LUMS Research On Boosting Bandwidth Featured In MIT Technology Magazine

[Babar]

September edition of the MIT Technology Review has featured the work
of Umar Saif, a professor at LUMS, on improving Internet connectivity
in the developing-world. The Project, dubbed DonateBandwidth, is a
follow-up project of Poor Man’s Broadband work which I wrote about
previously. Our congratulations to Dr. Saif and the team.

Umar shared his thoguhts in an e-mail:

With DonateBandwidth, users in the developing-world can help each
other by donating their unused bandwidth to those who need it. This
project received funding from the US State Department/NAS and HEC and
will be further developed in collaboration with UC Berkeley.

More At:
http://telecompk.net/2008/08/20/lums-research-on-boosting-bandwidth-featured-in-mit-technology-magazine/

Babar

[Rehan Allah Wala]

can we not do the same for wifi ?

http://blogs.rehan.com/2008/07/p2pi-peer-to-peer-internet-project.html

I was told today that Sir Syed University has included my request of the p2pi in one of there
batch for the 3rd year as a project.

I hope they can get in touch with Umer and get this going.

Rehan Ahmed AllahWala
Msn/Yahoo/GoogleTalk/Email: Re...@Rehan.com
http://www.supertec.com/ - Internet Telephony Solutions
Http://www.DIDX.net - DID Number Market Place.
Don't Remember Me ? Visit http://www.Rehan.com

~~~~~~~~~~~~~~~~~~~
"First they ignore you, then they laugh at you, then they fight you, then you win."
By Gandhi.

"Live as if you were to die tomorrow. Learn as if you were to live forever." - Gandhi

[Rehan Allah Wala]

well thank u for trying

i will lookup the projects u refered here, maybe they will help me

> Thanks for the note rehan.
> We are planning something vaguely similar and might even use the MIT
> roofnet nodes. Though we are focused on caching and multiplexing of
> bandwidth rather than just access to Internet (which several people
> have done before, including roofnet and TIER WiLD links).
>
> While I am always happy to bounce off ideas, my capacity to
> collaborate is severely limited these days due to many constraints on
> time.
>
> Thanks.
>
> Cheers,
> --
> Umar Saif | http://www.mit.edu/~umar

[Rehan Allah Wala]

Ok I read both,

One has turned into Meraki, Which i LOVE and use it in Baytaar.com

However the other one, I will read it when my eyes work well enough :) Later.


Now my p2pi Idea is to do what lums guys did on sharing the internet.

I want to use the internet which is in MY PERSONAL Computer, and Turn my computer in to
part of a MESH and use the radio in my PC to re broadcast the signal to the guy in next
room/apartment / house

I want to use this software to share the internet that I have, Specially if my pc is NOT using
the internet at the time .

That was the idea, any thoughts on this ?

Rehan

hanks for the note rehan.
> We are planning something vaguely similar and might even use the MIT
> roofnet nodes. Though we are focused on caching and multiplexing of
> bandwidth rather than just access to Internet (which several people
> have done before, including roofnet and TIER WiLD links).
>
> While I am always happy to bounce off ideas, my capacity to
> collaborate is severely limited these days due to many constraints on
> time.
>
> Thanks.
>
> Cheers,
> --
> Umar Saif | http://www.mit.edu/~umar
>
> Quoting Rehan Allah Wala <re...@supertec.com>:
>

[Mahmood Ahmed]

Can anyone please shed light on the status of USSD Gateway availability in
Pakistan with operators? I am interested to know about USSD V2.0 (with
interactive menu sessions). Telogic has provided to one operator in 2006,
not sure which one?

http://www.telogic.com.sg/web/New-ussd.html

Thanks,

[Tee Emm]

Paktel is using one for quite some time now.

[Zaeem Arshad]

Rehan,

Now my p2pi Idea is to do what lums guys did on sharing the internet.

I want to use the internet which is in MY PERSONAL Computer, and Turn my computer in to
part of a MESH and  use the radio in my PC to re broadcast the signal to the guy in next
room/apartment / house

Pardon my ignorance but isn't that the same as having a LAN (wifi or wired) and setting up ICS in Windows XP or installing a similar internet sharing software with bandwidth throttling. The static gateway assignment or DHCP can solve the issue of default gateway and nameserver assignment to clients. Or is it something entirely different that you are suggesting and I am failing to grasp it?


Cheers

--
Zaeem

[Sameer Bokhari]

My congrats to Dr. Saif and his team

I have gone through the first 3 pages and I am pasting below some lines from the details available on the project on the web, along with my novice questions/commments, that in my opinion are noticable.

 

I would appreciate a comment from experienced members in the ISP/telco field.

 

