Sign up with GoogleMy thoughts and comments on iBurst
From some testing, research and generally bashing around I'd like to venture the following concerning iBurst...
Background (basics):
The iBurst system makes use of multiple towers with multiple antennas on each tower. The signalling system employed allows the tower or the modem to decide on a different path for packets to flow. This is based on numerous factors such as tower load, distance signal has travelled and quality of signal.
The iBurst system also supports semi-transparent handover between towers (base stations), similar to the way cell phones work (except that there is a momentary halt of traffic).
All in all, a well thought out and effective system
Why are many complaining about the speed?
There are numerous factors here…
Firstly the ones that the helpdesk are quick to mention:
Upstream Providers
From testing done by me and many others on the forum it is quite obvious that international has deteriorated over the past weeks. A large part of the problem lies with the IS & UUNet, both have indicated problems on their Network Notices with their international peering partners.
I have specifically not mentioned lack of international bandwidth as problem as I don’t believe iBurst have yet saturated their capacity.
Usage Patterns
While the service has been touted as broadband service, iBurst has themselves geared the service for “corporate” use:
Australia initially also punted the “broadband” aspect of the service… Of late however the packages offered by many of those providers have now also changed to a more corporate road-warrior style service.
What does all this mean? Simple, the service is not suited to continuous HIGH bandwidth solutions like p2p. This is due to the way the signalling works… Those with good signals close to the tower will invariably be able to get better service than those with poorer signals… If those users are thrashing the tower with continuous downloads the aggregate available bandwidth (Kyocera indicates 20Mbs per base station) for other users is reduced.
Those users at the edge of the coverage areas will know the pain I explained above, between 8:15am and 23:30 service is often almost unusable with numerous disconnects and general slow throughput.
Even the product name implies the usage it’s suited to … iBurst … bursts of traffic …
Tower Contention
This is what most of us suspect is the root of the slow general throughput experienced by some users … This may however not be completely accurate… While many users do contend for the services of one tower some users are happy, others less so. The reason behind this goes back to the way the iBurst network handles signalling… Those further (poorer return signal) use a different modulation than those closer to the tower. This means that users at the edge get fewer opportunities to actually send/receive data.
Tower Hunting
Many have seen the “Christmas Tree” effect on our modems … 5lights … 0lights …5lights. This is due to the fact that while our modems receive a strong signal from the tower it often struggles to get a signal back to the tower. When the modem detects that data isn’t getting back to the tower it actively tries to hand over to another tower…often with the same return path problem … The modem must then switch back to the “poor” tower in order to get any connectivity.
Conclusions
Those of us who made it through the MyWireless experience will read the above and grin … this is exactly what happened to Sentech’s network… Too many users per tower; Modems handing off to out-of-reach towers…
To fix the problems requires more towers – Until more towers go live things are just going to get worse and worse as the number of users on the iBurst network grows.
What is needed is a public admission by WBS’s executive that there ARE problems. They need to provide their users with a roadmap on how they are going to fix things … Otherwise they are headed for a Sentech-like roller-coaster.
As much as I hate saying it, Telkom’s ADSL service is currently the only true broadband in this county !
Any comments, corrections or feedback ?
From some testing, research and generally bashing around I'd like to venture the following concerning iBurst...
Background (basics):
The iBurst system makes use of multiple towers with multiple antennas on each tower. The signalling system employed allows the tower or the modem to decide on a different path for packets to flow. This is based on numerous factors such as tower load, distance signal has travelled and quality of signal.
The iBurst system also supports semi-transparent handover between towers (base stations), similar to the way cell phones work (except that there is a momentary halt of traffic).
All in all, a well thought out and effective system
Why are many complaining about the speed?
There are numerous factors here…
Firstly the ones that the helpdesk are quick to mention:
Upstream Providers
From testing done by me and many others on the forum it is quite obvious that international has deteriorated over the past weeks. A large part of the problem lies with the IS & UUNet, both have indicated problems on their Network Notices with their international peering partners.
I have specifically not mentioned lack of international bandwidth as problem as I don’t believe iBurst have yet saturated their capacity.
Usage Patterns
While the service has been touted as broadband service, iBurst has themselves geared the service for “corporate” use:
Australia initially also punted the “broadband” aspect of the service… Of late however the packages offered by many of those providers have now also changed to a more corporate road-warrior style service.
What does all this mean? Simple, the service is not suited to continuous HIGH bandwidth solutions like p2p. This is due to the way the signalling works… Those with good signals close to the tower will invariably be able to get better service than those with poorer signals… If those users are thrashing the tower with continuous downloads the aggregate available bandwidth (Kyocera indicates 20Mbs per base station) for other users is reduced.
Those users at the edge of the coverage areas will know the pain I explained above, between 8:15am and 23:30 service is often almost unusable with numerous disconnects and general slow throughput.
Even the product name implies the usage it’s suited to … iBurst … bursts of traffic …
Tower Contention
This is what most of us suspect is the root of the slow general throughput experienced by some users … This may however not be completely accurate… While many users do contend for the services of one tower some users are happy, others less so. The reason behind this goes back to the way the iBurst network handles signalling… Those further (poorer return signal) use a different modulation than those closer to the tower. This means that users at the edge get fewer opportunities to actually send/receive data.
Tower Hunting
Many have seen the “Christmas Tree” effect on our modems … 5lights … 0lights …5lights. This is due to the fact that while our modems receive a strong signal from the tower it often struggles to get a signal back to the tower. When the modem detects that data isn’t getting back to the tower it actively tries to hand over to another tower…often with the same return path problem … The modem must then switch back to the “poor” tower in order to get any connectivity.
Conclusions
Those of us who made it through the MyWireless experience will read the above and grin … this is exactly what happened to Sentech’s network… Too many users per tower; Modems handing off to out-of-reach towers…
To fix the problems requires more towers – Until more towers go live things are just going to get worse and worse as the number of users on the iBurst network grows.
What is needed is a public admission by WBS’s executive that there ARE problems. They need to provide their users with a roadmap on how they are going to fix things … Otherwise they are headed for a Sentech-like roller-coaster.
As much as I hate saying it, Telkom’s ADSL service is currently the only true broadband in this county !
Any comments, corrections or feedback ?
.
