[Ict4d] Next Network Research group meeting - Monday 15 June 2:30PM ICT4D meeting room
David Johnson
djohnson at csir.co.za
Wed Jun 10 10:21:42 SAST 2015
Hi all
Our next Network Research group meeting should be a pretty lively one.
We are going to be discussing this very interesting email thread we've
been having on
- Is DTN dead
- How to make applications / TCP stacks / operating systems / servers
developing-world network friendly ... well that was my take on the
topic of the discsussion
Alette Schoon has been having problems with uploads of audio/video
from hip-hop artists timing out over mobile networks which resonated
with some of our current discussions on poor performance of TCP on
certain operating systems ... so all this is tying together nicely
Alette will be skyping in from Rhodes
Any suggestions for a specific paper(s) to read - Melissa had some
interesting paper suggestions from Marshini Chetty ... I vote for "Why
is my internet slow?: making network speeds visible"
All suggestions welcome
David
___________________________
David Johnson
Principal Researcher
Networks and Media group
Meraka, CSIR
Adjunct Senior Lecturer
ICT4D, Computer Science Department
University of Cape Town
---------- Forwarded message ----------
From: Alette Schoon <A.Schoon at ru.ac.za>
Date: Wed, Jun 10, 2015 at 9:47 AM
Subject: Re: [Ict4d] Fwd: Is DTN dead?
To: Thomas Reitmaier <treitmaier at gmail.com>
Cc: Melissa Densmore <melissa.r.densmore at gmail.com>, David Johnson
<djohnson at csir.co.za>, Marion Walton <the.marion.walton at gmail.com>,
Richard Maliwatu <richardmaliwatu at hotmail.com>
Very interesting, Thomas, but I'm afraid I have no idea what it means
and how this affects the upload. I somehow assumed that the internet
is totally immaterial, and that as your signal bounces around in
cyberspace it doesn't really matter where it goes, it's all the same
for the network. Hopefully you will enlighten me on Monday.
Quoting Thomas Reitmaier <treitmaier at gmail.com>:
> Hi Alette,
> look at the following traceroute, which prints out at the steps (hops in
> networking speak) and time it takes in milliseconds for me to reach the
> servers of kasimp3.co.za from my internet access point at home. Don't worry
> too much about the various numbers, just focus on the domain names. Steps
> 1-3 show my home router and how it is connected to apartment complex's
> internet connection. In Steps 4 - 10, you'll notice that my apartment
> complex is connected to the internet through MTN business. All pretty
> standard so far. Then it gets interesting: The next hops (11) are in
> Amsterdam and (12-14) in Phoenix, Arizona (USA).
>
> » traceroute kasimp3.co.za
> traceroute to kasimp3.co.za (69.64.85.104), 64 hops max, 52 byte packets
> 1 10.0.0.4 (10.0.0.4) 1.731 ms 1.386 ms 1.112 ms
> 2 hotspot.albionnet.co.za (192.168.1.254) 4.193 ms 35.796 ms 15.092 ms
> 3 * * *
> 4 41.181.53.149 (41.181.53.149) 119.075 ms 169.156 ms 140.773 ms
> 5 ipc-recieve-tb-1a.mtnbusiness.net (41.181.53.150) 86.456 ms 24.188
> ms 16.527 ms
> 6 tb-dca-2.za--qux-i.za.mtnbusiness.net (66.8.11.186) 28.924 ms
> tb-dca-2.za--qux-a.za.mtnbusiness.net (41.181.184.28) 74.133 ms
> 32.935 ms
> 7 compj-cpt-1.mtnns.net (196.44.18.2) 23.149 ms 24.591 ms 12.113 ms
> 8 ls-cr-2--tb-cr-1.uk-b.mtnns.net (196.44.31.113) 159.815 ms 196.263
> ms 168.846 ms
> 9 am-cr-1.nl--ls-cr-2.uk-a.mtnns.net (196.44.31.183) 182.104 ms 192.057
> ms 260.215 ms
> 10 am-tpr-1.nl--am-cr-1.nl-a.mtn.net (209.212.111.141) 194.530 ms
> 172.623 ms 213.045 ms
> 11 xe-4-1-0-135.edge5.amsterdam1.level3.net (212.72.41.89) 172.244 ms
> 201.675 ms 194.240 ms
> 12 ae-0-11.bar2.phoenix1.level3.net (4.69.148.114) 351.307 ms 325.790 ms
> *
> 13 aph-inc.dba.bar2.phoenix1.level3.net (4.28.82.158) 355.671 ms 335.862
> ms 516.751 ms
> 14 edge1_cr1.phx.codero.com (216.55.184.98) 370.878 ms 310.218 ms
> 303.125 ms
> 15 69.64.66.26 (69.64.66.26) 358.165 ms 342.422 ms 339.437 ms
> 16 kasimp3.co.za (69.64.85.104) 323.978 ms 395.737 ms 319.839 ms
>
>
> So in effect your participants are uploading their stuff to, and
> downloading it from, a datacenter in Arizona.
