
A client has a problem with a device that generates data faster than it's 3G modem can upload it. Is there a good LTE AKA 4G modem that can connect to a Linux system? Can you run two 3G modems next to each other and expect that each of them will get full bandwidth? Or will they interfere with each other in some way? -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

On 02/07/2013, at 2:34 PM, Russell Coker <russell@coker.com.au> wrote:
A client has a problem with a device that generates data faster than it's 3G modem can upload it.
Will a queuing system help is or this sustained throughput?
Can you run two 3G modems next to each other and expect that each of them will get full bandwidth? Or will they interfere with each other in some way?
Each tower has a fixed amount of spectrum and capacity that can be shared between users. 4G, like 3G and even GPRS, allocates capacity using a slot system, so each device on the network that wants to use the data service is allocated a slot. When I was living in the bush I had 2 3G modems which meant a lot of the time I was competing with myself. On the weekends when there were more users on the network I noticed a throughput drop on each link, but I got more bandwidth than I would have with 1. If there are 3 other users I would have got at least 2/5s of the slots, if there were 6 other users I would have got at least 1/4 of the slots. If your client is in a congested area, then there is a chance that 2 modems will improve their throughput, but if the network is lightly used, then they will see little benefit. If your client is serviced by multiple cells with poor signal, throughput will be compromised as the device continuously cell hops. Cheers Dave

On Tue, 2 Jul 2013, Dave Hall <dave.hall@skwashd.com> wrote:
On 02/07/2013, at 2:34 PM, Russell Coker <russell@coker.com.au> wrote:
A client has a problem with a device that generates data faster than it's 3G modem can upload it.
Will a queuing system help is or this sustained throughput?
Sustained throughput for a period that's long enough to fill all storage.
If your client is in a congested area, then there is a chance that 2 modems will improve their throughput, but if the network is lightly used, then they will see little benefit. If your client is serviced by multiple cells with poor signal, throughput will be compromised as the device continuously cell hops.
Thanks for the advice, it doesn't sound worth persuing the multiple 3G device option then. So I'm back to 4G/LTE support. Had anyone had a good experience with LTE modems on Linux for uploading data? I don't particularly care about receiving data except for ACKs. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

Just thinking out of the box here. If the limitation is on the tower, would going with two different carriers work? Hopefully if they are not trunking together, you should get better throughput. Regards Rasika On 2 July 2013 14:56, Dave Hall <dave.hall@skwashd.com> wrote:
On 02/07/2013, at 2:34 PM, Russell Coker <russell@coker.com.au> wrote:
A client has a problem with a device that generates data faster than it's 3G modem can upload it.
Will a queuing system help is or this sustained throughput?
Can you run two 3G modems next to each other and expect that each of
them will
get full bandwidth? Or will they interfere with each other in some way?
Each tower has a fixed amount of spectrum and capacity that can be shared between users. 4G, like 3G and even GPRS, allocates capacity using a slot system, so each device on the network that wants to use the data service is allocated a slot.
When I was living in the bush I had 2 3G modems which meant a lot of the time I was competing with myself. On the weekends when there were more users on the network I noticed a throughput drop on each link, but I got more bandwidth than I would have with 1. If there are 3 other users I would have got at least 2/5s of the slots, if there were 6 other users I would have got at least 1/4 of the slots.
If your client is in a congested area, then there is a chance that 2 modems will improve their throughput, but if the network is lightly used, then they will see little benefit. If your client is serviced by multiple cells with poor signal, throughput will be compromised as the device continuously cell hops.
Cheers
Dave _______________________________________________ luv-main mailing list luv-main@luv.asn.au http://lists.luv.asn.au/listinfo/luv-main
-- ============================== || Rasika Amarasiri, PhD || Rasika.Amarasiri@gmail.com ==============================

On Tue, 2 Jul 2013, Rasika Amarasiri <rasika.amarasiri@gmail.com> wrote:
Just thinking out of the box here. If the limitation is on the tower, would going with two different carriers work? Hopefully if they are not trunking together, you should get better throughput.
That's an interesting idea, but not viable in this case. Some of the devices I'm deploying will be in areas where only Telstra gives good reception. My general impression of the state of 3G in Australia is that Telstra will outperform all other options in most situations. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

I am assuming these are part of a large sensor network. If the devices should definitely have some processing capability. It may be good to do some local processing/compression to reduce the required bandwidth. Regards Rasika On 2 July 2013 15:59, Russell Coker <russell@coker.com.au> wrote:
On Tue, 2 Jul 2013, Rasika Amarasiri <rasika.amarasiri@gmail.com> wrote:
Just thinking out of the box here. If the limitation is on the tower, would going with two different carriers work? Hopefully if they are not trunking together, you should get better throughput.
That's an interesting idea, but not viable in this case. Some of the devices I'm deploying will be in areas where only Telstra gives good reception. My general impression of the state of 3G in Australia is that Telstra will outperform all other options in most situations.
-- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/
-- ============================== || Rasika Amarasiri, PhD || Rasika.Amarasiri@gmail.com ==============================

