Group Actions
Engagement
  • Author Author: Catwell
  • Date Created: Date Created
  • Views 573 views
  • Likes 1 like
  • Comments 2 comments
Related
Recommended

MIT uses math to gain a ‘bandwidth breakthrough’ over net congestion

Catwell

image

 

It’s happened to all of us at one time or another. You’ll be sitting in traffic or walking around town talking when all of a sudden the connection is lost and web-pages take forever to load. You remind yourself that you did purchase your mobile device with an ultra-fast 4G or Wi-Fi connection, but still to no avail it feels like your back in the 90s with a dial-up connection. The loss of connection or slow-down is usually the result of ‘net congestion’ or rather ‘packet-loss,’ and until now there was no easy fix for it other than installing more base stations or upping the towers wattage. Thanks to the brains at MIT this issue was solved using math, specifically an algebraic equation rather than more hardware.

 

When too many people are using a network at the same time it becomes congested and results in packet loss. To overcome the problem, the network resends the packet to the mobile device in order to reacquire the missing data. MIT’s solution sends a linear algebraic equation that ‘describes’ the lost packet and the device being used reconstructs that missing information on its own. This process actually shifts the processing load to the device rather than the tower or base station, which not only alleviates the net congestion but improves the overall connection speed.

 

The MIT researchers tested their ‘math’ using MIT’s own Wi-Fi networks with a typical 2% data packet loss. The team found that using the equation resulted in an overall increase in speed of 1 to 16Mbps! They also found that in area’s that suffered a 5% loss resulted in a speed increase of 0.5 to 13.5Mbps after applying the algebraic system. According to MIT, there have been several companies that have already licensed their system but have signed non-disclosure agreements so exactly who these companies are will not be known for quite a while (Google, Apple, we know it's you). One thing is for certain, while the technology will help alleviate current net congestion, it may or may not forestall the ‘spectrum crunch’ (available wireless frequencies) Cisco Systems, Bell Labs and the FCC expect anywhere from a few years to 2016.

 

It's elegance over brute force. We should all learn from their example.

 

Cabe

http://twitter.com/Cabe_e14

  • This is a favorite topic of mine.  The way you describe it sounds like forward error correction. 

     

    I've seen this problem on Wi-Fi.  Once you have a few collisions, the contention window increases, i.e. it waits longer before sending the next packet.  Some algorithms fall back to a lower data rate when there are lots of errors, causing each packet to tie up the channel for longer and be at more risk of interference.  This is especially a problem in industrial environments where there are lots of devices sending very short packets at a high packet rate. 

  • It was only a matter of time before they needed to build out a coordination scheme for the cell coordination.

    The initial cell design was good for limited use, but with the explosive growth in phones and data usage, more command and control is needed.

    There is plenty of capacity within the existing structure if they begin using apriori knowledge to set up reservations between the cell blocks to keep mobile users connected.

    It is not rocket science, but it does require a coordinated network to keep everything allocated.

    One caveate though, if they do begin reserving connections, you might find you cannot get a line during peak times.  Still, that would be better than letting calls be dropped after they connected a previous user.

     

    It will be interesting to see how much they can actually expand usage with the new algorithms.

     

    DAB