>
> Thomas
>
>
>
> On Tue, Jun 9, 2015 at 5:43 PM, Alette Schoon <A.Schoon at ru.ac.za> wrote:
>
>> I would love to write up something, Melissa, especially as our university
>> now recognizes ComSci conferences in its points system. CHI or AfriCHI
>> sound like the best option to get it out before anyone scoops me. Any
>> chance that we could have a 'How to write a ComSci conference paper'
>> workshop during my next visit in July? What do you think, Marion?
>>
>>
>>
>> Quoting Melissa Densmore <melissa.r.densmore at gmail.com>:
>>
>> I love it - Alette you should write a paper on p2p pedestrian networks! I
>>>
>>> think it's widely understood as something that happens, but there really
>>> isn't a lot written about it, and it's important that it gets documented
>>> as
>>> a phenomenon! It's related to some of the work on intermediaries (of
>>> which
>>> there is also not a lot), and as David mentioned there hasn't been a lot
>>> of
>>> work on DTNs recently. Would you be interested in writing up something
>>> for
>>> CHI2016 (due this september) or CSCW2017 (due probably May next year)? Or
>>> maybe ICTD2016 (due september?) or AfriCHI2016 (due 5 october)
>>>
>>> Also related - and in response to David's last email, I've been thinking
>>> for a while about a paper on appropriate design paradigms for applications
>>> to be developing-world-network friendly. I sort of think of it as POP3 vs
>>> AJAX mail. It's not that complicated - longer timeouts, no chatty
>>> mechanism, forgiving of network drops, but I'm not sure how to frame an
>>> investigation around it. Is the answer really to push everything to a
>>> gateway rather than getting software developers to be aware of parts of
>>> the
>>> Internet that aren't always on and bandwidth unlimited?
>>>
>>> Melissa
>>>
>>> On Tue, Jun 9, 2015 at 4:20 PM, Thomas Reitmaier <treitmaier at gmail.com>
>>> wrote:
>>>
>>> Thanks Everyone for the interesting questions and discussions!
>>>>
>>>>
>>>> Alette, I think it might be helpful if you could give a few more details
>>>> on the problems your participants are having. Off the top of my head,
>>>> these
>>>> questions will help us direct our discussions:
>>>>
>>>> What types of phones are they using? And, if possible, what Operating
>>>> System version are they running?
>>>>
>>>> What cell phone network are they using (e.g. MTN, vodacom)? I'm assuming
>>>> pre-paid? Are they purchasing data-bundles?
>>>>
>>>> What type of network are they connecting to (e.g. 2G, 3G)?
>>>>
>>>> What service are they trying to upload to?
>>>>
>>>> Are they uploading through an App or through the web browser?
>>>>
>>>> What type of content and file sizes?
>>>>
>>>>
>>>> And please do let us know if you'd like to skype into our meeting. I'd be
>>>> more than happy to set that up.
