Attaining better speed over EDGE/GPRS network with multi-slot class devices

Hi,

I have been using EDGE for sometimes here in Bangladesh with Siemens
MC75 EDGE card which claims to have a maximum bit rate of 474 kbps
compared to GPRS with 171 kbps. I've been using Nokia GPRS phone
before this EDGE card.

Yes, GSM can deliver 14.4 CS data on a single slot. Which turns out to
be maximum of 64 kbit/s. So, in real life scenario it should be often
3*14.4 kbit/s (43.2) or at most 4*14.4 kbit/s (57.6). As someone
mentioned in the newsgroup, Nokia GPRS phones, those are limited to 3
downlink CS slots in most of the networks - even if the model supported
more slots with GPRS. When it comes to downlink speed, it provides
maximum of 3*14.4 or 43.2 kbit/s. Since the data is checksummed, what
would be the actual payload?

As I understand, getting 474 kbps would be a dream when there's
bottleneck with Internet gateway at EDGE provider. Does that mean that
this speed will remain when I try connecting local networks, i.e.
corporate networks?

How would I know the number time-slots allotted to this device from my
EDGE provider? Can providers limit number of time slots to the devices
used? Is that true that providers configure their network to knock-off
the data service when voice calls are prioritized to use more
bandwidth? Shouldn't be they using some threshold levels for data
service?

Thanks in advance for any pointer.


--
Raqueeb Hassan
Bangladesh

"Raqueeb Hassan" <wideangle@gmail.com> wrote in message
news:1150297865.356668.204420@i40g2000cwc.googlegr oups.com...
> Hi,
>
> I have been using EDGE for sometimes here in Bangladesh with Siemens
> MC75 EDGE card which claims to have a maximum bit rate of 474 kbps
> compared to GPRS with 171 kbps. I've been using Nokia GPRS phone
> before this EDGE card.

Where did you find this claim?

Siemens' webpage
(http://www.siemens.com/index.jsp?sdc...d=26181128089&)
states the following:

"Thanks to EDGE, users can easily conduct their business activities on the
Internet or on the company intranet using their smartphone, PDA or laptop
PC - in fact, this can be done three times faster than when using GPRS; EDGE
theoretically has a maximum bit rate of 474 kbps compared to GPRS with 171
kbps."

(I don't know where they got the 474 kbps from...)

and further down "The main features of the MC75 include: ... EDGE (E-GPRS)
Multislot Class 10 ".

With simple calculation (and knowing that multislot class 10 means max 4.
slots downlink, max. 2 slots uplink, and maximum sum of timeslots is 5 at
any given moment), we get the following maximum speeds (with 8-PSK MCS9) for
this Siemens device :

downlink : 59,2 kbps/timeslot * 4 timeslots = 236,8 kbps
uplink : 59,2 kpbs/timeslot * 2 timeslots = 118,4 kbps

>
> Yes, GSM can deliver 14.4 CS data on a single slot. Which turns out to
> be maximum of 64 kbit/s. So, in real life scenario it should be often
> 3*14.4 kbit/s (43.2) or at most 4*14.4 kbit/s (57.6).

> As someone mentioned in the newsgroup, Nokia GPRS phones, those are
> limited to 3
> downlink CS slots in most of the networks - even if the model supported
> more slots with GPRS. When it comes to downlink speed, it provides
> maximum of 3*14.4 or 43.2 kbit/s. Since the data is checksummed, what
> would be the actual payload?

How come do you mix between CS data and GPRS/EGPRS? Here are the facts:
- most Nokia models suppot HSCSD (High-Speed Circuit-Switched Data), i.e.
multislot connections for CS data with multislot class 6 (max. 3 slots
downlink, max. 2 slots uplink, max. 4 simultaneous timeslots).
- Siemens MC75 only supports CS data with 1 timeslot
- Siemens MC75 supports EGPRS multislot class 10, and GPRS multislot class
12 (max 4 timeslots, max 4 uplink timeslots, with extended dynamic
allocation or fixed allocation - I don't know for sure which, but fixed
allocation is not supported by any live networks out there!)
- most Nokia phones support EGPRS and GPRS multislot class 10.

>
> As I understand, getting 474 kbps would be a dream when there's
> bottleneck with Internet gateway at EDGE provider. Does that mean that
> this speed will remain when I try connecting local networks, i.e.
> corporate networks?
>

Well, I don't know about the gateway but as described above, the maximum
bitrates achievable over the radio interface with EGPRS multislot class 10
are what were described above (downlink 236,8 bkps ; uplink 118,4 kbps).

Also, bear in mind that the maximum bitrates possible over the radio
interface are far and distant from what is the real case, especially when
looking from the application level and also depending on the distance to the
basestation and the amount of traffic in the network.

