Saturday, 4 June 2011

"Packet loss" as a fault

Our favorite telco once again show their total ineptitude by again failing to understand the basics of IP networking.

We have reported packet loss, something we can  easily test as we do LCP echos on every line every seconds.

Packet loss is normal when a link is full and the way TCP adapts to a full link, but random packet loss fools TCP so one of the main effects of packet loss is low speed on file transfers using TCP. It is not the only impact, as random packet loss can cause drops in VoIP, delayed DNS lookups, missed syslogs, slow throughput, and so on.

Our favourite telco are insisting that we identify the end users perceived problem. They cannot understand that the we have a technical end user that perceives "random packet loss" as "the problem". It is not "when I got to facebook it is slow", it is "I can see there is random packet loss". They also fail to understand that *WE* are their customer and what we as the customer perceived as the fault is measurable random packet loss on the line...

Oh what fun.

As an ISP I would far rather a customer came to me saying "I see 2% packet loss" or "I see 2% packet loss to XXX". I can test and diagnose a report like that. What I don't want is someone saying "The internets are slow". It seems out favourite telco want the latter and when you provide actual detailed reports of exactly what is wrong they cannot cope.

Arrrrg!

9 comments:

  1. "Oh hai BT teh Interwebz is slow pls can u fix it my techie frnd sez its random packet loss whtvr tht means lol. luv revk"

    ReplyDelete
  2. Stop whining. It's never been better than 'best effort'. Deal with it twonk!

    ReplyDelete
  3. Yes, and a service that is randomly dropping a few percent of packets is FAR from their BEST efforts and is a fault.

    I have to assume that is just a troll post as nobody thinks random packet loss is acceptable if they have any clue. If you are genuine then there is no point being an A&A customer ever as we do take such things seriously.

    ReplyDelete
  4. Who are you to judge what their best efforts are? If you feel so passionate about it, put your money where your mouth is and provide your own national infrastructure - dickhead.

    ReplyDelete
  5. Now you have to be a troll.

    Their best efforts are clearly what we can see that they can do - if we have many thousands of lines that do not have random packet loss, and one that does, then clearly that line has a fault.

    "Best efforts" has always been a rather odd way to describe the service anyway. The "best" effort is to put dedicated fibres and backups in to every home - they don't do that. they do something far less than the "best" that they could do. It is also not a measurable target which is not good as a service description.

    We pay them to convey packets - it is pretty simple. If they introduce random packet loss then they make the service unacceptable and we would stop using them if they thought it acceptable to offer the service like that. It is one of the reasons we also deal with other telcos.

    My "rant" here was that they do not understand packet loss as a type of fault, which is crazy as it is a measurable network metric. What they want is a vague end user perception as a fault report rather than a tangible and measurable problem. That is what is crazy.

    ReplyDelete
  6. How about "Customer is losing some Multicast market pricing data"? An example of time-sensitive, unidirectional data that can't be retransmitted, and isn't suitable for sending over TCP.

    ReplyDelete
  7. Well, yes, we could make something up, and "drop-outs in VoIP" is another, slow DNS lookup is another, and so on. But the issue is that we have done the work for them - identified the underlying problem is packet loss, and they want to bugger about.

    ReplyDelete
  8. Well - this is the same "FaveTelco" that lost a shelf in the DSLAM/MSAN in my exchange late last week (with my line and about 90 others attached) were told about it within 20 minutes (via ISP) yet contrived to take 21 hours to restart that shelf.

    Syslog? SNMP? Seems they're too new-fangled for FaveTelco.

    ReplyDelete
  9. "Best efforts" is a very misunderstood expression, and it's very dangerous if it's part of a contract.
    People seem to think it means "well we'll try, but if we can't fix it that's tough". Wrong! That's "Reasonable efforts".
    "Best efforts" means that they will use everything in their power to fix it, up to and including committing the entire company's resources to it (is it could bankrupt them) - anything less isn't "Best"!

    ReplyDelete