Forgot your password?
typodupeerror

Comment: Re:Nasa and Belcore got it right (Score 1) 374

by Brymouse (#37434326) Attached to: Ask Slashdot: Clever Cable Management?

+1 for cable lacing. I'll show a couple pictures of lacing i did during a lab evaulation of some gear a few years ago.

168 DS1's all in a rack

Before:
http://gallery.keekles.org/d/4624-2/94-ds1s.jpg

After:
http://gallery.keekles.org/d/4615-1/8620-laced-out.jpg

Power Cables:
http://gallery.keekles.org/d/4628-1/Power-cables.jpg

The entire setup:
http://gallery.keekles.org/d/4634-1/complete.jpg

Comment: Bandwidth != airtime (Score 2, Insightful) 309

by Brymouse (#33914116) Attached to: Can Apps Really Damage a Cellular Network?

The issue with all Cellular networks (and any half duplex shared media) is that the time it takes to send 256 bytes over the air is not 1/4 the time it takes to send 64 bytes, it's more like .6 to .8 times. The signaling setup and tear down takes time to transmit packets over the air, which is fixed no matter the amount of bytes you send.

This impacts the network as the real bandwidth of a cellular network is not in BPS but airtime. If all the airtime is used up for signaling small packets for marginal signal customers, even the customers that have strong signals and want to send a http request will have to wait. Stateless protocols cause the worst problems as once a flow is established the PDSN/HA/etc does not have to do anymore work. With a app that generates a new flow for each data transfer of 10 bytes to say "hey im still online", the signaling bandwidth is used up and the network quickly falls to it's knees.

This massive use of third party apps and data is still quite new to the providers. This scares them, as you can't just turn on netflow, setup nfsen and see what's going on. Lucent is about the only company out there with a ntop like solution for the providers, but it's new and still being deployed.

I know the IP people are asking how they don't know what's on their network, but it's not just IP traffic you need to monitor, as all the carriers do so. monitoring the IP traffic only gives you the 10000 foot level view, to actually say how the loading on the radio layer relates to the applications in use is a very new requirement. While you can pull hundreds of data point for voice traffic from each radio and switch, at best you can find an error rate and total transfer for the busy hour on the data counters.

It's the providers problem for selling a data plan based on bytes transferred , rather than airtime used.

Is your job running? You'd better go catch it!

Working...