Actually, this point alone would merit for a lenghty discussion, but
typically e.g. throughput "over" TCP protocol depends on bandwidth, latency
and packet loss. When considering this, it is important to realise that even
with significant increase in bandwidth, the overall throughput gains are
moderate if the latency of the system is high - as it is in GPRS/EDGE.
Therefore, increasing bandwidth without improvements to (read: reduction of)
latency (or round-trip-time, if you will) is a game of diminishing returns.

Improvements to GPRS/EGPRS in terms of latency have been standardised, but
many of them require support from both the network and the mobile device. I
do not know whether the network you are using in Bangladesh supports these
enhancements (most importantly, the extended uplink TBF mode ; delayed
downlink TBF release is likely supported, as it works with all GPRS and
EGPRS -capable terminals), and also I can't say this for the Siemens card in
question (the data sheet states that the card is GSM release 99, which
suggest the extended uplink TBF mode is not supported as it is a GSM Rel-4
*) feature ; *) Rel-4 comes after R99). What I do know is that most
(relatively new, e.g. models introduced to the markets in 2005 or later)
Nokia EGPRS-capable devices do support these enhancements.

I'm not saying all this in order to bash or advertise anything, I'm just
trying to say that Internet rumors and speculations may be just that -
rumors and speculation.

> How would I know the number time-slots allotted to this device from my
> EDGE provider?

You don't, unless you have a network monitor or other engineering tools at
your disposal.

Besides, the amount of timeslots does not directly equal to the available
data throughput due to the fact that in GPRS/EGPRS the radio resources (i.e.
timeslots) can be shared between several mobile stations. Therefore, e.g.
having 2 timeslots for one mobile may actually result in faster connection
speed than having 4 timeslots allocated but shared with some other mobiles.

> Can providers limit number of time slots to the devices
> used?

Depends, but chances are that the radio network supports multislot class
operation (up to multislot class x), but depending on traffic conditions ,
the negotiated QoS parameters for the packet data traffic, and the quality
of the radio connection, the amount of allocated timeslots and used
(modulation &) coding scheme is continuously changed.

> Is that true that providers configure their network to knock-off
> the data service when voice calls are prioritized to use more
> bandwidth?

In most cases, yes. This is logical as users require certain (low) blocking
rates for CS connections - and would be very dissatisfied if more call
attempts would be blocked. Also, speech calls may bring more profit to the
operator than packet data services.

> Shouldn't be they using some threshold levels for data
> service?
>

Which thresholds are those?

> Thanks in advance for any pointer.
>

You're welcome.

>
> --
> Raqueeb Hassan
> Bangladesh
>

Antti

Thanks very much for your in depth analysis of my queries.

> How come do you mix between CS data and GPRS/EGPRS?

Yes, you are right. I did mix-up with CSD and EDGE the way they are
used. It was based on the calculation when I was using Cingular's CSD
service. Well, sometimes I wonder, who would use CSD (slow, uses
minutes) when EDGE provides better service? I think CSD is getting
phased out.


> > Can providers limit number of time slots to the devices
> > used?
>
> Depends, but chances are that the radio network supports multislot class
> operation (up to multislot class x), but depending on traffic conditions ,
> the negotiated QoS parameters for the packet data traffic, and the quality
> of the radio connection, the amount of allocated timeslots and used
> (modulation &) coding scheme is continuously changed.

So, you are saying that it depends on the ability of the devices used
with all other parameters as you have mentioned.

> In most cases, yes. This is logical as users require certain (low) blocking
> rates for CS connections - and would be very dissatisfied if more call
> attempts would be blocked. Also, speech calls may bring more profit to the
> operator than packet data services.

That's what I wanted to know. Thanks.

>
> > Shouldn't be they using some threshold levels for data
> > service?
> >
>
> Which thresholds are those?

I mean, providers should allocate some minimum possible of bandwidths
per user, rather than getting nothing at all - in busy days.

Thanks once again!


--
Raqueeb Hassan
Bangladesh

Raqueeb Hassan wrote:
>
> As I understand, getting 474 kbps would be a dream when there's
> bottleneck with Internet gateway at EDGE provider. Does that mean that
> this speed will remain when I try connecting local networks, i.e.
> corporate networks?
>
Sounds like somebody ( in marketing?) has noticed that the maximum data
rate available for EDGE within a single air interface timeslot is 59.2
kbps, and that there are 8 timeslots within the GSM frame. So 59.2
multiplied by 8 is pretty close to 474kbps.

This conventiantly ignores the facts :-

59.2 kbps data has pretty much no error correction, and needs a very
strong signal to be achieved.

Using all 8 timeslots on the downlink wouldn't leave any radio
resources on the uplink to transmit anything back to the network - a
highly improbable scenario in real life.