On Tue, 2 Jul 2013, Rasika Amarasiri <rasika.amarasiri@gmail.com> wrote:
I am assuming these are part of a large sensor network. If the devices should definitely have some processing capability. It may be good to do some local processing/compression to reduce the required bandwidth.
Good theory, but the devices in question are taking pictures. We need all the pics so there's not much we can do to reduce the volume of JPEGs. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

Russell Coker <russell@coker.com.au> writes:
On Tue, 2 Jul 2013, Rasika Amarasiri <rasika.amarasiri@gmail.com> wrote:
I am assuming these are part of a large sensor network. If the devices should definitely have some processing capability. It may be good to do some local processing/compression to reduce the required bandwidth.
Good theory, but the devices in question are taking pictures. We need all the pics so there's not much we can do to reduce the volume of JPEGs.
Well you could reduce the quality of the JPEGs. It's a lossy representation format after all :-) (If it's anything like the photos I've seen taken by cellphones, they're ridiculously large and low quality, but still encoded as high quality, so using imagemagick + jpegoptim to reduce it to a 75% 1080xthingy might not be as silly as it first sounds.)

On Wed, 3 Jul 2013, "Trent W. Buck" <trentbuck@gmail.com> wrote:
Good theory, but the devices in question are taking pictures. We need all the pics so there's not much we can do to reduce the volume of JPEGs.
Well you could reduce the quality of the JPEGs. It's a lossy representation format after all :-)
That's a nice theory, but my client knows what they want in terms of image quality. It's my job to arrange the data transfer. The pictures are actually of very high quality. We're talking about good DSLRs used for taking the pictures. But it's not the issue for me, I just have to do what my client wants whenever it's possible. I'm sure that they also have some significant Telstra 3G bills. But again it's not my issue, they are paying for the results that they want. -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/

On 04/07/2013, at 20:26, Russell Coker <russell@coker.com.au> wrote:
Well you could reduce the quality of the JPEGs. It's a lossy representation format after all :-)
That's a nice theory, but my client knows what they want in terms of image quality. It's my job to arrange the data transfer.
The pictures are actually of very high quality. We're talking about good DSLRs used for taking the pictures.
Do they need to be uploaded as soon as they're shot? Maybe batching the upload overnight while there's less 3G contention could alleviate the backlog

On 4/07/2013 8:26 PM, Russell Coker wrote:
That's a nice theory, but my client knows what they want in terms of image quality. It's my job to arrange the data transfer.
The pictures are actually of very high quality. We're talking about good DSLRs used for taking the pictures. But it's not the issue for me, I just have to do what my client wants whenever it's possible. I'm sure that they also have some significant Telstra 3G bills. But again it's not my issue, they are paying for the results that they want.
Would thumbnails be out of the question? Anything found interesting through thumbnail versions could trigger an upload of the fuller image.... Mass storage and a courier could handle things outside of 3G/4G -- fibre or whatever else is available. A.

On 04/07/13 20:26, Russell Coker wrote:
On Wed, 3 Jul 2013, "Trent W. Buck" <trentbuck@gmail.com> wrote:
Good theory, but the devices in question are taking pictures. We need all the pics so there's not much we can do to reduce the volume of JPEGs.
Well you could reduce the quality of the JPEGs. It's a lossy representation format after all :-)
That's a nice theory, but my client knows what they want in terms of image quality. It's my job to arrange the data transfer.
The pictures are actually of very high quality. We're talking about good DSLRs used for taking the pictures. But it's not the issue for me, I just have to do what my client wants whenever it's possible. I'm sure that they also have some significant Telstra 3G bills. But again it's not my issue, they are paying for the results that they want.
So to restate the facts you've given over the course of this thread: * You have a certain amount of data to transfer, and you are not able to reduce the size of this data. * The data is generated faster than can be uploaded over 3G, even including some buffering to local storage to allow for uploads during quieter periods. * The devices are often in remote areas with only a single 3G carrier (Telstra). You cannot load-share over multiple mobile networks. 4G will not be available either. The way I see it, you have to either reduce the size of the data or increase the capacity of the network link, but you seem to have ruled those options out. I really think you might have to give a little on the data reduction strategy. Have you considered using alternative image compression algorithms? ie. Instead of JPEG, look at fractal or wavelet compression? Those will be more CPU intensive, but reduce the final size, I think. Are your images fairly similar to each other? If so, could you use a video codec and treat them like film frames? x264 compression can maintain a pretty high level of quality with enough bandwidth.

