 joakoPremium
join:2000-09-07
/dev/null
kudos:5 | Comcast Loves Packet Manipulation PowerBoost really fudges the numbers. Comcast customers see above 100% of the advertised speeds *on speed tests*
--
PRescott7-2097 |
|
| Actually, they are pretty careful to take such shaping into account in the measurement:
The test is over 30 seconds of full rate, with three concurrent TCP streams. The bandwidth delivered is only measured in the last 5 seconds.
(There is also reporting on the first 5 seconds, which captures the boost effect). |
|
| reply to joako
said by joako:PowerBoost really fudges the numbers. Comcast customers see above 100% of the advertised speeds *on speed tests*
Read the report a bit more. That data is for sustained speeds and taken after the Powerboost effect. If you add PowerBoost back in they exceed Fiber as much as 150% (130%-200% over provisioned speed for upload)
--
"Too often we... enjoy the comfort of opinion without the discomfort of thought." - John F. Kennedy |
|
|
|
| reply to joako
I totally agree with you! When I saw Comcast over 100% I thought it's a joke! There is no way Comcast is delivering over 100%!! |
|
| said by Scuty :I totally agree with you! When I saw Comcast over 100% I thought it's a joke! There is no way Comcast is delivering over 100%!!
I think the facts show different and they are.
--
"Too often we... enjoy the comfort of opinion without the discomfort of thought." - John F. Kennedy |
|
plat2on1
join:2002-08-21
Hopewell Junction, NY | reply to joako
comcast customers see above 100% because they over-provision. standard 12/2 is actually 13.2/2.2. |
|
 jlivingoodPremium,VIP
join:2007-10-28
Philadelphia, PA
kudos:1 | reply to yt
said by yt:said by Scuty :I totally agree with you! When I saw Comcast over 100% I thought it's a joke! There is no way Comcast is delivering over 100%!!
I think the facts show different and they are. Indeed - the facts do show otherwise. These results ran tests on end points that SamKnows and the FCC controlled, to M-Labs servers. There were 1,109 test units in our network. The average peak period and 24-hour sustained (i.e. without PowerBoost) speeds met or exceeded advertised speeds - both upstream and downstream. This is on page 15 of the FCC report.
The numbers with PowerBoost are on page 20 and 21 of the report. One of our tiers achieved 152% of the advertised download speeds, and all were over 100%. In the upstream, one of the tiers delivers over 220%. On a related note, the recent ShaperProbe paper shows similar data, and demonstrates how regularly PowerBoost is delivered.
--
JL
Comcast |
|
 BHNtechXpertBHN StaffPremium,VIP
join:2006-02-16
Saint Petersburg, FL
kudos:82 | reply to nweaver
said by nweaver:Actually, they are pretty careful to take such shaping into account in the measurement:
The test is over 30 seconds of full rate, with three concurrent TCP streams. The bandwidth delivered is only measured in the last 5 seconds.
(There is also reporting on the first 5 seconds, which captures the boost effect).
Boost technology thresholds vary from provider to provider. It is IMPOSSIBLE for them to adjust for it and get meaningful data. The only way to work around it would be to test each time for an extended period of time attempt to detect the drop then start the benchmark...a very difficult task. Not only are the speeds skewed but also the consistency of service. In the case of consistency or QoS anyone using boost technology would have their connection QoS poorer than someone not using boost technology.
This is pretty much why this whole thing is waste of time because virtually all the majors are now using some form of boost technology on one or more tier offerings.
--
"I can’t give you a surefire formula for success, but I can give you a formula for failure: try to please everybody all the time."
~ Herbert Bayard Swope |
|
| reply to Scuty
Why not, since they hardly let their transfer any data? |
|
 KrKHeavy Artillery For The Little GuyPremium
join:2000-01-17
Tulsa, OK | reply to joako
That's the cool thing about this method of data collection.
It's not being faked out by caching or packet shaping. This is real time, long term data.
--
"Fascism should more properly be called corporatism because it is the merger of state and corporate power." -- Benito Mussolini
|
|
iansltx
join:2007-02-19
Golden, CO
kudos:2 | reply to nweaver
PowerBoost may still be in effect 30 seconds after a multistream download starts. After PB dies speeds are roughly as advertised *shrugs* |
|
iansltx
join:2007-02-19
Golden, CO
kudos:2 | reply to BHNtechXpert
Comcast, TWC and Cox use PB. Others don't. |
|
BHNtechXpertBHN StaffPremium,VIP
join:2006-02-16
Saint Petersburg, FL
kudos:82 | reply to iansltx
said by iansltx:PowerBoost may still be in effect 30 seconds after a multistream download starts. After PB dies speeds are roughly as advertised *shrugs*
It's not that simple my friend. I and many others have tried to work around PB in broadband benchmarking and it's extremely difficult and differs like I said earlier with each provider. It also tends to cause havoc with streaming and other applications where the streaming engine is inferior or intolerant of wide datarate fluctuations and/or slower or misconfigured PC's. Simply put PB is great for marketing and real crappy for real world performance where consistency of datarate is important.
--
"I can’t give you a surefire formula for success, but I can give you a formula for failure: try to please everybody all the time."
~ Herbert Bayard Swope |
|
| Comcast's implementation is a "leaky token bucket" shaper, which is over all flows on the user. Since the burst capacity is generally around ~20 MB of data, this should be gone after 10-15 seconds at the data rates we are talking about, so at least for Comcast the measurements are highly likely to exclude powerboost effects.
Additionally, this is data which, IIRC, is or will be made available, including packet traces, through Mlabs. If so, you can validate for yourself that the sustained data rate has dropped down, excluding the burst effects.
And this may make consistent datarate apps unhappy (OTOH, constant datarate apps need to adapt anyway because of cross-traffic), but it makes bursty-interactive apps, like web surfing, very happy. |
|
