WA and incorrect allocation of capped bandwidth

I have a capped account with WA and noticed that a few days ago - according to the logs they provide that 4 gigs was DL between 11:00 and 12:00. As I have a 6 meg line, my max DL is about 2.2. gigs per hour

Anyway I did post this on the WA forum and had no response as well as take it up with support. After chasing support a few times I was given an answer that "all ISP's use this system"; "part of the bandwidth could have been allocated to the 1/2 before" (interestingly enough from 10:00 to 11:00 my dl was 2 gigs so close to the max).

At the end of the day I have been told all is in order and systems are fine, so no problem, even though this has happened twice this month

For me 2 gigs is not end of the world stuff and now that there are problems with international connectivity I would not be able to use the bandwidth anyway. My issue is the principle. Capped user pay for a certain allocation of bandwidth and should be able to use it all if they so choose. If the system is in anyway not recording something correctly, then the service provider should take the fall, not the customer.

For uncapped users this is not an issue (except I guess with WA and rolling windows); for capped users it is an issue of principle as we are not getting all we pay for.

Rant over

I was under the impression this was being sorted out here http://forums.webafrica.co.za/showt...age-of-cap-by-WA?p=67706&viewfull=1#post67706
Not sure why another thread, on another forum needs to be created?

I'll have things looked at on Monday, thoroughly. As I said in your original thread on our forums - we'll do what we can.
Would love to do an investigation on this now and throw some ideas and fixes around with the devs, but we're all a little occupied with the SAT3 issues at the moment.

On the 1st of this month I apparently used 508.71 MB during 00:00 and 01:00. I'm on 1mbps home uncapped (so max 370MB/hour), so it doesn't really bother me, but I've noticed it happen on more than one occasion.

WAJeff said:

I was under the impression this was being sorted out here http://forums.webafrica.co.za/showt...age-of-cap-by-WA?p=67706&viewfull=1#post67706
Not sure why another thread, on another forum needs to be created?

I'll have things looked at on Monday, thoroughly. As I said in your original thread on our forums - we'll do what we can.
Would love to do an investigation on this now and throw some ideas and fixes around with the devs, but we're all a little occupied with the SAT3 issues at the moment.

Click to expand...

I logged a call in May with a similar problem, I recieved a reply to say it was escalated to the developers. It is now the 29 June, I never had a reply since then. I dont use your data anymore so the point is moot since I moved to Mweb, but thought I would let you know that the "unparalleled level of support" your website claims is not consistent with reality.

This was one of the reasons I left WA originally... the sudden inexplicable mass disappearance of capped bandwidth. All they could tell me was that it was definitely downloaded by me. Right.
This capped bandwidth measurement model is horribly broken and no-one wants to admit to that or fix it.

Can the guys who are having these issues please PM me with their client codes and ticket reference numbers so I can have this looked into on Monday please.

Will see if I can get things sorted for everyone.

WAJeff said:

I was under the impression this was being sorted out here http://forums.webafrica.co.za/showt...age-of-cap-by-WA?p=67706&viewfull=1#post67706
Not sure why another thread, on another forum needs to be created?

I'll have things looked at on Monday, thoroughly. As I said in your original thread on our forums - we'll do what we can.
Would love to do an investigation on this now and throw some ideas and fixes around with the devs, but we're all a little occupied with the SAT3 issues at the moment.

Click to expand...

Nope not sorted, and as I stated in the first post, I was not satisfied with the "conclusion" to this matter.

Why start a new thread on this forum - sadly sometimes this is the only way to get a matter resolved. In addition it has a much bigger readership and hence seems to be taken more seriously.

Interestingly looking at the responses above it seems to be a relevant issue that WA has not taken too seriously (I have had this issue quite a few times over the years)

Thanks though for looking into on Monday as I do understand you have issues on the go at the moment.

I barely read the thing. I don't work for WA. But I do know the answer.

Telkom send an interim every hour, this hour is not on the hour (could be any time during the hour). These interims are also UDP packets without guaranteed delivery.

That interim contains the total of your usage over the hour, if the packet doesn't survive the trip, it could be 2 hours worth, or 3 or 4 etc...

Turning off your modem makes things work because they also send accounting info when a session stops.

WA would need to process the information when it arrives, I am not sure which hours count here, but lets say whatever cut off time your talking about is 6am, if Telkom send the interim at 5:01 the next interim will be 6:01.. If you downloaded 500MB between 5:01 and 6:01.. theoretically that's 6am+ usage. (There are ways to deal with some of this in code, there's also very expensive and cunning tricks to deal with it).

--------------------------------------------------------------------
Now here is the actual crux of the matter... everything above the line is actually Telkom's fault. It's my personal opinion that this was done to ensure over usage on their original reseller model of pay-per-gig thus making more profits.
--------------------------------------------------------------------

To summarise, every HOUR Telkom sends a little UDP packet which MIGHT make it to the ISP containing your usage. < If you can not see a small room for error there then I don't know. It sucks ass and the ISP's have to deal with the fallout.

So 'sorting' an issue out basically entails telling someone 'I'll look into it in a few days?' No wonder he;s making more threads elsewhere.