>>>>
>>>> Thomas
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Tue, Jun 9, 2015 at 4:02 PM, David Johnson <djohnson at csir.co.za>
>>>> wrote:
>>>>
>>>> These are great discussions for our next network research group
>>>>>
>>>>> meeting - I'll schedule one for Monday 15 June now that our Masi trip
>>>>> will be moved to Friday 19 June.
>>>>>
>>>>> I've been having some conversations about performance issues with some
>>>>> of my colleagues at CSIR on other devices like XBOX that don't like
>>>>> high-delay links and even androids apps that have issues ... I sense
>>>>> there is a lot to uncover at various layers of the stack to really
>>>>> understand why apps / various browser-based services perform poorly
>>>>> when the network isn't perfect .. see conversation below...
>>>>>
>>>>> >>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>>
>>>>> Hi Kobus - that's interesting. I checked my Android TCP settings and
>>>>> the TCP receive window is fairly generous
>>>>>
>>>>> cat /proc/sys/net/core/rmem_max = 2,097,152 so basically 2MB ... this
>>>>> would mean on your 5 Mbps link the receive window of your android
>>>>> phone would only become a bottleneck when your delay is higher than
>>>>> about 4s.
>>>>>
>>>>> Apps sometimes have their own statically defined internal flow
>>>>> control buffers that don't like high delay - for example you will
>>>>> notice ssh often has different performance to iperf with high delay
>>>>> nets - that's because its not very well tuned to high delay .. some of
>>>>> the phone apps are probably the same. In addition servers also don't
>>>>> like lots of hanging sockets (they consider a few seconds as a long
>>>>> time) and kick you off. This was something that was really frustrating
>>>>> in Zambia - servers would kick you off and you would have to keep
>>>>> retrying a download with no differential recovery - The TCP connection
>>>>> actually was surviving but the server threw you off.
>>>>>
>>>>> Of course your Microsoft box will suffer from the above + a TCP stack
>>>>> (I assume it is the same as Windows) not well tuned to the occasional
>>>>> high delay when the mesh is congested.
>>>>>
>>>>> My sense is that every home and village could do with an intelligent
>>>>> box (connected to your 3G / ADSL ... modem) with the kind of TCP
>>>>> acceleration and network optimization we see in satellite modems +
>>>>> local cloud (like our VillageShare/owncloud) for sync which then
>>>>> intelligently syncs to your favourite cloud service like Dropbox when
>>>>> your net is not used + allows some level of priority on which folders
>>>>> must sync immediately (like work docs) or can wait for periods of low
>>>>> use (e.g. photos / videos etc.) + detecting OS queries for updates /
>>>>> virus updates triggering a sync to the intelligent box at low usage
>>>>> period (e.g. night) which then are synced to your device later.
>>>>>
>>>>> Something many people also don't know is the importance of throttling
>>>>> their outbound link to around 90% of the known capacity of the link.
>>>>> This prevents the outbound being driven to congestion when uploading
>>>>> (e.g. dropbox sync) and dropping ACKS for web / file downloads - which
>>>>> can have a massive impact on your incoming capacity. This is
>>>>> particularly vital on ADSL modems.
>>>>>
>>>>> I think there is a business opportunity in an intelligent box like the
>>>>> one above that will have a place in every home in places in the world
>>>>> where home connections are not fibre yet and the device could also be
>>>>> used at a rural village level in the case where connectivity is spread
>>>>> from a shared gateway.
>>>>>
>>>>> There are products for windows e.g. SpeedConnect Internet Accelerator
>>>>> - http://www.cbs-soft.com/ - but it would make much more sense to have
>>>>> these features in a gateway
>>>>> ___________________________
>>>>>
>>>>> David Johnson
>>>>> Principal Researcher
>>>>> Networks and Media group
>>>>> Meraka, CSIR
>>>>> Adjunct Senior Lecturer
>>>>> ICT4D, Computer Science Department
>>>>> University of Cape Town
>>>>>
>>>>>
>>>>> On Tue, Jun 9, 2015 at 3:35 PM, Melissa Densmore
>>>>> <melissa.r.densmore at gmail.com> wrote:
>>>>> > which paper? the one that david sent?
