Re: [OLPC-AU] limits on ad-hoc connections
On Wed, Feb 8, 2012 at 10:59 PM, James Cameron qu...@laptop.org wrote: No, that's not how ad-hoc works. I'll simplify and translate for you. Your explanation is correct but doesn't exactly match the buggy behaviour of our wireless hardware/firmware. As far as I can tell, the ad-hoc nodes in our setup make no attempt to synchronize clocks/beacons, nor do they stop transmitting their own beacons in the presence of another node transmitting beacons too. I guess this is why our ad-hoc networks perform slower and less reliably than the XO-1 mesh, which at least had synchronization. Daniel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
limits on ad-hoc connections
Ad-hoc connections only scale to a limited number of participants before problems begin to occur. What would be the maximum number of participants that an ad-hoc network can reliably handle? Can we impose a hard limit on the number of clients to prevent too many XOs connecting to a single ad-hoc session? Sridhar Sridhar Dhanapalan Engineering Manager One Laptop per Child Australia ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: [OLPC-AU] limits on ad-hoc connections
On Wed, Feb 08, 2012 at 10:16:53PM +1100, Sridhar Dhanapalan wrote: Ad-hoc connections only scale to a limited number of participants before problems begin to occur. What are the problems you observe? It may be that the problems you observe are not due to the ad-hoc network, but due to something else as well. What would be the maximum number of participants that an ad-hoc network can reliably handle? There's no maximum that I know of. A well placed set of laptops that can hear each other, with no outside noise, can operate an ad-hoc network to a quite large size. One node will be the beacon. Once you place traffic on the network, things will slow down. Once the slow down is enough, certain applications may fail. Can we impose a hard limit on the number of clients to prevent too many XOs connecting to a single ad-hoc session? No, there is no control for that as far as I know. -- James Cameron http://quozl.linux.org.au/ ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: limits on ad-hoc connections
On Wed, Feb 8, 2012 at 6:16 AM, Sridhar Dhanapalan srid...@laptop.org.au wrote: Ad-hoc connections only scale to a limited number of participants before problems begin to occur. The technically correct answer is it depends. And it is true, it depends on a ton of factors. As a rule of thumb, I've seen it work for groups of 5~6 units, physically close and without interference sources or reflective materials. I would not aim higher than that -- 5~6 units in a channel. You have 3 channels, so 3 groups of 5~6 units. To clarify: keep any other laptops and cordless phones in the vicinity _off_, to allow these 18 users to work. In practice, it won't work in a school, but if you invite a few schoolmates home after school, or in the park, you're fine. No warranties expressed or implied. There's a long laundry list of things that can interfere, and make things not fine. For example, professional TV cameras from that friendly news crew transmit in the 2.4GHz band. That battery pack feeds a powerful antenna to get the signal back to the van that has the uplink, and it paves over consumer-grade wifi. So don't count in wifi (of any kind!) to work for a demo or show-and-tell when you get TV coverage at a school :-) Can we impose a hard limit on the number of clients to prevent too many XOs connecting to a single ad-hoc session? As James says... unfortunately no. m -- martin.langh...@gmail.com mar...@laptop.org -- Software Architect - OLPC - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: limits on ad-hoc connections
On Wed, Feb 8, 2012 at 4:23 AM, Martin Langhoff martin.langh...@gmail.com wrote: On Wed, Feb 8, 2012 at 6:16 AM, Sridhar Dhanapalan srid...@laptop.org.au wrote: Ad-hoc connections only scale to a limited number of participants before problems begin to occur. The technically correct answer is it depends. And it is true, it depends on a ton of factors. As a rule of thumb, I've seen it work for groups of 5~6 units, physically close and without interference sources or reflective materials. I would not aim higher than that -- 5~6 units in a channel. You have 3 channels, so 3 groups of 5~6 units. wiki.laptop.org/images/a/a3/Country_Technical_Support.pdf recommends 10 per channel on mesh. Given that 802.11s draft vs ad-hoc is really a layer 2 issue, the numbers should be in that neighborhood? cheers, Sameer To clarify: keep any other laptops and cordless phones in the vicinity _off_, to allow these 18 users to work. In practice, it won't work in a school, but if you invite a few schoolmates home after school, or in the park, you're fine. No warranties expressed or implied. There's a long laundry list of things that can interfere, and make things not fine. For example, professional TV cameras from that friendly news crew transmit in the 2.4GHz band. That battery pack feeds a powerful antenna to get the signal back to the van that has the uplink, and it paves over consumer-grade wifi. So don't count in wifi (of any kind!) to work for a demo or show-and-tell when you get TV coverage at a school :-) Can we impose a hard limit on the number of clients to prevent too many XOs connecting to a single ad-hoc session? As James says... unfortunately no. m -- martin.langh...@gmail.com mar...@laptop.org -- Software Architect - OLPC - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: limits on ad-hoc connections
On Wed, Feb 8, 2012 at 2:56 PM, Sameer Verma sve...@sfsu.edu wrote:
wiki.laptop.org/images/a/a3/Country_Technical_Support.pdf recommends
10 per channel on mesh. Given that 802.11s draft vs ad-hoc is really a
layer 2 issue, the numbers should be in that neighborhood?
