FCC Chooses SamKnows For Broadband Speed Testing > Not Real Data

said by pnh102:

ISPs will simply adjust connections to this testing site to make things look peachy keen.

There is plenty that the FCC can do to help consumers without resorting to wacky BS that wont' help anyone. The main one being cracking down on unfees that are not made known to the user until the first bill is received.

said by Ignite:

Hasn't happened in the UK. No way for ISPs to know which customers have the boxes installed and would need them to install DPI kit or implement QoS to prioritise.

It would be noted in the results if ISPs were playing games

said by pnh102:

And how would they know? If the ISP knows what the IP addresses of the testing servers are, all they have to do is prioritize all the traffic going to them. All of which is very easily done.

I don't know how they do this in the UK, but the FCC here already has plenty of enforcement and investigative powers at its disposal to help ensure that customers are cheated less by ISPs. If the FCC really wanted to improve the average US resident's broadband experience, it would use the tools already at its disposal to do so.

said by fiberguy:

And you don't think the ISPs can't get their hands on those devices??? if you don't believe that, then you're not trying or thinking very hard.

said by Ignite:

There is no evidence of this happening in the UK. I don't see why the US would be any different nor what the point of doing such would be. An ISP that goes out of their way to rig the results would be very obvious on the statistics in that they would be anomalous.

An ISP in the UK actually hired Samknows to assist them assessing their own performance.

I don't get why you think that all UK ISPs are apparently paragons of virtue while all US ISPs will frantically go out of their way to rig the results.

Still let's talk technically.

For cable QoS isn't going to work, there is no way to prioritise at DOCSIS level based on IP address it can only be done by using this to trigger a prioritised service flow which requires DOCSIS 1.1 or higher and would be obvious to a technically aware customer, their modem would report the service flow addition. The same applies to DSL that isn't going through an IP DSLAM and having PPPoE and ATM terminated there. A congested DSLAM backhaul isn't avoidable until it breaks out into IP further down the chain where the QoS can take effect.

This is only really of value for hiding core network congestion - which shouldn't be an issue anyway unless ISPs are idiots. Samknows can use multiply homed testing servers and even if need be extremely low TTL DNS entries along with multiple subnets to make life harder for ISPs that want to diddle with the transit prioritisation.

Really worth the ISPs time just to mitigate potential transit / peering when the vast majority of the issues on the networks are at access layer?

Nah.

said by fiberguy:

The majority of backups and slow downs in the U.S. on cable is the last mile, not the cloud.

Again, most of the slowdown, in cable at least, is in the last mile, not the core network. As for DSL, it's pretty good about delivering the speeds needed. When it comes to DSL, MY opinion as to the reason people don't get the speeds they are advertised has to do with distance from the CO or RT or the quality of the line. That in itself is not a national issue.. its an issue of the phone operators repairing any damaged twisted pair issue. ... don't really think there is an issue from the CO out to the cloud either.

It's generally a last mile issue - which again, is why I believe that the whole testing thing is silly anyway. All it's going to really measure is the last mile which is more of an issue of the ISP's, no, wait.. the "phone company and cable company" maintaining their local plants.

I haven't been too familiarized with entire cities complaining of slow performance.. just individual users or localized nodes, etc. When was the last time you hear "All of Chicago slowed to a crawl last night"...?