Fulton, did they refund you the lost bandwidth?

encrypted said:

I barely read the thing. I don't work for WA. But I do know the answer.

Click to expand...

Thanks for your "answer" but having not read the post, the answer is not really all that helpful. My issue is with the fact that the figures given as to bandwidth used, were simply impossible for my line. My issue is with the fact that I am told I have used bandwidth when it is impossible to use that much in an hour. Thanks though for your explanation of timings

Graal said:

So 'sorting' an issue out basically entails telling someone 'I'll look into it in a few days?' No wonder he;s making more threads elsewhere.

Fulton, did they refund you the lost bandwidth?

Click to expand...

Graal, my email response was basically everything is right, so there is no problem. Trust you are happy!! I think the next comments were made because I then started this thread etc.

Nope they did not. I think at this stage of the month, my issue is more the principle. The right thing would have been to say - yep there is a problem and here is some cap to make up for it.

Anyway let's see what comes up on Monday

Fulton I reported a similar problem 6 weeks ago. I've since looked into the problem in a bit more detail and concluded that their story checks out fine - essentially its what encrypted says above.

You can actually see the delay in your usage logs. Note the missing hours, keeping in mind that 144mb is the max you can download on a 384k line per hour:



Its not related to D+...happens at all times of the day and shows up in usage logs even before D+ was introduced. Sometimes it works to your advantage and sometimes to your disadvantage.

The problem is fairly pervasive unfortunately as can be see from the last 2 years of my usage logs:


@Jeff: As a solution I'd suggest WA implement the following:

I can see the delay in my usage logs, therefore WA can too. Where there is a gap in the logs (see first image above), the total in the first hour after the gap should be split evenly over the hours in question. As a gesture of good will I'd also suggest that where such a gap spans both normal and D+ plus hours, the total be only allocated to D+. This ensures that where there is a problem with the SAIX updates, it will always be in the customers favour. Sure, that will cause WA losing out on a couple of gigs here & there but that is much better than the horrible PR this issue is currently causing. I'd also suggest creating a dedicated KB article for this & highlighting it to support staff so that they are aware of the issue and can respond accordingly since the way my support ticket regarding this was handled was somewhat unsatisfactory (Closed without meaningful feedback).

Thanks Havoc for your detailed response. Sorry but not sure how to do those PNG's. For me on my 6 meg line my max is about 2.2 gigs per hour

Below is the one time in question, so is different to what you are mentioning I think as this was one long download. As you can see I had not used my WA account since 25th; connected 27th and did a download for a few hours.
2012 6 25 15 54 3.33
2012 6 27 10 1981.34 30.5
2012 6 27 11 4049.06 70.08
2012 6 27 12 1663.9 30.71
2012 6 27 13 33.78 0.87

Here is another example in the same month (but during DL+ hours)
2012 6 4 4 2251.46 42.28
2012 6 4 5 4086.89 75.25
2012 6 4 6 1218.86 22.13
2012 6 4 23 106.52 2.17

In this case I would have switched off my router before 7 as I go to work.

I think the DL+ one did not bother me as much as there is a good amount of cap available; but during my normal cap which is limited, I do feel much more strongly about the issue.

Maybe at the end of the day I have to be more philosophical about it all and say you win some, and you loose some I do just think it is a matter of principle and these little amounts add up over time.

This problem dates back to the free 10GB trial exercise that WA launched after implementing their own network. My trial was prematurely suspended at 6GB, supposedly to have used up all 10GB.

After analysing the log files, I have pointed out that I would need at least three times my actual speed to be able to use the bandwidth indicated for certain time intervals. I can remember a few other testers reporting similar experiences.

On the WA forum, promises were made to investigate and correct the matter, which never happened. My guess is that they still have this 'buggy' bit-meter in place.

Still discussing this with our Devs and Product Teams.
I'll have feedback soon.

DawieS said:

This problem dates back to the free 10GB trial exercise that WA launched after implementing their own network.

Click to expand...

Yeah I vaguely recall commenting on something like that during the trial too.

@Jeff: While the devs are fiddling with the usage system the following changes would be awesome too:
Complete split in the detailed usage file (Freezone etc).
An actual excel file or comma serparated values (CSV) for the detailed usage. Some wiseguy took an .html file and simply renamed it to .xls. Excel works around this by opening it as an HTML file, but then you guys might as well just open it in the browser window if its HTML anyway.

Fulton17 said:

Sorry but not sure how to do those PNG's.

Click to expand...

That actually required quite a bit of black magic to get those right due to the above ".xls" issue.

Fulton17 said:

I do just think it is a matter of principle and these little amounts add up over time.

Click to expand...

Certainly. The usage tracker should be 100%. I've crunched the numbers though & I can tell you that even if all of it is unfavourable its still a fairly modest amount especially for people on RedAfrica.

I've looked at the logs some more: The gaps in the hours aren't reliable either. In some cases there are low usage hours followed by a high usage. Presumably the SAIX/WA system missed a bit of data & then just jammed it into a subsequent hour. That would not be fixed by the "solution" I proposed above.