I would fix that document instead, to bring it from dreams of happy
fluffy bunnies to the realities of what's been seen to work.
Back in the day, we spent a considerable time at OLPC waiting for mesh
to work, looking at our tests outcomes with optimistic eyes, and the
hope that all we needed was just one more bugfix. So don't trust the
networking performance statements of documents dating back to that
era.
{ Perhaps I wrote that paragraph you refer to. I wrote a few like
that. I am sorry. Mea culpa. }
m
--
martin.langh...@gmail.com
mar...@laptop.org -- Software Architect - OLPC
- ask interesting questions
- don't get distracted with shiny stuff - working code first
- http://wiki.laptop.org/go/User:Martinlanghoff
___
Devel mailing list
Devel@lists.laptop.org
http://lists.laptop.org/listinfo/devel
Re: limits on ad-hoc connections
On 8 February 2012 23:23, Martin Langhoff martin.langh...@gmail.com wrote: On Wed, Feb 8, 2012 at 6:16 AM, Sridhar Dhanapalan srid...@laptop.org.au wrote: Ad-hoc connections only scale to a limited number of participants before problems begin to occur. The technically correct answer is it depends. And it is true, it depends on a ton of factors. As a rule of thumb, I've seen it work for groups of 5~6 units, physically close and without interference sources or reflective materials. I would not aim higher than that -- 5~6 units in a channel. You have 3 channels, so 3 groups of 5~6 units. Great, that's what I was thinking. To clarify: keep any other laptops and cordless phones in the vicinity _off_, to allow these 18 users to work. In practice, it won't work in a school, but if you invite a few schoolmates home after school, or in the park, you're fine. No warranties expressed or implied. There's a long laundry list of things that can interfere, and make things not fine. For example, professional TV cameras from that friendly news crew transmit in the 2.4GHz band. That battery pack feeds a powerful antenna to get the signal back to the van that has the uplink, and it paves over consumer-grade wifi. So don't count in wifi (of any kind!) to work for a demo or show-and-tell when you get TV coverage at a school :-) Interesting - definitely worth knowing! Can we impose a hard limit on the number of clients to prevent too many XOs connecting to a single ad-hoc session? As James says... unfortunately no. This is possible on many wireless access points. Why isn't it possible on the XO's ad-hoc? Is it because WAPs do it by limiting DHCP leases, whereas ad-hoc uses link-local? Sridhar Dhanapalan Engineering Manager One Laptop per Child Australia ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: limits on ad-hoc connections
On Wed, Feb 8, 2012 at 10:39 PM, Sridhar Dhanapalan srid...@laptop.org.au wrote: Can we impose a hard limit on the number of clients to prevent too many XOs connecting to a single ad-hoc session? As James says... unfortunately no. This is possible on many wireless access points. Why isn't it possible on the XO's ad-hoc? Is it because WAPs do it by limiting DHCP leases, whereas ad-hoc uses link-local? More generally, when you have a central node (the AP) there's a node that can carry the accounting, and has the authority to say who's welcome and who's not. I don't know if 802.11a/b/g/n has a mechanism to reject association, or if it's a dirty hack with only giving a liminted number of DHCP leases. Either way, ad-hoc peer model isn't well equipped for this limitation. cheers, m -- martin.langh...@gmail.com mar...@laptop.org -- Software Architect - OLPC - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: limits on ad-hoc connections
On 9 February 2012 14:59, Martin Langhoff martin.langh...@gmail.com wrote: More generally, when you have a central node (the AP) there's a node that can carry the accounting, and has the authority to say who's welcome and who's not. I don't know if 802.11a/b/g/n has a mechanism to reject association, or if it's a dirty hack with only giving a liminted number of DHCP leases. Either way, ad-hoc peer model isn't well equipped for this limitation. Hmm I am thinking that my understanding of the ad-hoc implementation might be incorrect. I was under the assumption that one XO acts as the ad-hoc host, and the others connect to it. That made me wonder whether that host could limit how many clients connect to it. What I gather from what you're saying is that there's more of a peer-to-peer connection happening, similar to the old mesh on the XO-1s. Or am I confusing my network layers? Sridhar ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: limits on ad-hoc connections
