General : Oversubscribed areas?

[Chrysalis]

qas is right, in the url I linked to and ignition also confirmed on a tbb post that lib global still use the VM escalation thresholds, so it didnt even get approved as a capacity problem until the congestion was high enough, when the customer first raised it on the forums he had clear congestion but the support staff couldnt raise it due to it been below the threshold.

This is why I said if they need a year or so to fix the problem, the work needs to start at a much lower threshold.

[one2escape]

My oversubscribed area has made me leave now. Being going on 9 months and the date to get fixed keeps on getting put back. My amazing speed for 152mb!

[Kushan]

That's disgraceful, you are right to leave.

[one2escape]

The frustration is that I didnt want to leave as its a hassle to get other things in.

[Kushan]

More hassle than the pathetic speeds you're getting and the even more pathetic response Virgin is giving?

[japitts]

Just a spot of thinking out loud here, partly because my connection is often (but not always) congested at peak times and anything from 5>20Mb/s is the going rate for my 50Mb. I'm also on my 3rd or 4th review date. I had problems with TiVo 10Mb last year too, but that's been fine for a while.

So it seems that VM have been leaving the congestion trigger points too late which causes problems with long lead times. And there's apparently a rolling program to deal with these, fair enough. But what, if anything, is being done in the long-term to learn lessons? I'm thinking of the triggers being tweaked so that by the time the issues become customer-affecting, plans are already well in-hand.

I guess that, already, if "planned" capacity increases in a given area are upcoming, then they'll be done at the same time as a "congestion fix" job.. Mind you, the same was supposed to be the case in my area and it didn't happen.

Like I said, just thinking out loud.

[horseman]

Originally Posted by japitts

..... I'm thinking of the triggers being tweaked so that by the time the issues become customer-affecting, plans are already well in-hand.

I guess that, already, if "planned" capacity increases in a given area are upcoming, then they'll be done at the same time as a "congestion fix" job.. ......

Capacity planning and congestion relief are different sides of the same coin? VM appear to have developed/inherited a number of ryo back end systems so planning almost appears to become a "black art"!

Given NTL/VM and now LG always have run contention extremely warm (if not hot) then we don't know how much VM currently use MRTG, Cacti, NetFlow or OSS-Iris or similar capacity planning in conjunction with their own ryo predictive back end systems?
Apart from obvious 5 yr rolling trend analysis there must be RGU/"Customer churn" feeds because despite long/protracted "congestion relief works" the revenue hasn't dropped and churn has actually decreased in previous quarters business results?

[Ignitionnet]

MRTG, Cacti and Netflow aren't used for these purposes at all as far as I'm aware.

Combination of own tools and feedback from Samknows boxes.

LGI use ServAssure.

[General Maximus]

Originally Posted by japitts

if "planned" capacity increases in a given area are upcoming, then they'll be done at the same time as a "congestion fix" job.

you have hit the nail on the head. I can remember a couple of years ago that there was an area which was absolutely diabolical for speed and oversubscription and it took them months and months to sort out. The sad thing was that the planned work was not capacity relief for the oversubscription, it was for the pending speed upgrades. One the extra capacity was added everything was great for a week or so before they doubled everyone's speed and then it went back to being **** again.

[Chrysalis]

maximus reminds me of my area some years back.

Constant upstream congestion for years on end, then was the 10:1 network upgrade program, and the upgrade work was carried out months before they upped the speeds. I remember for about 3-4 months experiencing the level of service some get on here all year round and was great, sadly they moved some users onto my node from the blackbird road node as those guys were getting congestion (instead of actually upgrading that node they did the cheapest option), then I was starting to see signs of congestion even before the speeds got ramped up, then the day they activated the higher speeds there was clear congestion the moment they flicked the switch. Either they were/are very incompetent or its the beancounters saying no go on upgrades unless is for marketing reasons.

Really this need to hit 90% utilisation is such a silly threshold, and its still in place now.

[Kushan]

If it's 90%, then it's an improvement on what it used to be. It used to be (If I recall) 95% for more than I think 10% of the time.

[Chrysalis]

It also was 80% once upon a time. It may have been lower than that before.

95% is just plain stupid, only makes any sense if your only concern is cost.

The only time any of these figures makes sense be it 80,90,95% is if an upgrade is ready to go at a push of a button but with it taking VM dozens of months it should actually probably be an ongoing process so when one upgrade is done start on the next. Funny the two largest transit providers in the UK for broadband both dont have a clue on capacity planning, BT and VM.

If I was the CEO of VM this is what I would do.

Keep the same "up to" packages. Take advantage of the very soft ASA guidelines which allow an isp to only provide the marketed speed to 10% of customers, then in high utilisation areas, roll out lower speed modem configs, if the ASA moan, point out 10% of the customer base can get the speed. I would say I would have a auto upgrade program for every single node, but I am sure someone would tell me that would bankrupt the company so I wont say that the business plan probably only works when parts of the customer base get screwed over.

[qasdfdsaq]

Originally Posted by Chrysalis

It also was 80% once upon a time. It may have been lower than that before.

95% is just plain stupid, only makes any sense if your only concern is cost.

If your only concern is cost you'd run it at 100% and let congestion self-limit.

Funny the two largest transit providers in the UK for broadband both dont have a clue on capacity planning, BT and VM.

In my personal experience BT have been far better. Over the most recent 3 years with each ISP I've experienced a sum total of 4 weeks' of congestion on BT and about 24 months on VM.

[Chrysalis]

qas but they do let it run at 100%, as by the time it gets work raised it will then be 100% for many months before any relief work. Remember the problem still exists after its raised. Which is the point I was trying to make, any one with intelligence knows the capacity relief thresholds are inadequate by a large margin they clearly set at that level for congestion to limit usage.

My own experience BT is better also, it seems just about every address I witness a VM service there is congestion. But BT have let their service rot for some users for 6+ months also, the saving grace tho the congestion can be avoided via hopping thanks to redundant routing.

[Kushan]

Originally Posted by Chrysalis

The only time any of these figures makes sense be it 80,90,95% is if an upgrade is ready to go at a push of a button but with it taking VM dozens of months it should actually probably be an ongoing process so when one upgrade is done start on the next. Funny the two largest transit providers in the UK for broadband both dont have a clue on capacity planning, BT and VM.

If I wanted to be pedantic, I'd state that none of those figures make sense on their own. I dare say that even in the worst areas, "90%" utilisation rarely occurs because utilisation alone is almost meaningless without something else to correlate it. I.e. if a connection sits at 20% utilisation between midnight and 5pm during the week but skyrockets to 100% between 5pm and 10pm, then it's not going to have 90% utilisation overall but (in my opinion) needs capacity relief anyway.

Part of Virgin's issue (in my opinion) is the second figure is far too high even at 10%. I would happily tolerate 99% utilisation once in a blue moon, during a critical event or something but it shouldn't go above 90% more than 5% of the time.