Transcript Block-switched Networks: A New Paradigm for Wireless Transport

Block-switched Networks: A New
Paradigm for Wireless Transport
Offense
Alok Rakkhit and Patrick Wong
Hop and the Link Layer

Designed for wireless link-layer protocols
specifically

How will it work over other LL protocols?
Replace Hop with TCP/UDP under wired protocols?
 If LL bursts behave differently, will there still be
performance gains?

Hop and the Network Stack

Breaks hierarchy by skipping over network layer
and directly interacting with link layer




How will issues at IP layer affect Hop
Never explicitly stated
At every hop a packet packets have to be processed
down to the transport layer instead of the link layer
Implemented over UDP, making it a second
Transport Layer?

Are the comparisons representative?
Caching


Makes heavy use of caching at each router
When there is lots of traffic won’t there be
issues with the caches filling up?
Performance gains

Not clear whether the performance is due
to Hop itself or the underlying routing
protocol

“In conjunction with a disruption-tolerant
routing protocol…” it does well
Experiment



Why 802.11b? Why not conduct all the
experiments on 802.11g instead of shoving that
as a sub-experiment?
Table 1: Median and mean improvements are so
different. Some of the means are just 1x = no
improvement!
Multiple researches on TCP modifications for
multi-hop networks.


Would have been nice to compare those instead
Makes a stronger argument that Hop is
*fundamentally* better
Conclusion


Not easily deployable
Paper does not conclusively prove Hop is
fundamentally better than TCP