From: Scott, Tyson C (tyson.scott@hp.com)
Date: Tue Jun 01 2004 - 20:49:56 GMT-3
Kenneth,
What you have stated is my understanding of Serialization.
You confirmed what I thought but I have one more question.
So if serialization is the amount of time it takes for a
packet to be transmitted out of the interface on to the line if you have
multiple PVC's coming out of the interface wouldn't you want all of the
PVC's be < or equal to the maximum serialization delay you want on the
interface.
For this example I have two PVC's and the interface rate is 1536. I
want to split that between the two VC's. So if I set the Bc on the
mission critical to be 7680 and max fram size to coincide at 960 bytes
would that also mean I would want to set the same on the second VC
because if I have it at something greater than 7680 (Both CIR's being
equal to each other) I am going to cause serialization delay for the
Critical RTP packets going out the mission critical circuit. Is this a
correct understanding?
Regards,
Tyson Scott
Agilent Problem Management Team
Managed Network Services
Phone: 313-583-5812
Pager: 877-997-0811
-----Original Message-----
From: Kenneth Wygand [mailto:KWygand@customonline.com]
Sent: Tuesday, June 01, 2004 7:39 PM
To: Scott, Tyson C; ccielab@groupstudy.com
Subject: RE: Question about QOS taken from IE LAB 7
Scott,
I'm not sure I understand all the details of the task you are trying to
solve. However, you must make sure you understand the difference
between _queuing_ delay and _serialization_ delay... Queuing delay is
the delay experienced sitting in the output queue when the Tx ring is
full. Serialization delay, however, is the time from the moment the
first bit is placed on the line until the final bit is placed on the
line of any given packet. Propegation delay is the time it takes an
electronic pulse (and thus a bit) to traverse the medium (which is
almost the speed of light which is negligible).
Serialization delay is a direct result of the size of the packet and the
speed of the link, where serialization_delay_in_seconds =
packet_size_in_bits / speed_in_bits_per_second. Ideally you want to
keep serialization delay at 10ms, so you set your Bc to be 1/100th of
your CIR (or access rate if you are not shaping/policing). So in your
case, the CIR is 768000, so your Bc should be 768000/10 = 7680. Of
course, you don't want a single packet to take up more than a single Bc,
so you fragment all packets down to this size.
If there is any more to your question that I can help you with, please
feel free to ask!
Thanks!
Ken
-----Original Message-----
From: nobody@groupstudy.com on behalf of Scott, Tyson C
Sent: Tue 6/1/2004 6:54 PM
To: ccielab@groupstudy.com
Cc:
Subject: Question about QOS taken from IE LAB 7
Group,
Quick question that is plaguing me on this scenario I
have
done and I am not sure I understand how Serialization occurs on
an
interface when it is done on a per VC basis. Serialization is
stating
the amount of time it takes the interface to queue the packet
and send
it out I believe. So with multiple classes applied on the
interface is
each queue serviced independently or are they all the same.
For example you are asked to minimize the
serialization of
VOIP traffic going from Router4 to Router5 and your CIR is
768000. To
minimize this I would set my Bc = 7680 and fragment any packets
going
out that are greater than 960 bytes since 7680 bits = 960 bytes.
There
is a second VC that goes from Router3 to ROUTER5 but with the
second VC
they don't specify that the Bc should be set to a minimum. In
the
solutions guide for this question it shows the Bc being set to
48384 but
wouldn't it also need to be set to 7680 and fragment anything
above 960
to meet the requirements of TASK 1 or is each queue serviced
independently and I am not understand the term serialization
delay?
Please help this is bugging me that I can't answer this question
to
myself?
This archive was generated by hypermail 2.1.4 : Sat Jul 03 2004 - 19:40:31 GMT-3