802.11 Ad-hoc vs Infrastructure: Latency, throughput, and scaling.

Started by
4 comments, last by Emergent 13 years, 6 months ago
Hi,

I have a situation where a number of computers in close proximity (in a single room, in fact) need to communicate with one another. They happen to be robots, but they might as well be PCs at a LAN party. They have 802.11 adapters, and I'm trying to understand the tradeoffs involved between ad-hoc and infrastructure mode wireless networking.

Latency
Infrastructure mode gives you a star topology. Ad-hoc gives you a complete graph (it's a good assumption in this example that all stations really can talk to one another). All agents are two hops apart with infrastructure, but only one hop away with ad-hoc. Advantage: Ad-hoc?

Throughput
I've read contradictory things about throughput. Some sources imply that the halved latency of ad-hoc also means double throughput vs infrastructure mode (this doesn't make sense to me). Others say that you get 54 Mbps in infrastructure and only 11 Mbps in ad-hoc. Who's right? Advantage: Infrastructure?

Scaling (Collisions)
Despite the different network topologies of infrastructure and ad-hoc, the physical reality is that all stations are sharing the same medium for either one -- so you'd expect roughly the same number of collisions with either, and in both cases this scales with the square of the number of stations, right? If anything, since twice the hops are required to get a packet from point A to point B in infrastructure mode, would you expect fewer collisions in ad-hoc mode? Advantage: Ad-hoc?

Thanks in advance for sharing any insight.

[Edited by - Emergent on October 16, 2010 12:28:58 AM]
Advertisement
My limited experience with Ad Hoc has been that it doesn't work very well, and you really want to be in infrastructure mode.
enum Bool { True, False, FileNotFound };
Quote:Original post by Emergent
Latency
Infrastructure mode gives you a star topology. Ad-hoc gives you a complete graph (it's a good assumption in this example that all stations really can talk to one another). All agents are two hops apart with infrastructure, but only one hop away with ad-hoc. Advantage: Ad-hoc?
Invalid assumption.

Assume crappy cards and drivers that can only connect to a single node. You may have people who only attach to each other as disconnected meshes. Assume that people's cell phones might get picked up as ad-hoc devices. (It may help your bandwidth but upset your guests.) Your network will be a collection of one or more partial meshes where communications are poorly forwarded through many nodes, if they are forwarded at all.

Better to have a fixed step of exactly one hop in a star, or minimal hops between your own gigabit wireless routers, than a hodge-podge of potentially unreachable nodes.

Quote:Throughput
I've read contradictory things about throughput. Some sources imply that the halved latency of ad-hoc also means double throughput vs infrastructure mode (this doesn't make sense to me, but let's move on). Others say that you get 54 Mbps in infrastructure and only 11 Mbps in ad-hoc. Who's right? Advantage: Infrastructure?
Basically correct.

A good wireless 802.11n routers will have multiple antenna that can provide 11, 54, 150, and 300 Mbps simultaneously.

They are not expensive. Get enough to properly cover the floor and the surrounding areas, and configure them so they frequencies don't overlap too much.


Quote:Scaling (Collisions)
Despite the different network topologies of infrastructure and ad-hoc, the physical reality is that all stations are sharing the same medium for either one -- so you'd expect roughly the same number of collisions with either, and in both cases this scales with the square of the number of stations, right? If anything, since twice the hops are required to get a packet from point A to point B in infrastructure mode, would you expect fewer collisions in ad-hoc mode? Advantage: Ad-hoc?
No, you want the channels spread out. You can do this with a floor plan and concentric circles indicating where each channel is in use. Carefully planned you can cover a very large area with no channel overlap. A planned infrastructure of multiple wireless hubs on multiple channels scales best.
Quote:Original post by frob
Assume crappy cards and drivers that can only connect to a single node.


Thank you very much for your reply. However, I probably didn't describe the situation clearly enough. I actually have control over the hardware. If the drivers are crappy, I'll find that out, but there won't be variability.

Quote:You may have people who only attach to each other as disconnected meshes.


All devices in question are in a single room. I'd be very surprised if they were not all within radio range of one another; we're talking about distances of maybe 60 ft in the absolute worst case. Is this still a problem?

Quote:communications are poorly forwarded through many nodes, if they are forwarded at all.


Is mesh routing part of 802.11 ad-hoc mode? I'd been under the impression that out-of-range computers simply dropped packets. At what level does the routing happen? Is it part of 802.11, or is it built on top of it at the TCP level?

Quote:your own gigabit wireless routers


There will be at most one router if I go infrastructure mode, since as I said we're talking about a single room. The "scaling" -- maybe bad word choice on my part -- is not about enlarging the spatial coverage of the network, but rather about increasing the number of peers concentrated in the same small area. Imagine 50+ computers in the same small room.

Not a typical situation, I know!
What frob says still holds under your requirements, in my experience. No 802.11 providers actually optimize or tune for ad hoc mode. You can spend months to try to find compatible hardware and work on drivers (assuming you're using open source and are comfortable with drivers), or you can spend $70 on a 802.11n MIMO wireless access point, and have it all work almost right away, almost all the time.

I would choose infrastructure for robustness reasons. The actual throughput is likely to be at least as good as ad hoc; probably better in practice. Note that ad hoc nodes generally won't connect on multiple frequencies, so even though there may be "point to point" connections, it's all a single physical aether.

I'd probably also splurge for a higher-quality base station from some well known upscale hardware provider (I don't have experience with that particular device, but it looks okay), paying more money, unless I was really budget constrained. This is because "enterprise" devices give more analytics (metrics) and possibly have better debugged firmware (or at least answer when you try to call them).
enum Bool { True, False, FileNotFound };
Thanks frob & hplus; sounds like there's consensus on this.

This topic is closed to new replies.

Advertisement