The joy of distributed architecture
Distributed architectures are wonderful when they work.
And let "God stand between you and harm in all the empty places where you must walk" when they go belly up and you have to find a way to fix them.
Skype is the process of recovering from a major outage that left users across the world unable to log into the VoIP service on Wednesday. Many computers that act as supernodes, the systems that provide directory information on Skype, were taken offline "by a problem affecting some versions of Skype". As a result many Skype …
This supernode crap is what made me ditch skype years ago. My pc became supernode overnight, and began consuming my complete upload allowance.
I agree free stuff is give and take, certainly with a distributed environment, but saturating my internet connection without giving me *any* control over it was a bridge too far.
And anyway, skype is closed as hell, I find standard SIP compliant VOIP much better - pick any provider or client, or even run your own server :)
It is a standard type of failure in distributed systems.
If a distributed system failure is triggered by exceeding a capacity threshold and if clients perform some form of service discovery this is unfortunately a standard feature. Positive feedback loop. N clients trigger a bug in a node, node falls over, clients go to next node adding M clients in the process. N+M bottles sitting on the wall... and so on
As a result distributed makes it only worse because you get all kinds of wonderful stuff starting from "standing waves" of collapse and finishing with dead zones. There is a way to design around each and every one of these, but it is not something people do before it has bitten them.
Biting the hand that feeds IT © 1998–2019