>>>>> >
>>>>> > Actually - none of them really relate to why phones time out when you
>>>>> do
>>>>> > large uploads. Delay Tolerant Networking (DTN) for rural networks
>>>>> largely
>>>>> > refers to what geeks call a sneakernet approach - basically when
>>>>> people
>>>>> use
>>>>> > SD cards and flash drives to share files instead of email, cloud-based
>>>>> file
>>>>> > systems, etc. For you - this would probably be only useful in the
>>>>> context
>>>>> > of people going to media kiosks to get movies, media and software.
>>>>> >
>>>>> > For a paper on third-party kiosks, I would read:
>>>>> > Smyth, Thomas N., et al. "Where there's a will there's a way: mobile
>>>>> media
>>>>> > sharing in urban india." Proceedings of the SIGCHI Conference on Human
>>>>> > Factors in Computing Systems. ACM, 2010. (attached)
>>>>> >
>>>>> > For this problem of stuff timing out, you might look at Marshini
>>>>> Chetty's
>>>>> > work on limited bandwidth. There's a couple of explanations - the
>>>>> phone
>>>>> > might be low on battery and suspend the network connection. They
>>>>> might
>>>>> have
>>>>> > poor network connectivity and just not be able to send a large file
>>>>> with the
>>>>> > connection they have. They could be running out of airtime in the
>>>>> middle of
>>>>> > an upload. Perhaps they are running too many applications and the OS
>>>>> killed
>>>>> > the upload process before it was finished. They could be uploading to
>>>>> a
>>>>> > website or server that is unfriendly to disconnections and timeouts -
>>>>> that
>>>>> > could be a server implementation problem or a problem more inherent to
>>>>> the
>>>>> > communication protocol, such as the one David is documenting right now
>>>>> in
>>>>> > his paper on why skype for windows doesn't work over satellite and
>>>>> skype for
>>>>> > unix/osx does.
>>>>> >
>>>>> > For Marshini's work:
>>>>> > Chetty, Marshini, et al. "You're capped: understanding the effects of
>>>>> > bandwidth caps on broadband use in the home." Proceedings of the
>>>>> SIGCHI
>>>>> > Conference on Human Factors in Computing Systems. ACM, 2012.
>>>>> > Chetty, Marshini, et al. "Who's hogging the bandwidth: the
>>>>> consequences
>>>>> of
>>>>> > revealing the invisible in the home." Proceedings of the SIGCHI
>>>>> Conference
>>>>> > on Human Factors in Computing Systems. ACM, 2010.
>>>>> > Chetty, Marshini, et al. "Why is my internet slow?: making network
>>>>> speeds
>>>>> > visible." Proceedings of the SIGCHI Conference on Human Factors in
>>>>> Computing
>>>>> > Systems. ACM, 2011.
>>>>> > Wyche, Susan P., et al. "Deliberate interactions: characterizing
>>>>> technology
>>>>> > use in Nairobi, Kenya." Proceedings of the SIGCHI Conference on Human
>>>>> > Factors in Computing Systems. ACM, 2010.
>>>>> >
>>>>> > Eminently discuss-able at the next Networking meeting!
>>>>> >
>>>>> > Melissa
>>>>> >
>>>>> > On Tue, Jun 9, 2015 at 3:01 PM, Alette Schoon <A.Schoon at ru.ac.za>
>>>>> wrote:
>>>>> >>
>>>>> >>
>>>>> >> Hi all
>>>>> >>
>>>>> >> I tried to read the paper, but I'm afraid I did not understand much
>>>>> of
>>>>> it.