HavocXphere said:

Certainly. The usage tracker should be 100%. I've crunched the numbers though & I can tell you that even if all of it is unfavourable its still a fairly modest amount especially for people on RedAfrica.

Click to expand...

Fulton17 said:

I think the DL+ one did not bother me as much as there is a good amount of cap available; but during my normal cap which is limited, I do feel much more strongly about the issue.

Maybe at the end of the day I have to be more philosophical about it all and say you win some, and you loose some I do just think it is a matter of principle and these little amounts add up over time.

Click to expand...

I hear you Havoc - especially during DL+ hours. I think initally it did not bother me until it was impacting my normal cap which is far less than the after hours cap.

Anyway it seems like there is attention been given to it by various parties so hopefully things will get sorted out once and for all, and the matter can be put to bed

WAJeff said:

I'll have things looked at on Monday, thoroughly. As I said in your original thread on our forums - we'll do what we can.

Click to expand...

WAJeff said:

Can the guys who are having these issues please PM me with their client codes and ticket reference numbers so I can have this looked into on Monday please.

Will see if I can get things sorted for everyone.

Click to expand...

WAJeff said:

Still discussing this with our Devs and Product Teams.
I'll have feedback soon.

Click to expand...

Hi Jeff.
Before this slips off the radar, please can you give us an update and when the matter will hopefully be resolved and closed? Thanks

No where near "slipping off the radar"
Hoping for a resolution soon, we're going through logs etc on our end.

Will post up feedback when I have it.

Hey guys,

OK, so we've opened up an investigation on our side.. can you please include usage times when you have noticed the incorrect usage come through.
Thanks

encrypted said:

I barely read the thing. I don't work for WA. But I do know the answer.

Telkom send an interim every hour, this hour is not on the hour (could be any time during the hour). These interims are also UDP packets without guaranteed delivery.

Click to expand...

It's a bit unfair to characterize RADIUS as pure, un-adapted UDP. While RADIUS is UDP, RADIUS will typically retry. While DIAMETER (which is TCP-based) may have a more reliable transport, connections can still be interrupted (just as RADIUS packets can be lost). In both cases, the application is responsible to retrying, and nothing guarantees that you will always get all updates.

That interim contains the total of your usage over the hour, if the packet doesn't survive the trip, it could be 2 hours worth, or 3 or 4 etc...

Click to expand...

But, in this case, the previous interim made it ...

Turning off your modem makes things work because they also send accounting info when a session stops.

WA would need to process the information when it arrives, I am not sure which hours count here, but lets say whatever cut off time your talking about is 6am, if Telkom send the interim at 5:01 the next interim will be 6:01.. If you downloaded 500MB between 5:01 and 6:01.. theoretically that's 6am+ usage. (There are ways to deal with some of this in code, there's also very expensive and cunning tricks to deal with it).

Click to expand...

Yes, you could pro-rate the usage for any interim that covers multiple rating intervals.

Now here is the actual crux of the matter... everything above the line is actually Telkom's fault.

Click to expand...

I thought WA was on IPC? If they are, everything is not Telkom's fault. If this account is on SAIX Usage, well, then, why isn't WA using IPC here?

Also, remember that other changes were forced upon Telkom by ICASA due to complaints by users who were trying to effectively bully Telkom into providing uncapped and fixed IP addresses back in 2005 or 2006. The changes in the regulations prevent an ISP from disconnecting a user. In WA's case, if they were to limit the session time to end just before their free-rated time window, the ESR would terminate the PPPoE session, the modem would re-connect, and WA would get nice stops and starts to be able to perfectly rate their different intervals. But, they are not allowed to do that, because everyone lobbied for never-ending PPPoE sessions for all users, even those who would derive no benefit from them.

However, WA probably could require that customers on this plan opt in to being disconnected at the end of the rating interval, or accept that accounting will not be 100% accurate. As long as they don't have too many customers on this plan to PoD, it would work.

It's my personal opinion that this was done to ensure over usage on their original reseller model of pay-per-gig thus making more profits.

Click to expand...

This wasn't done intentionally. AFAIK, the ESRs don't have sufficient processing capacity for more frequent interims.

To summarise, every HOUR Telkom sends a little UDP packet which MIGHT make it to the ISP containing your usage. < If you can not see a small room for error there then I don't know. It sucks ass and the ISP's have to deal with the fallout.

Click to expand...

It's the same kind of packet that is sent when you connect or disconnect. Yes, not all packets make it, but if they aren't ack'ed, they are re-transmitted, and if they still aren't ack'ed, they are sent to the ISPs listed secondary RADIUS IP.

Sorry, but people knock Telkom all the time, and Telkom Internet's current setup (while we are currently not permitted to use IPC) is much fairer here.

Unfortunately, it is quite complex to ensure exact and always up-to-date usage tracking without change-of-authorization and never-ending sessions spanning accounting and billing cycles, but WA has options they are not using, so while there are improvements Telkom could make (once the ESRs are upgraded), WA honestly can't blame Telkom when other ISPs don't have these problems.