc.date.accessioned |
2022-04-08T07:49:38Z |
|
dc.date.available |
2022-04-08T07:49:38Z |
|
dc.date.issued |
2022-04-07 |
|
dc.identifier.uri |
http://hdl.handle.net/10900/126086 |
|
dc.identifier.uri |
http://nbn-resolving.de/urn:nbn:de:bsz:21-dspace-1260865 |
de_DE |
dc.identifier.uri |
http://dx.doi.org/10.15496/publikation-67449 |
|
dc.description.abstract |
In recent years, the requirements for publish/subscribe systems have changed enormously. Mainly, this is due to the vastly increasing number of smart objects, services, and apps being connected to the cloud and their growing demand for bandwidth and realtime message processing. Unfortunately, many publish/subscribe systems are hardly able to handle both necessary and available data rates when being built on conventional network stacks that suffer from significant system overhead.
In this paper, we present a novel architecture for fast publish/subscribe that consists of an edge broker for preprocessing notifications and a cloud broker that leverages eBPF to efficiently process network packages and speed up notification delivery. We address limitations of eBPF, discuss different filter options, and quantify the achievable performance. The evaluation confirms a significant performance improvement of the eBPF-enabled cloud broker when compared to a conventional implementation. |
en |
dc.language.iso |
en |
de_DE |
dc.publisher |
Universität Tübingen |
de_DE |
dc.subject.ddc |
004 |
de_DE |
dc.title |
Fast Publish/Subscribe Using Linux eBPF |
en |
dc.type |
ConferencePaper |
de_DE |
utue.publikation.fachbereich |
Informatik |
de_DE |
utue.publikation.fakultaet |
7 Mathematisch-Naturwissenschaftliche Fakultät |
de_DE |
utue.opus.portal |
kuvs-nets3 |
de_DE |
utue.publikation.noppn |
yes |
de_DE |