sridhar wrote: On 9 February 2012 14:59, Martin Langhoff martin.langh...@gmail.com wrote: More generally, when you have a central node (the AP) there's a node that can carry the accounting, and has the authority to say who's welcome and who's not. I don't know if 802.11a/b/g/n has a mechanism to reject association, or if it's a dirty hack with only giving a liminted number of DHCP leases. Either way, ad-hoc peer model isn't well equipped for this limitation. Hmm I am thinking that my understanding of the ad-hoc implementation might be incorrect. I was under the assumption that one XO acts as the ad-hoc host, and the others connect to it. That made me wonder whether that host could limit how many clients connect to it. What I gather from what you're saying is that there's more of a peer-to-peer connection happening, similar to the old mesh on the XO-1s. Or am I confusing my network layers? think of it as all the XOs plugging into the same ethernet hub. no router needed, and they all see each other's traffic. you just plug in and start talking (of course to really do that, you'll need a link-local address, or a static address, since there's probably no DHCP). it's quite similar to the mesh -- what the mesh adds is some topology awareness and routing, so if A can see B and B can see C, then B will forward packets from A to C. that can't happen with ad-hoc. paul Sridhar ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel =- paul fox, p...@laptop.org ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: [OLPC-AU] limits on ad-hoc connections
On Thu, Feb 09, 2012 at 03:11:42PM +1100, Sridhar Dhanapalan wrote: Hmm I am thinking that my understanding of the ad-hoc implementation might be incorrect. I was under the assumption that one XO acts as the ad-hoc host, and the others connect to it. That made me wonder whether that host could limit how many clients connect to it. No, that's not how ad-hoc works. I'll simplify and translate for you. In an 802.11 wireless ad-hoc network, each node has the duty and right to be the beacon, especially if there is no other beacon heard. The beacon is used for timing the transmissions, so that they occur in empty time. Transmissions that occur simultaneously would interfere with each other, and the receivers would be more likely to miss them. Always, the first node to begin an ad-hoc network begins by being the beacon. If a node cannot hear a beacon, then after a very short while it will try to become the beacon. In effect, they compete for the job, in a psuedo-random fashion. This is implemented in the wireless device firmware, not in the host, not in the CPU, not in the kernel, not in the user-space networking tools. If a cluster of XOs that have formed an ad-hoc network, are slowly spread out physically, then eventually the responsibility for beacon should tend to be in the centre of the cluster. If a cluster of XOs is split in two, and the two groups slowly moved apart from each other, then eventually two beacons will be active and there will be two networks. References: http://en.wikipedia.org/wiki/Timing_Synchronization_Function_%28TSF%29 http://en.wikipedia.org/wiki/Beacon_frame http://en.wikipedia.org/wiki/Traffic_indication_map (ad-hoc is IBSS) -- James Cameron http://quozl.linux.org.au/ ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: [OLPC-AU] limits on ad-hoc connections
To expand on James' excellent notes... On Wed, Feb 8, 2012 at 11:59 PM, James Cameron qu...@laptop.org wrote: The beacon is used for timing the transmissions, so that they occur in And that is _all_ it can do. It just broadcasts a beacon, like a metronome for a band recording in a studio. This is implemented in the wireless device firmware, not in the host, not in the CPU, not in the kernel, not in the user-space networking tools. In other words, in practice we can't patch it to be smarter. m -- martin.langh...@gmail.com mar...@laptop.org -- Software Architect - OLPC - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