more than 40 countries have less than 10Mbps of international Internet bandwidth, whereas in Belgium, a 9Mbps ADSL high-speed Internet package is available for just USD 80 a month [1]. "The Digital Divide at a Glance", World Summit on the Information Society, Tunis 2005.

 

Dated source?

 

A 56Kbps modem can typically achieve an average throughput of close to 40Kbps,more than twice the bandwidth available over a typical (10-20Kbps) dialup Internet connection in Pakistan.

Is it true?. On a good telephone line isn't the download speed attained on a typical connection close to 36kbps? if not 40kbps . Doesn't this speed vary depending on conditions.

 

 

That's the case even when a Pakistani user is browsing websites hosted in his or her own country. "The packets can get routed all the way through New York and then back to Pakistan," Saif says.

 

  It was there when only PIE and flag were around which i guess was later fixed. Not sure about the current scenario. Need a comment on this.

 

the scarcity of available bandwidth in the developing world is not a "last-mile" problem

 

Is it true? What is a typical percentage of calls received by an ISPs helpline based on last mile or other issues?

In our experiments, we routinely achieved a peer-to-peer dialup bandwidth of 32kb/sec (symmetric), but the bandwidth allocated by various dialup ISPs rarely exceeded 15kb/sec (typically between 8-15kb/sec)

If the entire problem is due to "limited" upstream bandwidth available to the ISPs and if this 15kb/sec mentioned is upstream bandwidth available to the end user. Whats the point in comparing a p2p dialup bandwidth with that availabe to end user on the upstream link?  what about the p2p bandwidth available to 2 users connected to the same ISP? and if this is not a last mile issue at all then the same can be achieved on a p2p link for 2 users connected to the same ISP. Even the "routine" speed mentioned is 32kbps but what about the max min and average?

 

Last but not least the research is based on the theme Poor Man's Broadband. What would be the cost of data/internet on a traditional dialup as compared to this soluton?

 

SB

 

[Salman Ansari]

I tend to agree with Sameer on the issues he has highlighted from the perspective in Pakistan. The situation on the ground is considerably different on many counts from that assumed in this article. The analogy of using computing cycles to ‘sharing’ bandwidth is not valid: one uses spare capacity where it can be effectively used without any external network constraints and the results (which are small files) do not interfere with the scarce resource of the ‘last mile’ capacity.

 

Most medium to large operators use caching and the small ISPs ‘peer’ into them in some manner. The number of users in the totally ‘uncached’ territory are very few.

 

Most of the ‘speed’ of up and downlink of asynchronous links is needed by the users to browse themselves. This sharing of bandwidth by distributed caching has to be experimented over a large study sample to be accepted as being valid.

 

But am open minded to be excited enough to see if this theory or a modified version can be put to practical use in a real environment.

 

Salman

<br

[Faried Nawaz]

On Aug 21, 5:26 am, "Rehan Allah Wala" <re...@supertec.com> wrote:

> I want to use the internet which is in MY PERSONAL Computer, and Turn my computer in to
> part of a MESH and use the radio in my PC to re broadcast the signal to the guy in next
> room/apartment / house
>
> I want to use this software to share the internet that I have, Specially if my pc is NOT using
> the internet at the time .

http://en.wikipedia.org/wiki/Netsukuku

[Qasim Khan]

Fully agreed with Salman Ansari, real environment modification and acceptability shall be necessary. It would also help if a little more explanation and segmentation is worked on!

 

Qasim

[Fawad Niazi]

Telogic provided it to Paktel and was also trying to get order from others esp Telenor

[umar]

Babar pointed me to this thread on our recent research project. I
typically do not post to forums, but since Salman had made a few
comments, I wanted to clarify the assumptions behind our work.

(1) Re: comment: "The analogy of using computing cycles to 'sharing'

bandwidth is not valid: one uses spare capacity where it can be
effectively used without any external network constraints and the
results (which are small files) do not interfere with the scarce
resource of the 'last mile' capacity".

The problem in the developing-world is not the last-mile capacity
(copper over v.92 can carry up to 64kb/sec) but the rate limiting
imposed by the ISP. On a pre-paid (scratch-card) dialup connection,
ISPs typically limit the BW in the range of 10-20kb/sec. This was
addressed by our previous work on modified bitTorrent to enable modem-
