Can my ISP fake the numbers on speedtest.net ?

[adesh]

Hi,

 

After changing my ISP, I have found my internet speed to not be upto par with what I am paying for.

When I tested my download speed with speedtest.net it showed the download speed as 24-25Mbps (tested many times) but when I test it on testmy.net I get the download speed as 4-5Mbps

 

I feel the results shown by testmy.net are accurate because that's how much download speed I usually get when I download from chrome (tested downloading from s3 servers)

 

Is there a way for my ISP to give priority to speedtest just to show fake speeds to me ?

[nox_]

18 minutes ago, adesh said:

Hi,

 

After changing my ISP, I have found my internet speed to not be upto par with what I am paying for.

When I tested my download speed with speedtest.net it showed the download speed as 24-25Mbps (tested many times) but when I test it on testmy.net I get the download speed as 4-5Mbps

 

I feel the results shown by testmy.net are accurate because that's how much download speed I usually get when I download from chrome (tested downloading from s3 servers)

 

Is there a way for my ISP to give priority to speedtest just to show fake speeds to me ?

No, the ISP can't fake speedtest.net results. Try this one : https://fast.com what kind of speed does it show?

Edited June 23, 2019 by noxdeouroboros

[WereCatf]

8 minutes ago, noxdeouroboros said:

No, the ISP can't fake speedtest.net results. Try this one : https://fast.com what kind of speed does it show?

Actually, they can affect those results quite heavily. There are many ISPs that redirect speedtest.net - traffic to their own dataservers in order to boost the numbers.

12 minutes ago, adesh said:

Is there a way for my ISP to give priority to speedtest just to show fake speeds to me ?

Yes, see above.

 

13 minutes ago, adesh said:

but when I test it on testmy.net

The one speedtest that I, personally, trust and have found to be very accurate is http://www.dslreports.com/speedtest -- give that a try as well.

[nox_]

1 minute ago, WereCatf said:

Actually, they can affect those results quite heavily. There are many ISPs that redirect speedtest.net - traffic to their own dataservers in order to boost the numbers.

Never encountered that, but thanks for correcting me.

[brwainer]

Things that ISPs can do:

1. Host their own speedtest servers on speedtest.net, which the system may automatically choose for you based on their lower latency

2. Detect traffic to their or other speedtest servers and treat it as higher priority

3. Redirect traffic to/from speedtest servers that don’t use HTTPS to their own servers

 

I recommend dslreports.com/speedtest

[adesh]

Hi,

 

Thanks a lot guys for the information.

 

2 hours ago, noxdeouroboros said:

Try this one : https://fast.com what kind of speed does it show?

It showed me 18 Mbps

 

2 hours ago, WereCatf said:

The one speedtest that I, personally, trust and have found to be very accurate is http://www.dslreports.com/speedtest -- give that a try as well.

This one showed me 8Mbps

 

I guess the difference is because of location. Anyway, I am thinking of just changing my ISP so that I get a better idea of things.

 

Again, thanks everyone.

[sleerf]

There is quite a bit of incorrect and/or misleading on this post so let me clear it up for others that stumble onto this site.

 

First of all, if you have shared internet - which is what most of us get at home - then the provider isn't obligated to provide the speed you are paying for. The service is listed as "best effort". For all practical purposes, this is the speed that throttle is set at and nothing more. You have no guarantee of a bottom or even an average speed. It says that the connection is capable of X and that they will do their best to provide you with X but if they can't get it, it's not a service or quality problem. 

 

Now, given the competition in the marketplace, they do try to deliver that. But there is only so much bandwidth on a connection that is shared among hundreds of users connected to each aggregation point. When traffic gets heavy, connections bog for all shared customers. And occasionally, you'll click a page, get a time-out error, refresh, and all is fine. This is due to the freak occurance when everyone clicks at the same time and the pipe is completely full. Fortunately this happens infrequently and they try to maintain utilization of about 10-20:1 and 80%.

 

Incorporated into these systems is a feature that for the first few seconds of a request, the data rate has higher priority than most of the other traffic. This is a fairness issue where sustained high usage is throttled back somewhat to manage these shared networks and allow others who are using less to get faster speeds momentarily for what they need. 