>>>>> >> Does this in any way relate to how people's phones time out when they
>>>>> try
>>>>> >> and upload large audio or video phones on the mobile internet?
>>>>> >> Because I'm trying to find some references that will explain why this
>>>>> >> happens among my hip hop artists in my study.
>>>>> >> Any explanations will be appreciated.
>>>>> >> All the best
>>>>> >> Alette
>>>>> >>
>>>>> >> Quoting Melissa Densmore <mdensmore at cs.uct.ac.za>:
>>>>> >>
>>>>> >>> I would hesitate to say "dead" but you are right, there hasn't been
>>>>> that
>>>>> >>> much work done recently in the space - pretty much since Keshav
>>>>> switched
>>>>> >>> over to environmental sensing. Even my Ghana Telemedicine Network
>>>>> stuff
>>>>> >>> was done back in 2008. The main difficulty I think we struggled
>>>>> with
>>>>> was
>>>>> >>> that it takes a ton of work to get a stable DTN going - and often by
>>>>> the
>>>>> >>> time we got something working, mobile phones or DSL would arrive and
>>>>> >>> immediately displace our efforts. At the same time, it's been 10
>>>>> years
>>>>> >>> and
>>>>> >>> some parts of Africa are still largely unconnected. I don't think
>>>>> the
>>>>> >>> applicability of DTN has reduced - but the number of researchers
>>>>> looking
>>>>> >>> to
>>>>> >>> leverage it in isolated places has greatly reduced. It's possible
>>>>> that
>>>>> >>> the
>>>>> >>> DTN reference architecture is simply too much overhead for rural
>>>>> >>> deployments, with no clear benefits (e.g. interoperability, etc) for
>>>>> >>> using
>>>>> >>> it. Tons of people use informal DTNs (flash drives, Netflix, media
>>>>> >>> kiosks)
>>>>> >>> without using a formal network architecture worthy of a paper.
>>>>> >>>
>>>>> >>> Some other relevant papers:
>>>>> >>> Luk, Rowena, Melissa Ho, and Paul M. Aoki. "Asynchronous remote
>>>>> medical
>>>>> >>> consultation for Ghana." *Proceedings of the SIGCHI conference on
>>>>> human
>>>>> >>> factors in computing systems*. ACM, 2008.
>>>>> http://arxiv.org/pdf/0801.1927
>>>>> >>>
>>>>> >>> Ho, Melissa, and Kevin Fall. "Poster: Delay tolerant networking for
>>>>> >>> sensor
>>>>> >>> networks." *Proc. of IEEE Conference on Sensor and Ad Hoc
>>>>> Communications
>>>>> >>> and Networks*. 2004.
>>>>> >>>
>>>>> >>>
>>>>>
>>>>> http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.59.3330&rep=rep1&type=pdf
>>>>> >>>
>>>>> >>>
>>>>> >>> Digital Study Hall was also originally a DTN-based system, most
>>>>> recently
>>>>> >>> presented in 2012, but focus has shifted to questions of interaction
>>>>> >>> rather
>>>>> >>> than content delivery
>>>>> >>> Richard Anderson, Chad Robertson, Esha Nabi, Urvashi Sahni, and
>>>>> Tanuja
>>>>> >>> Setia. 2012. Facilitated video instruction in low resource schools.
>>>>> In
>>>>> >>> *Proceedings
>>>>> >>> of the Fifth International Conference on Information and
>>>>> Communication
>>>>> >>> Technologies and Development* (ICTD '12). ACM, New York, NY, USA,
>>>>> 2-12.
>>>>> >>> DOI=10.1145/2160673.2160675
>>>>> http://doi.acm.org/10.1145/2160673.2160675
>>>>> >>>
>>>>> >>> There's also dtn stuff on SaamiNet (2009)
>>>>> >>> Doria, Avri, Maria Uden, and D. Pandey. "Providing connectivity to
>>>>> the
>>>>> >>> saami nomadic community." *generations* 1.2 (2009): 3.
