Broadband speed comparison

this is rather useless dont you think ?

Offcourse the ISP's that have the best peerings will have better speeds, even a ISP that has 50gbit more peerings than others can come dead last in this test!

If you want to have a fair runout place the facility at a IX where all the compared ISP's peer with.
 
pointless but interesting, we all know telkom is root of the problem and nothing can be done about it.
 
I say we talk with the DA that owns cape town and let them give cape town a fast cheap adsl solution :D rest of the country you voted ANC goodluck ;)
 
I can not help but to wonder about this article...

Hetzner is not an ISP and their SA hosting is based on the MTN Business(Verizon) backbone. But in the article they are referred to as separate networks....

"According to the Speedtest.net results Uninet has the best download speed amoung ISPs with 4.24 Mbps, followed by Hetzner with 4.15 Mbps and Verizon South Africa with 2.72 Mbps. The top 5 is completed by MTN with 2.41 Mbps and Telkom SA Ltd with 2.07 Mbps."

Am i missing something here or is there a distinct fishy smell in the air?
 
Why worry about speed?

Why worry about speed? We all know that our speed isnt the best in the world... What kills is me is that we have 3gig caps.... the other coutries dont even know what a cap is on a adsl line..

we need atleast a 10gig cap for the price of a one gig!
 
When schools/universities are included in the results the University of Potchefstroom reigns supreme at 7.67 Mbps, followed by two other universities which are not clearly identified at 5.32 Mbps and 4.29 Mbps respectively.
Click to expand...

staff writer, if you would like more accurate info on Tertiary Education institutes speeds then have a look at this site

UCT has a pipe of ~25MBps which (according to today's stats) has achieved an actual speed (albeit only a peak speed) of 27.5MBps, which is AFAIK the fastest (or one of the fastest) tertiary education pipes in SA...

I understand that the kernel of the article was assessing achieved speeds though results from web based speed tests (such as speedtest.net), however the link given above is, in a sense, a web based speed test (as it gives live, current, achieved speeds). :)
 
Sad

Well it is a sad state of affair in SA when compared to the rest of the world. Locally i think those Stats may be a bit off.

I am getting an average local download of above 5Mbps through a wireless internet service.

Internationally i am getting about 3Mbps.
 
rpm, please explain how these tests were done? when I connect to speedtest.net I see two test servers - Technology Concepts in JHB and Newtech Technology Services in CT.

Much obliged.

PDH
 
Hrm,

I read this article and have to say, I can't draw anything useful from its conclusions. Let's face some facts.

First of all, the speeds you get to the speed test site are largely dependent not only on YOUR connectivity, but the connectivity of the speed test site. Again, if the speed test site is badly peered thats going to skew the results. Add to the fact that the speed tests are also entirely load dependent on both the load on the line of the person doing the tests and the load on the site where the tests are hosted.

I'm looking for example at the university tests, mentioned, those are bizarre, because I can ASSURE you that those universities have far better bandwidths than are portrayed there, but that bandwidth is a.) heavily loaded and b.) I don't know which test site they were using so I cant check peering/capacity/latency.

The other thing to keep in mind is that those tests are to my knowledge run over TCP, that makes single session throughput HEAVILY latency dependent. With windows default tcp stack settings, you aren't going to get much better than 3mbit over international latency for example, where as a heavily tweaked stack is capable of giving you gigabit speeds in single session.

I would be FAR more interested in seeing an article on who peers with who and the inter-ISP connectivity / bandwidths / routing etc, as it would be far more useful in choosing a provider.

*shrug* just my 2c

Andrew
 
First of all, the speeds you get to the speed test site are largely dependent not only on YOUR connectivity,

My point - the test tests your link's speed - did rpm or whoever go to each ISP climb on their backbone and run a test to CT or NTS?

If not then how?

PDH
 
TheLoot said:
staff writer, if you would like more accurate info on Tertiary Education institutes speeds then have a look at this site

UCT has a pipe of ~25MBps which (according to today's stats) has achieved an actual speed (albeit only a peak speed) of 27.5MBps, which is AFAIK the fastest (or one of the fastest) tertiary education pipes in SA...

I understand that the kernel of the article was assessing achieved speeds though results from web based speed tests (such as speedtest.net), however the link given above is, in a sense, a web based speed test (as it gives live, current, achieved speeds). :)
Click to expand...

Just some more examples of just how skewed these tests are, I've just run some tests from a windows server that has a gigabit NIC into a router with 10gigabit of national peering capacity and 155mbit of international connectivity.

The results:

To the Cape Town server:
28ms latency, 0.47mbit downstream, 3.06mbit upstream.

To the Johannesburg server:
52ms latency, 6.11mbit down, 3.98mbit up

To the London server (maidenhead its labelled)
301ms latency, 4.57mbit down, 0.71mbit up

Then I ran the same tests from a linux box hosted in Johannesburg, with a 10gigabit NIC, into a router with 20gigabit of total peering bandwidth (10gig into JINX, 10gig via private peer), and 155meg of international connectivity. I also tweaked the tcp stack on the linux box slightly

To the Cape Town server:
37ms latency, 0.52mbit down, 3.09mbit up

To the Johannesburg server:
4ms latency, 10.53mbit down, 54.08mbit up

To the London server:
361ms latency, 9.61mbit down, 2.79mbit up

Now, these boxes have virtually identical amounts of theoretical bandwidth (more than their NIC's can carry), but the figures are startlingly different, based purely on location and peering.

I might also point out, that from that Johannesburg server if we do some other tests I see the following:

On an ISO download from ftp.is.co.za with wget, it completed at 8.97MBytes/second (71.76mbit/second). Again, thats pure peering and connectivity.

Basically, without knowing the speed of testing sites, and without knowing their peering/routing, the figures are meaningless!
 
Agreed. What exactly is the intention with this article in the first place now that we know:

1.) It's meaningless
2.) Results are inaccure
3.) It's meaningless.
 
Articles like those are here to keep up the click rate, clog our congested lines, waist our limited cap and generate revenue for MyBB...
 
durbandave said:
Well it is a sad state of affair in SA when compared to the rest of the world. Locally i think those Stats may be a bit off.

I am getting an average local download of above 5Mbps through a wireless internet service.

Internationally i am getting about 3Mbps.
Click to expand...

what isp you using?
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X