aside from the channel capacity, geostationary satellite links with two-way data service are limited by the latency, which is a minimum of about 490 ms round trip in a dedicated-bandwidth SCPC (single carrier per channel) link. anything that is highly transactional, unfortunately including for example loading webpages with many images on different servers requires dozens of back-and-forth communications between the http client and the httpd.
another example of how latency kills throughput: let's say you want to retrieve thousands of small files across a high bandwidth but also high latency satellite link. even if you have a 10 Mbps+ link, sequentially downloading a mirror of the debian wheezy apt cache (thousands of files 15KB to 30MB in size) would take a very long time because of the need to request each file,
retrieve each file and then ask for the next one with a minimum 500ms lag in between each small file. multi-stream http downloading and parallel download tools help with this, but a lot of software that exists today was decidedly not optimized for 500ms+ latency between client and server.
the 500ms+ latency issue is one of the reasons why the upcoming o3b network, a MEO Ka-band based system, may improve net access in developing nations that have a lack of terrestrial/marine fiber infrastructure.
Comments
Post a Comment