1. the service is up to xxxbps/yyybps, not guaranteed 2. where do you deploy the QoS? on your pc? then you should count the network link headers, because the 384kbps is for the network speed instead of payload
comcast 6m/1.5m? you kidding, when I was using comcast, my downstream is 16M , but the upstream is only 1M. For your regular 6M downstream speed, your upstream is 384k.
I can even reach 30Mbps down, 1.5Mbps up for a short while (ten seconds or longer). The upload rate got cut back to 370Kbps after a while, the download rate got cut to below 10Mbps similar. Not sure what kind of device Comcast uses to meter and shape traffic.
16M
【在 z**r 的大作中提到】 : comcast 6m/1.5m? you kidding, when I was using comcast, my downstream is 16M : , but the upstream is only 1M. For your regular 6M downstream speed, your : upstream is 384k. : : 到1
z*r
6 楼
how did you measure the speed? from what you said, I can tell that you got influenced by the cache. let me guess how you measured the speed, I guess you used web browser? when you click downloading, you usually get a chance to select where to store the file, at this moment, the clock is not starting yet, but the background downloading has been started. That's why you could see a fast speed in the very beginning. In addtion, no matter what kind of shaping comcast used, it IS shaping. How shaping
【在 o*********y 的大作中提到】 : I can even reach 30Mbps down, 1.5Mbps up for a short while (ten seconds or : longer). The upload rate got cut back to 370Kbps after a while, the download : rate got cut to below 10Mbps similar. : Not sure what kind of device Comcast uses to meter and shape traffic. : : 16M
o*y
7 楼
Thanks for the explanation. Nope, I did not use web browser to test speed, though I know of such tools, e.g, speedtest.net. What I used was a simple program that sends UDP at a constant bit rate to a receiver; the receiver reports the measured throughput. I suspect Comcast uses devices similar to token bucket to police traffic from home users, and shape out-of-profile traffic at a hard limit, e.g., 6Mbps/370Kbps. Simply my guess, no solid evidence at all.
when the How won' between
【在 z**r 的大作中提到】 : how did you measure the speed? from what you said, I can tell that you got : influenced by the cache. : let me guess how you measured the speed, I guess you used web browser? when : you click downloading, you usually get a chance to select where to store the : file, at this moment, the clock is not starting yet, but the background : downloading has been started. That's why you could see a fast speed in the : very beginning. : In addtion, no matter what kind of shaping comcast used, it IS shaping. How : shaping
z*r
8 楼
you need to find a better way to test the speed. what you said doesn't make sense. read my previous post, you can see comcast uses shaping for downstream traffic, this is a must for any provider to limit the rate.
though I know of such tools, e.g, speedtest.net. What I used was a simple program that sends UDP at a from home users, and shape out-of-profile traffic at a hard limit, e.g., 6Mbps/370Kbps. Simply my guess, no solid evidence at all.
【在 o*********y 的大作中提到】 : Thanks for the explanation. Nope, I did not use web browser to test speed, though I know of such tools, e.g, speedtest.net. What I used was a simple program that sends UDP at a : constant bit rate to a receiver; the receiver reports the measured : throughput. : I suspect Comcast uses devices similar to token bucket to police traffic from home users, and shape out-of-profile traffic at a hard limit, e.g., 6Mbps/370Kbps. Simply my guess, no solid evidence at all. : : when : the : How : won' : between
o*y
9 楼
Shaping is certainly used, but is it for per-user or per-group? Is it the same for home user or for business accounts? Will there be any interaction between traffic from different users? Except for Comcast engineers, probably seldom anyone else knows the details of how they enforce bandwidth. I was trying to apply some "black-box" testings, but obviously it cannot give me enough information. I was hoping to get some more precise information from people who know cable technology very well, or fro
【在 z**r 的大作中提到】 : you need to find a better way to test the speed. what you said doesn't make : sense. : read my previous post, you can see comcast uses shaping for downstream : traffic, this is a must for any provider to limit the rate. : : though I know of such tools, e.g, speedtest.net. What I used was a simple : program that sends UDP at a : from home users, and shape out-of-profile traffic at a hard limit, e.g., : 6Mbps/370Kbps. Simply my guess, no solid evidence at all.
z*r
10 楼
you don't have to be a comcast employee to get all this info, there is a term called standard. I agree each company may treak a little bit, but basically they follow the same or similar standard. usually it's per-user basis, not only shaping, also guaranteed bandwidth. That's why queuing function is so important for access equipment. It's very popular that an access interface is capable of thousands of queues. in addition, the DOCSIS has lots of QoS implementation as well, eg for priorities... I
【在 o*********y 的大作中提到】 : Shaping is certainly used, but is it for per-user or per-group? Is it the : same for home user or for business accounts? Will there be any interaction : between traffic from different users? : Except for Comcast engineers, probably seldom anyone else knows the details : of how they enforce bandwidth. I was trying to apply some "black-box" testings, but : obviously it cannot give me enough information. I was hoping to get some : more precise information from people who know cable technology very well, or : fro
o*y
11 楼
This is very helpful. Thanks a lot!
very this
【在 z**r 的大作中提到】 : you don't have to be a comcast employee to get all this info, there is a : term called standard. I agree each company may treak a little bit, but : basically they follow the same or similar standard. : usually it's per-user basis, not only shaping, also guaranteed bandwidth. : That's why queuing function is so important for access equipment. It's very : popular that an access interface is capable of thousands of queues. : in addition, the DOCSIS has lots of QoS implementation as well, eg for : priorities... I