 

The result however can mean that a speedtest when launched - which is a new connection to a new server - will initially skew to the high side provided the network itself , which is always bouncing in utilization, isn't bogged down. If it is, you'll experience a brief drag on the speedtest followed by a surge, and then when your initial connection priority times out it will begin to drop again.

 

Then you have the various hops between you and the speedtest server. Your packets are not traversing a direct path. They can take any route, different from the previous route, then have the same thing happen on return. There are any number of hops in between which is why selecting a server closer is important as it usually means fewer hops. But each hop is also a potential point of congestion. And as these routes change dynamically during the test, you really don't know where the problem is so  your first instinct is to blame the testing server or your carrier. 

 

Finally, you have the speedtest server which also is subject to the flow of traffic. 10,000 people hitting at once is drastically different from 10. They're connections can be impacted as well. 

 

The fact is, that the quality of shared internet service - which means you aren't buying dedicated internet for 10X the cost -  can't be determined by speedtests. If you run multiple tests over an extended period to multiple providers that are in different regions, you will have a better idea of what you're really getting. But unless you see a trend across multiple speedtest servers in geographically separate regions spanning a long period of time where speedtests average significantly lower than what you are paying for it's really not a service issue and shouldn't be a problem.

 

What is significant? Well, if you buy 100M x 10M and you see average speeds of 75M X6M, you don't have a service problem. If the next package sold is 50M x5M and you are consistently getting 45Mx3M, THEN you have a problem that the service departments will address. If you have 100M x 10M and you consistently get 60X6....it MAY be a service issue worth looking into. That will come down to how many tests are run and where. And they will all need to be close in the results. If one site shows a significantly lower test then it's likely that site. 

 

One more things. While a provider could technically fake a speedtest, it serves no purpose. They want to know when there are problems on the network so they can fix them before they become bigger problems. They want happy customers so they don't go elsewhere. With that said, they also get extremely frustrated at the weight most customers put into speedtests and the misinformation surrounding them. We've had people call in complaining about speeds only to find that they had a 10/100 switch in line, or their wifi was on the same channels as 50 others at an apartment building, or they had a bad cable, or jack, and countless other reasons which is why the policies exist that say that if we roll a truck and find it's not us, then you have to pay us. 

 

I really hope this helps people understand speedtests, what  they're good for and what they're not. 

[pythonmegapixel]

6 hours ago, sleerf said:

There is quite a bit of incorrect and/or misleading on this post so let me clear it up for others that stumble onto this site.

 

First of all, if you have shared internet - which is what most of us get at home - then the provider isn't obligated to provide the speed you are paying for. The service is listed as "best effort". For all practical purposes, this is the speed that throttle is set at and nothing more. You have no guarantee of a bottom or even an average speed. It says that the connection is capable of X and that they will do their best to provide you with X but if they can't get it, it's not a service or quality problem. 

This depends on what country OP is in.

 

In the UK, under new regulations, Ofcom (the communications regulator/watchdog) prescribes the advertised speed has to be achievable by (off the top of my head) a minimum of 50% of users, as at the busiest time. This effectively means that at less congested times you are able to actually exceed the advertised speeds. (Exact details may not be entirely accurate but it's something like that). Edit: This is not prescribed by Ofcom but by the Advertising Standards Authority

Anyway, the implication here is that if you are in the UK and pay for 100mbps up, 10 down and are consistently receiving noticably lower speeds than that at less busy times (assuming your testing prevents the weakest link from being anything other than the internet connection - i.e. you're not using wifi or a really outdated device) then they do have a service issue.

 

Also, while the insight you've given us into the delays caused by packet switching is correct, it's also worth noting that these delays are not specific to speed testing. Any data you send anywhere will encounter these delays, and I believe in the UK they're supposed to be factored into the advertised speed somehow (again - just off the top of my head, might not be correct). Surely this is - or at least should - also be the case in the US?

 

Quote

One more things. While a provider could technically fake a speedtest, it serves no purpose. They want to know when there are problems on the network so they can fix them before they become bigger problems. They want happy customers so they don't go elsewhere.

In the US, customers  can't really go elsewhere in many cases... they have a choice between either crap and crap (big cities) or just a single crap (other areas).

In the UK, we can change providers but most use the same lines, so if it's a customer service issue then we're better off, if it's a hardware issue then we're not.

 

I really can hardly blame US customers for relying on speedtests anyway - ISPs don't exactly have the best reputation over there!