Tools for testing FTTH service

Jungfla

I'm looking for recommendations in my search for tools suitable for testing a 1000/100 Mbit FTTH service. The "OOKLA" test gives erratic results, the one provided/hosted by the ISP gives better results fairly consistently.

As a crude test, I initiated an ftp session with heanet.ie and a 9.6Gb ISO file took 12 minutes at 0700 on a Saturday, suggesting an effective speed well below 1Gb/s

Reported speeds range between 90Mb (OOKLA) and 980Mb (ISP), ping times to Dublin servers vary from 4ms (according to the ISP) and 18ms (OOKLA).

I'm familiar with the techniques employed to test private or dark fibre, but how can I obtain reliable performance measurement from FTTH?

Johnboy1951

As I understand it, you are not going to get a 1Gb/s constant data throughput on a home fibre connection.
For that you need a commercial, uncontended, connection which has huge cost.

Add in that the server must also be capable of providing that constant data throughput, and that there are no delays in between you and the server.

Jungfla

I understand that, thank you for taking the trouble to respond.

To clarify, at times I see reported speeds of less than 10% that claimed for the service. I wouldn't expect a whole lot of contention at 0700 on a Saturday, and I'd be surprised if the HeaNet ftp servers even noticed the workload.

My connection appears to be 2 hops from the HEA network, hence I believe it's a useful indicator of performance.

I'm open to correction, and suggestions.

Johnboy1951

Did Heanet not have a page or two with real live stats showing a connections max throughput and how much was being used currently?

I have only vague recollections, and might be mis-remembering, but possibly worth checking out just to be sure.

Jungfla

That's jogged a memory - yes, you're right I think you can see their aggregate traffic via INEX. Must monitor that next time.

ED E

Jungfla wrote: »

I'm familiar with the techniques employed to test private or dark fibre, but how can I obtain reliable performance measurement from FTTH?

Performance measure of what?

Your package is 1000Mb. Your CDR is somewhere around 70-100Mb. Thats at the wholesale level. You can't measure Phy traffic, only the LT can do that.

Talking about two hops is a little misleading. OLO traffic is encapsulated to the handoff in most cases so IP only sees 1 hop where there might be 6.


From prior testing with Eir and Vodafone they seem to have provisioned peering to handle their GPON end users, I'm guessing you're with sky who've traditionally skimped here time and again?

Marlow

Also .. You can't just ditch Ookla's Speedtest as a whole.

It depends on what servers you test against. A lot of them may only have 1 Gbit/s themselves.

Known servers in Ireland, that have at least 10 Gbit/s uplink: Carlow (Blacknight), Galway (Airwire) and Sligo (Northwest Broadband). Blacknight and Airwire are on INEX directly.

So make sure to test against those. And always test to more than one.

You don't have access to any low layer that you can do more precise testing. And if your provider uses PPPoE and doesn't use baby-jumbos, you might already loose about 6% at that. So about 940 Mbit/s is the most you'll ever see. Not calculating for contention.

Also the public graphs you're referring to are 5 minute averages. That means, you would have to load the line for 10-15 minutes to even see a spike anywhere. What you want to achieve requires that you have access to your providers router infrastructure. It's a non-runner.

If you start to load another networks infrastructure constantly at that rate, their DDoS protection may also kick in and filter you off.

/M