>>>>> >>>
>>>>> >>>
>>>>> >>> David - I would pose this question on the gaia mailing list, which
>>>>> >>> includes
>>>>> >>> a lot of the people that originally worked on the DTN architecture
>>>>> for
>>>>> >>> both
>>>>> >>> space and rural motivations.
>>>>> >>>
>>>>> >>> Melissa
>>>>> >>>
>>>>> >>> On Mon, Jun 8, 2015 at 3:56 PM, David Johnson <djohnson at csir.co.za>
>>>>> >>> wrote:
>>>>> >>>
>>>>> >>>> Hi all
>>>>> >>>>
>>>>> >>>> One of my lectures in my Net4D honours modules is on the use of DTN
>>>>> as
>>>>> >>>> a connectivity option for poorly connected areas. The slides I have
>>>>> >>>> are mostly about historical work such as DakNet (2004) and KioskNet
>>>>> >>>> (2007)
>>>>> >>>>
>>>>> >>>> My sense is that this research has mainly continued to have
>>>>> traction
>>>>> >>>> in the realm of Vehicular networks and space networks and that its
>>>>> no
>>>>> >>>> longer considered as a serious contender for rural/poorly connected
>>>>> >>>> regions ... there are some interesting new developments that mix
>>>>> >>>> social networks and delay tolerant networks - but they seem mostly
>>>>> >>>> like playful ideas than something to take seriously.
>>>>> >>>>
>>>>> >>>> Then you get weirdness like this 2014 paper from Disney on a DTN
>>>>> >>>> solution for mico-entrepeneurs using DTN-enabled projects (cinemas
>>>>> in
>>>>> >>>> a backpack) - the example being moving content between urban areas
>>>>> in
>>>>> >>>> Pretoria and an under-served ex-homeland North of Pretoria - there
>>>>> is
>>>>> >>>> not one African researcher on the paper ... I think I will use this
>>>>> as
>>>>> >>>> a prime example of what goes wrong when communities don't
>>>>> participate
>>>>> >>>> in the research process.
>>>>> >>>>
>>>>> >>>> So DTN for rural/under-serviced areas - dead?
>>>>> >>>>
>>>>> >>>> David Johnson
>>>>> >>>> Principal Researcher
>>>>> >>>> Networks and Media group
>>>>> >>>> Meraka, CSIR
>>>>> >>>> Adjunct Senior Lecturer
>>>>> >>>> ICT4D, Computer Science Department
>>>>> >>>> University of Cape Town
>>>>> >>>>
>>>>> >>
>>>>> >>
>>>>> >>
>>>>> >>
>>>>> >> _______________________________________________
>>>>> >> ICT4D mailing list
>>>>> >> ICT4D at cs.uct.ac.za
>>>>> >> http://mailman.cs.uct.ac.za/cgi-bin/mailman/listinfo/ict4d
>>>>> >
>>>>> >
>>>>> >
>>>>> > --
>>>>> > This message is subject to the CSIR's copyright terms and conditions,
>>>>> e-mail
>>>>> > legal notice, and implemented Open Document Format (ODF) standard.
>>>>> > The full disclaimer details can be found at
>>>>> > http://www.csir.co.za/disclaimer.html.
>>>>> >
>>>>> >
>>>>> > This message has been scanned for viruses and dangerous content by
>>>>> > MailScanner,
>>>>> > and is believed to be clean.
>>>>> >
>>>>> >
>>>>> > Please consider the environment before printing this email.
>>>>>
>>>>>
>>>>
>>>>
>>
>>
--
This message is subject to the CSIR's copyright terms and conditions,
e-mail legal notice, and implemented Open Document Format (ODF)
standard.The full disclaimer details can be found at
http://www.csir.co.za/disclaimer.html.
This message has been scanned for viruses and dangerous content by
MailScanner,and is believed to be clean.
Please consider the environment before printing this email.
More information about the ICT4D
mailing list