On 02/07/13 15:59, Russell Coker wrote:
On Tue, 2 Jul 2013, Rasika Amarasiri <rasika.amarasiri@gmail.com> wrote:
Just thinking out of the box here. If the limitation is on the tower, would going with two different carriers work? Hopefully if they are not trunking together, you should get better throughput.
That's an interesting idea, but not viable in this case. Some of the devices I'm deploying will be in areas where only Telstra gives good reception. My general impression of the state of 3G in Australia is that Telstra will outperform all other options in most situations.
On average, yes, but in certain neighbourhoods people have all bought into Telstra and overloaded the local cells, and so carriers using Optus or Vodafone networks are better.

On 2/07/13 4:46 PM, Toby Corkindale wrote:
On average, yes, but in certain neighbourhoods people have all bought into Telstra and overloaded the local cells, and so carriers using Optus or Vodafone networks are better. As is the case in much of Bendigo. :)
-- 73 de Tony VK3JED/VK3IRL http://vkradio.com

On Tue, 2013-07-02 at 15:59 +1000, Russell Coker wrote:
On Tue, 2 Jul 2013, Rasika Amarasiri <rasika.amarasiri@gmail.com> wrote:
Just thinking out of the box here. If the limitation is on the tower, would going with two different carriers work? Hopefully if they are not trunking together, you should get better throughput.
That's an interesting idea, but not viable in this case. Some of the devices I'm deploying will be in areas where only Telstra gives good reception. My general impression of the state of 3G in Australia is that Telstra will outperform all other options in most situations.
I might be right off the mark, but here goes ... could a streaming application work better for the uploads? I was under the impression that streaming performed a sort of "on the fly" compression and may make for a better result. Andrew Greig

Hi Russell On Tue, Jul 2, 2013 at 5:59 AM, Russell Coker <russell@coker.com.au> wrote:
On Tue, 2 Jul 2013, Rasika Amarasiri <rasika.amarasiri@gmail.com> wrote:
Just thinking out of the box here. If the limitation is on the tower, would going with two different carriers work? Hopefully if they are not trunking together, you should get better throughput.
That's an interesting idea, but not viable in this case. Some of the devices I'm deploying will be in areas where only Telstra gives good reception. My general impression of the state of 3G in Australia is that Telstra will outperform all other options in most situations.
I you are deploying in regional areas (my assumption from you mentioning that only Telstra coverage is available) then you will find that LTE is only in metro and some major regional areas so LTE may not even be an option. One way to increase throughput in a marginal area is to use a 3G modem with the facility for an external antenna and then use a directional Yagi/beam over an omni directional antenna and aim it at the closest tower. -- Mark "Pockets" Clohesy Mob Phone: (+61) 406 417 877 Email: hiddensoul@twistedsouls.com G-Talk: mark.clohesy@gmail.com GNU/Linux..Linux Counter #457297 - "I would love to change the world, but they won't give me the source code" "Linux is user friendly...its just selective about who its friends are"

On 02/07/13 14:34, Russell Coker wrote:
A client has a problem with a device that generates data faster than it's 3G modem can upload it.
Is there a good LTE AKA 4G modem that can connect to a Linux system?
A lot of the USB-based ones work pretty well on modern Linux systems now; they present as a USB serial driver, accept modem-style AT commands, and then provide a ppp connection.
Can you run two 3G modems next to each other and expect that each of them will get full bandwidth? Or will they interfere with each other in some way?
You wouldn't get full bandwidth, but you'd be getting a bigger share than if you just had one. eg. Say there are 8 other active people in your cell, then instead of you getting 1/9th of the total, you'd get 2/10th (ie 1/5th). You'd be much better off with LTE though -- I get uploads speeds that are far more than just doubling of 3G and with a lot less packet loss.

On 02/07/13 16:44, Toby Corkindale wrote:
On 02/07/13 14:34, Russell Coker wrote:
A client has a problem with a device that generates data faster than it's 3G modem can upload it.
Is there a good LTE AKA 4G modem that can connect to a Linux system?
A lot of the USB-based ones work pretty well on modern Linux systems now; they present as a USB serial driver, accept modem-style AT commands, and then provide a ppp connection.
Can you run two 3G modems next to each other and expect that each of them will get full bandwidth? Or will they interfere with each other in some way?
You wouldn't get full bandwidth, but you'd be getting a bigger share than if you just had one. eg. Say there are 8 other active people in your cell, then instead of you getting 1/9th of the total, you'd get 2/10th (ie 1/5th).
You'd be much better off with LTE though -- I get uploads speeds that are far more than just doubling of 3G and with a lot less packet loss.
For eg, here's 12mbit upstream on Optus LTE in brunswick west: https://fbcdn-sphotos-e-a.akamaihd.net/hphotos-ak-prn1/32331_101515439027664... Same location gets around 1.5mbit upstream on Optus HSPA! (And 0.6 mbit on ADSL2+, damnit!)
participants (10)
-
Andrew Greig
-
Andrew McGlashan
-
Dave Hall
-
hannah commodore
-
Hiddensoul (Mark Clohesy)
-
Rasika Amarasiri
-
Russell Coker
-
Toby Corkindale
-
Tony Langdon
-
trentbuck@gmail.com