Jitter and Relationship with Link Speed

krishna.vv · ‎04-30-2010

Hi Netpro champs,

Will the speed of a link effect the jitter?

I have an existing 2 Mbps circuit connecting two of my routers and am going to upgrade to a DS3 bandwidth. Will the jitter value come down due to increased bandwidth? Is there any reduction in the serialization delay when i move to a high speed link? Utilization is well below 70% in both 2 Mbps as well as DS3 scenario. Is there any relationship between Jitter and link speed.

Tx.

Kris

Giuseppe Larosa · ‎05-02-2010

Hello Kris,

>> Is there any reduction in the serialization delay when i move to a high speed link?

yes this is sure it comes from serialization delay definition, but this a constant given a link speed.

you need to take in account also propagation delay and queueing delay

The first is constant (unless you are dealing with aerial propagation), the second can vary widely over time.

Generally speaking, I would not see a direct relationship between link speed and delay variation.

You can expect some reduction of jitter, but once a bigger pipe is used IP traffic will increase going over the limits of old links.

For sure, you should use QoS tools to provide bounded jitter and delay to selected flows, like VoIP RTP streams, using a low latency queue

Only adding bandwidth does not solve the jitter problem and this is what is wise to note.

Hope to help

Giuseppe

lamav · ‎05-02-2010

Hello:

Serialization delay is measured using the following formula:

number of bits sent / speed of link

So, if you are increasing the speed of the link by a factor of 24, the serialization delay will decrease considerably.

Jitter is the difference in delay times between packets. For example, you can have phenomenal delays with no jitter, because the delay between packets is constant. Or you can also have minimal delays but a lot of differentiation in delay times, which means you will have a lot of jitter.

To illustrate further, if 3 consecutive packets of voice traffic consistently have a delay of 400 ms between them, the jitter is still zero, even though the delay is tremendous. If however, the first two packets experience only 10 ms of delay between them but then experience 100 ms between the second and third, the jitter will have increased, even though the overall delay is much lower.

So, increasing the bandwidth may very well decrease the delay, but not necessarily effect the jitter....sort of. Read on...

When you consider jitter, you want to think of the type of delay that may vary. Propagation delay will not vary, it is a constant. Electrical signals travel on a wire or optical medium at roughly the speed of light, more specifically 2.1 X 10 E 8 meters/sec.

Propagation delay is dtermined by using the formula:

length of link (meters) / 2.1 X 10 E 8 (meters/sec).

So, as long as the link length stays the same, the propagation delay will also be the same. As a rule of thumb, propagation delay is almost negligible anyway.

The variation in delay (aka jitter) is largely attributed to queueing and serialization delays. Queueing delay is the time a packet spends in a queue, waiting to be sent. It varies depending on the size of the packets in the queue and the speed of the link, which determines the serialization delay. Having decreased the serialization delay so drastically -- by increasing the bandwidth by a factor of 24 -- the queueing delays will be negligible and, so, too, will the jitter.

Take note that jitter will always exist -- it's just a matter of degree that we are concerned with.

HTH

Victor