speed dialup networking, much like the pre-Internet FidoNet. You may
want to read our paper which appeared in the Oct'7 Issue of the ACM
Computer Communication Review Journal: http://www.sigcomm.org/ccr/drupal/?q=node/282

(2) Re: comments: "Most medium to large operators use caching and the

small ISPs 'peer' into them in some manner. The number of users in the
totally 'uncached' territory are very few".

We were also surprised by the fact very few ISPs are doing adequate
caching in Pakistan. Of course, the largest ISP (PTCL) does not cache
at all! Even when the smaller ISPs cache, the caches are too
fragmented to be useful. We could not find a single ISP in Pakistan
which peers over ICP or CARP with another ISP's cache. Pakistan does
not have an IXP, so there is of course no peering at the routing level
either. And given that the upstream ISP (invariably PTCL) does not
cache itself, there is essentially no consolidated cache of Internet
content in Pakistan. This role is typically played by companies such
as Inktomi (in the late 90s) and Akamai (currently) in US. There is no
equivalent of this in Pakistan. Hence our proposal.

(3) re: comment: "Most of the 'speed' of up and downlink of

asynchronous links is needed by the users to browse themselves. This
sharing of bandwidth by distributed caching has to be experimented
over a large study sample to be accepted as being valid".

On any Internet connection, a large fraction of the time is spent by
users reading the content they download over the Internet (Email,
News, blogs). The Internet link, in a typical browsing session, is
typically utilized in the range of 50-60% when the user is simply
reading what he has just downloaded. There is certainly spare
capacity, which may be used utilized by someone else if a donation
mechanism existed. Hence our work.

Since I try not to post to forums, please feel free to contact me
directly at um...@mit.edu if you have more questions.

Cheers,
-Umar

On Aug 21, 8:19 pm, "Qasim Khan" <qa...@qasimkhan.com> wrote:
> Fully agreed with Salman Ansari, real environment modification and
> acceptability shall be necessary. It would also help if a little more
> explanation and segmentation is worked on!
>
> Qasim
>
> On Thu, Aug 21, 2008 at 2:06 PM, Salman Ansari <sal...@super.net.pk> wrote:
> >  I tend to agree with Sameer on the issues he has highlighted from the
> > perspective in Pakistan. The situation on the ground is considerably
> > different on many counts from that assumed in this article. The analogy of
> > using computing cycles to 'sharing' bandwidth is not valid: one uses spare
> > capacity where it can be effectively used without any external network
> > constraints and the results (which are small files) do not interfere with
> > the scarce resource of the 'last mile' capacity.
>
> > Most medium to large operators use caching and the small ISPs 'peer' into
> > them in some manner. The number of users in the totally 'uncached' territory
> > are very few.
>
> > Most of the 'speed' of up and downlink of asynchronous links is needed by
> > the users to browse themselves. This sharing of bandwidth by distributed
> > caching has to be experimented over a large study sample to be accepted as
> > being valid.
>
> > But am open minded to be excited enough to see if this theory or a modified
> > version can be put to practical use in a real environment.
>
> > Salman
>

> > *From:* telecom-gr...@googlegroups.com [mailto:
> > telecom-gr...@googlegroups.com] *On Behalf Of *Sameer Bokhari
> > *Sent:* Thursday, August 21, 2008 12:45 PM
> > *To:* telecom-gr...@googlegroups.com
> > *Subject:* Re: LUMS Research On Boosting Bandwidth Featured In MIT

> > Technology Magazine
>
> > My congrats to Dr. Saif and his team
>
> > I have gone through the first 3 pages and I am pasting below some lines
> > from the details available on the project on the web, along with my novice
> > questions/commments, that in my opinion are noticable.
>
> > I would appreciate a comment from experienced members in the ISP/telco
> > field.
>
> >  more than 40 countries have less than 10Mbps of international Internet
> > bandwidth, whereas in Belgium, a 9Mbps ADSL high-speed Internet package is
> > available for just USD 80 a month [1]. "The Digital Divide at a Glance",
> > World Summit on the Information Society, Tunis 2005.
>
> > Dated source?
>
> > A 56Kbps modem can typically achieve an average throughput of close to

> > 40Kbps,more than *twice the bandwidth available over a typical (10-20Kbps)
> > dialup Internet connection in Pakistan.*

> > On Wed, Aug 20, 2008 at 7:48 PM, Babar <babar.bha...@gmail.com> wrote:
>
> > September edition of the MIT Technology Review has featured the work
> > of Umar Saif, a professor at LUMS, on improving Internet connectivity
