Revisiting MARS: Storing Coded Packets In-Advance for IPFS
Research output: Contribution to book/Conference proceedings/Anthology/Report › Conference contribution › Contributed › peer-review
Contributors
Abstract
The Interplanetary File System (IPFS) is one of the largest peer-to-peer networks currently in use. In networks with heterogeneous peers, IPFS may take a long time to obtain all packets. Random Linear Network Coding (RLNC) can be added to IPFS to mitigate the effects of the suboptimal requesting strategy, as proposed by the Multi Access Recording System (MARS) protocol. We measure the power consumption, and the file processing and receiving delay for MARS in a test bed and find that MARS introduces significant delay overhead because it encodes its packets only after receiving a request for a file. As a modification of MARS, we propose Storing Coded Packets In-advance for IPFS (SPIFI) that addresses this issue by offloading the computation of the RLNC encoded packets to a point in time before the packets are requested. This way, the power consumption, and the delay incurred by the RLNC encoding does not deteriorate performance at run time. In additional measurements, we find that SPIFI reduces the energy consumption at run time and the data retrieval time compared to both MARS and IPFS. The encoding time can be selected freely, to perform the computation when power grids have spare capacity, avoiding peak loads in the grid. SPIFI works best for small files of a few MB and the benefit vanishes for larger file sizes due to the decoding complexity. If the client only requests single code packets to increase the file redundancy in the network, then SPIFI always outperforms IPFS and MARS.
Details
Original language | English |
---|---|
Title of host publication | ICC 2024 - IEEE International Conference on Communications |
Pages | 4692-4697 |
Number of pages | 6 |
ISBN (electronic) | 978-1-7281-9054-9 |
Publication status | Published - 2024 |
Peer-reviewed | Yes |
External IDs
ORCID | /0000-0001-5980-8731/work/166764105 |
---|---|
ORCID | /0000-0001-8469-9573/work/166764225 |
Scopus | 85202903239 |