I have a 6x2x2x2 network that has moderate bandwidth utilization. Peak traffic shows utiliztion of about 70% on the downstream, and 40% on the upstream.

Despite these numbers, I'm seeing an abnormally large number of time slots being used. Of course, someone made mention of this to the wrong people, and now it has become a hot topic, even though there seems to be zero service degredation.

I need a simple explanation of the correlation between time slots and bandwidth utilization. The iDirect website has proven fruitless on this one. Any help would be appreciated.

The number of timeslots available is related to the TPC FEC block size. What is the modulation and coding as well as size in information rate of your inroutes?

A further question to consider is : What is the typical length of a customer message (on the inroute) ?
If customer messages are all very short (e.g. mouse click on url) and the TDMA burst length is 5 times longer, then the efficiency in filling up the burst length with meaningfiul data will be 20%.  If the typical customer data is big files being uploaded then multiple bursts will be needed to send just one customer data file and the fill factor (burst content efficiency) will be near 100%.
In a real network there is a distribution of small, medium and long customer messages and the burst length will need to be chosen as a compromise, so the question could be improved by saying: What is the probability distribution of your incoming customer messages.

Best regards, Eric

and another question, what OS version are you running on?

you can try "pad stats" command on the modem to see the filling efficiency, that may give you an understanding on how well your timeslots are being used

Let me see if I can answer all your questions here.

I am still running 8.2.2.0. Yes, I know it's old.

The inroute is QPSK .79 with a FEC block size of 512. That should yield 75 timeslots.

Eric, is there a way I can monitor the length of a typical customer message? Also, where do I find my burst length? (I should know that one, but it escapes me now. I'm kind of new to this whole TDMA thing.)

Thanks!

You need an answer from an iDirect user.  I have no experience or expert knowledge of iDirect.

If a burst comprises one FEC block of 512 information bits the actual transmitted burst will comprise: a header comprising a carrier and bit timing recovery sequence, a unique word and then the content block of FEC encoded data (e.g. 683 bits representing 512 bits of user data at FEC ratio of 0.75). There may be a checksum byte at the end.  Such an example burst would carry 64 bytes of user data.

Clarify if you are talking about bits or bytes !.

TDMA bursts may be of a wide variety of lengths.

I've just accessed the internet and requested a forum page.  I sent: 66, 66, 54, 718, 60, 54, 54, 66, 54, 54, 54, 54 bytes, according to wireshark.

Best regards, Eric.  

You would use iNPT to calculate exactly what you have. If I plug a 2 Mbps IP data rate carrier in with QPSK TPC 0.793 FEC and a frame length of about 120 ms it shows me that you have 78 timeslots of 25.96 Kbps each. This is a carrier of 3.58 Mbps information rate.

The biggest problem you most likely have here is that you are using QPSK 0.793 which has a block length of 4K while changing to QPSK 0.66 which has a block length of 1K would yield many more smaller timeslots and may be more effective for your traffic profile.

RidingDonkeys wrote on Apr 10^th, 2011 at 9:39am:

I have a 6x2x2x2 network that has moderate bandwidth utilization. Peak traffic shows utiliztion of about 70% on the downstream, and 40% on the upstream. Despite these numbers, I'm seeing an abnormally large number of time slots being used. Of course, someone made mention of this to the wrong people, and now it has become a hot topic, even though there seems to be zero service degredation. I need a simple explanation of the correlation between time slots and bandwidth utilization. The iDirect website has proven fruitless on this one. Any help would be appreciated.

OK, so Dot makes some very good points, typically upstream traffic profiles better suit a smaller block carrier, the efficiencies of using the large block (.793/4K) are lost if your slot allocation means the remotes are not filling the available bandwidth.

you can measure the traffic leaving the PP blade and hitting  the upstream router, capture this in wireshark/ Network Observer (free/paid) and you will be able to get an idea of the traffic profile.

So the next question is from this "Despite these numbers, I'm seeing an abnormally large number of time slots being used."  
Where are you seeing this?  what is telling you abnormally large slot allocation is occurring?

RidingDonkeys, did you try "pad stats" command, on the version 8 it might be "qos pad stats"or something similar