> > in the developing-world. The Project, dubbed DonateBandwidth, is a
> > follow-up project of Poor Man's Broadband work which I wrote about
> > previously. Our congratulations to Dr. Saif and the team.
>
> > Umar shared his thoguhts in an e-mail:
>
> >    With DonateBandwidth, users in the developing-world can help each
> > other by donating their unused bandwidth to those who need it. This
> > project received funding from the US State Department/NAS and HEC and
> > will be further developed in collaboration with UC Berkeley.
>
> > More At:
>

> >http://telecompk.net/2008/08/20/lums-research-on-boosting-bandwidth-f...
>
> > Babar
> > <br

[Mahmood Ahmed]

Thanks Mr. Fawad Niazi and Mr. Tariq Mustafa.

 

What applications are offered, like balance enquiry, etc?

From: Fawad Niazi

Sent: Friday, August 22, 2008 12:04 PM

To: telecom-gr...@googlegroups.com

[Sameer Bokhari]

 

Would have been more comfortable if the word typical had been backed up by some stats or references.

 

Let alone the ambiguities in the published research work, some of which i pointed out earlier, what i fail to understand is that how does the poor man's broadband intend to reach the poor, when the total no of potential dialup internet subscribers(ie those having access to fixed telephone lines) in the country is 4,405,161(source: pta) out of a population of approx 164,741,924 (July 2007) .(2.5%approx).

[Ayaz Ahmed]

I think that is the same we use with ZOng *100# . I tested that its a
pretty good service for users.

Regards,

Ayaz Toor

[Tee Emm]

I can add some 'contemporary practices and conditions' of the ISPs in Pakistan as of today:

1. The dialup ISPs have long 'stopped' limiting transit bandwidth they provide to their users in both T1 and T2 cities. However, T3 cities where very very small POPs have been established by some really small ISPs (2/4/8 PRIs) are still limited by the intercity DPLC (leased line) capacities' high cost and these are the only places where the users actually do not get enough bandwidth that they would need on a 56k dialup connection. But still, the 'bandwidth limiting' is not being done by the ISP on a per user basis rather it is just a matter of bandwith being not available to the entire POP itself.
   
2. With transit bandwidth getting cheaper, I do not know of any ISP at least in Karachi that transparently caches its users. Optional caches might be available but it used by a few knowledgeable users only. The hassle to operate caches in a youtube world is far more than adding bandwidth. (This state, however, does not mean that Pakistan does not need an IX!)

-T

On Fri, Aug 22, 2008 at 6:41 PM, umar <saif...@gmail.com> wrote:

--
Tariq Mustafa MSN:t...@hotmail.com | Follow me here: http://friendfeed.com/abunet
Active Cell # http://tinyurl.com/2ob35t | Stuff I am sharing: http://tinyurl.com/2kagdu
Industry Blog: http://pakng.wordpress.com


Jay London  - "I saw a stationery store move."

[Rakesh Kumar]

Rehan,

There is something else from Microsoft called MCL (Mesh Connectivity Layer). I have tried it back in 2005 during a university course project, however MCL back then was severely buggy but their mailing list was quite nice help. MCL basically adds another network layer which supports the Mesh logic, it is being used as research test-bed but integrates nicely with Windows XP environment and just does that. Turn on your PC and you are a part of Mesh.

Rakesh.
http://gopchandani.wordpress.com

[Sameer Bokhari]

On Sat, Aug 23, 2008 at 10:45 AM, Tee Emm <tariq....@gmail.com> wrote:

1. With transit bandwidth getting cheaper, I do not know of any ISP at least in Karachi that transparently caches its users.
   

I know atleast one that was doing it in the past. Not sure if that one or others are doing it currently.

Cyber along with another company launched a super speed software for dialup users which cybernet called the CyberSpeed which also based on caching the content to improve the browsing experience.

 

Please do correct me if i am wrong.

 

SB

[Ashar]

On Saturday 23 August 2008 04:42:23 Sameer Bokhari wrote:
> On Sat, Aug 23, 2008 at 10:45 AM, Tee Emm <tariq....@gmail.com> wrote:

> > 1. With transit bandwidth getting cheaper, I do not know of any ISP at

> > least in Karachi that transparently caches its users.
> >
> > I know atleast one that was doing it in the past. Not sure if that
>
> one or others are doing it currently.
> Cyber along with another company launched a super speed software for dialup
> users which cybernet called the CyberSpeed which also based on caching the
> content to improve the browsing experience.
>
> Please do correct me if i am wrong.
>
> SB

I think it was based on compression - at least the software compressed the
transmission between the ISP and the dialup user.

Ashar