Latency VS Free Timeslots

(Read 7712 times)

Jun 29^th, 2011 at 11:47am

Hi,

I manage a iDirect network. Sometimes some satmodems have a high latency time.
In these timeranges all satmodem make a big IP traffic, but the Downstream total rate is lower than the Bandwidth MIR configured in QoS settings, In the same time I can't see a free timeslots in the Upstream Inroute Timeplan.
Is there a relation between no free timeslots and high latency time?

thanks

Jun 30^th, 2011 at 1:45am

Yes. You need timeslots to carry IP Payload. Example: A single timelsot at QPSK .793 (Large block) is capable of 394 Bytes....and there are approximately 78-82 slots associated with a 2048k QPSK/.793 upstream (2048k info rate...not IP Data rate).

On the flipside, small block coding (1k) is not nearly as efficient. But the answer to your question is - you need to have free slots in the network to move traffic off the remote.

The reason for the increase latency is due to your management traffic (remote --> NMS reporting) having to wait for a free slot to report into the NMS.

Jul 4^th, 2011 at 12:40pm

Are there some commands from the Falcon that can help me to performs some check?

Jul 4^th, 2011 at 3:31pm

You can view the burst timeplan and number of free timeslots from iMonitor.

Jul 4^th, 2011 at 5:16pm

Agree with dot. iMonitor gives you a decent (monitoring) graphical. Just right click the inroute group and view the timeplan.

Jul 13^th, 2011 at 12:56am

let me add few cents to this.

The NMS traffic is running in priority queue, which means that it should be processed before any other traffic, and when you see latency in iMonitor look odd, as you would expect NMS icmp rtt be very stable even if the link (network) is full

The main reason why you high latency when the inroute is full is a "sticky demand allocation" (not a sticky-CIR) - this reserves for the particular host the demanded amount of timeslots for 1 second, and therefore even high priority traffic may suffer.

You may disable sticky demand since version 8, but this should be done carefully.

As for timeslots usage monitoring, it is very poor - the only way is to observe it in iMonitor in real time and unfortunately it is not reported in SQL or SNMP and you cannot get historical data.

Jun 28^th, 2012 at 4:40pm

Hello,
I believe the relation should be like:
"The probability to experience high latency decreases with the availability of free slots".
I woud worry a bit if you had free slots available during high latency periods.
Make senses?

Jul 4^th, 2012 at 5:04pm

Hi,
the latency increases only when all the time slots are allocated.
Speaking with an iDirect technician, I understood that setting the Minimum Information Rate (Upstream QoS tab) = 1/2 slot rate, I can assign one slot to 2 satmodem.
If I well understand, this is a Multitrama allocation.
I do not understand what value I have to set.

Regards,

Jul 5^th, 2012 at 2:13pm

Hi,
I found half solution.

The MinIR is available in the remote's QoS tab.
There is a range, with a min and a max value.

Regarding the Max value (i.e.):
SR = 1352 Ksym
MOD = QPSK
Error Correction = 2D16S-170B (3/4):
• FEC payload = 170 B
• IP payload = 158 B
• 175 slots per frame

Frame length >> 125 msec/frame
Slots >> 175 slots/frame
0.7143 msec/slot

MAX = 158B/0.7143msec = 1769.6 Kb/s >> 221.2 KB/s >> (221.2 * 1000)/(8 * 158 ) = 175 slots/frame

Now I have some questions:

Is the MinIR a Static CIR? This means that the bandwidth not used is lost?

Is the CIR configured in the Group QoS tab (Inroute) a Dynamic CIR? This means that, if not required, the bandwidth may be allocated to other remotes?

Regards

Email me:eric@satsig.net

Disclaimer, Terms of Use and Privacy

Old Forum rules

Cookie policy.