GPRS DataTransfer Average Speed

Thanks OnixExp!!

That´s exactly the values I'm getting during my tests. I also find out that when I'm exchanging data inside the GPRS network (i.e. between two terminals connected to the GPRS service) the delay is increased when compare to the values I get while sending and receiving packets from the Internet (outside GPRS network).

I was not able to find a very good explanation for that (maybe too much "Over the Air" phases???).

Anyway, here are my data:

Accessing Internet: 700ms to 1500ms roundtrip;
Between GPRS Terminals: 3000 to 4000ms roundtrip :(


Thanks again,

Ric

P.S. I didn't wait for another answer to this question before accepted for it was posted more than a week ago, and this answer brings me what I was expecting!

>I was not able to find a very good explanation for that (maybe too much "Over the Air" phases???).

Well, this is probably close to the answer. Logically, there will be 4 over the air phases instead of 2 if you go from GPRS -> GPRS instead of GPRS -> Internet.

So then you have to start by doubling your roundtrip values: (700 to 1500) * 2 = 1400 to 3000
You say you're getting an extra 1000 to 1500 ms roundtrip on top of that, so how to explain that.

One think I can think of is that they could be interfering with eachother if they are in the same cell. Or maybe they have additional delays in acquiring their uplink channel because the reply is taking so long to return. Finally, there could be something in the GPRS core network that is adding delays due to a possible routing issue between GPRS terminals. Especially if the GPRS terminals are connected to different operators or something.

It's difficult to say. The technology is so complex that there could be many many problems. In the project where I'm working we had strange troubles with a GPRS that was probably related to the GPRS wireless pack hardware because after a firmware upgrade suddenly things improved by like 20%.

"One think I can think of is that they could be interfering with eachother if they are in the same cell."..

The was my first thought. But I run the tests from different points around the city and there were no changes.

I've tried the "tracert" app to find out more about the route took by the packets:

1) While Connecting to the Internet:

10.0.0.10 - MS
10.0.0.1 - ??Router??
10.200.0.50 - ??Router??
200.193.0.1 - Outside GPRS core
...
200.50.1.2 - Internet target

2) While connecting to another MS

10.0.0.10 - MS1
10.0.0.1 - ??Router??
10.200.0.50 - ??Router??
10.200.0.51 - ??????????
10.200.0.50 - ??Router??
10.0.0.1 - ??Router??
10.0.0.200 - MS2

I'm not sure about the last three steps for I don't have the system installed right now. But I was wondering what could be the "10.200.0.51" node for I just reach it while connecting to another GPRS mobile. Maybe an Internal firewall (I also noticed that my IP ports are changed whitin the packet).

Regards,

Ric

If you can use tracert, doesn't it show the RTT to each machine? (so you could see where the biggest part of the delay is at).

The only thing for .51 that I can think of is maybe that it's a core network border router (they are checking to see if it needs to go to a peering core network maybe). But the route seems a bit silly going through the same IP's 2 times on one journey..

As I mentioned I'm not sure about the route...I do remember they always reach the 10.200.0.51 node ( only during intra-gprs communication) and there is where the biggest delay is at.

The other problem I faced is that somehow I can't reach another terminal sometimes, i.e, after the two of them get connected to the GPRS network I can't even ping each other. The solution is to disconnect one of them, reconnect (more than once sometimes) and then they can be reached by the other terminal. Weird!!

Regards,

Ric

Well, I was surprised you were even able to ping from one terminal to the other at all. After all, you're getting a private IP address and what guarantees that both GPRS terminals get hooked up to the same LAN.. Obviously if the operator has only one core network that all the traffic across the country is routed to then implicitly all the phones will be on the same LAN, but I don't know how to be sure of that. And maybe the strange delays are coming from the fact that the operator didn't count on people trying this and never tested it or something.

In the past I spoke to network operator and I was told that there were only one core network and every attempt to connect to the external IP (internet) network was done thourgh only one GGSN. However I believe they didn't count on this type of data transfer for there are lots of people (from the operator or any other representative of them) that are not aware of this possibility.

I was also told that maybe it would be possible to config. an APN so that I could have a group MS attached to a special sub-network and then able to reach each other. My concern is still the speed/delay. Would this special config. speed up the data transfer between them??

As far as I know the GPRS specs takes into account this type of communication. Maybe this is an option left to the operator of manufacture.

Ric

P.S. Apparently they are using a Nokia plataform for the GPRS network.