Content Delivery Networks
Michal Krsek
michal at krsek.cz
Tue Aug 7 14:05:07 UTC 2007
Hi Patrick,
>> 5) User redirection
>> - You have to implement a scalable mechanisms that redirects users to
>> the closes POP. You can use application redirect (fast, but not so much
>> scalable), DNS redirect (scalable, but not so fast) or anycasting (this
>> needs cooperation with ISP).
>
> What is slow about handing back different answers to the same query via
> DNS, especially when they are pre-calculated? Seems very fast to me.
Yes DNS-based redirection scales very pretty.
But there are two problems:
1) Client may not be in same network as DNS server (I'm using my home DNS
server even if I'm at IETF or I2 meeting on other side of globe)
2) DNS TTL makes realtime traffic management inpossible. Remember you may
not distribute network traffic, but sometimes also server load. If one
server/POP fails or is overloaded, you need to redirect users to another one
in realtime.
> Application redirection is far, far slower. (I am assuming you are
> talking about something like HTTP level redirects. Did you mean
> something else?)
Yes, it is slower, but in some scenarios (streaming hours of content for
thousands of spectators) works nicely.
Currently I'm using both mechanisms. Typically I'm using application level
redirets and when I know there will be significant peak (e.g. ten times more
than average), I'll provide DNS FQDN :-)
> As for anycast, with your own backbone, you don't need any cooperation.
> Even if you don't, the cooperation you need from your providers & peers
> is minimal at worst. (At least relative to writing the code for, say,
> DNS redirection.) But anycast assumes "BGP knows best", and we all know
> that is not even close to the truth.
I'm definitelly not talking about situation when you are running your own
backbone.
Anycast in general works for small transactions (root DNS fits perfectly),
but Internet is "living beast" and if you provide live stream that lasts two
hours, one flipping BGP session makes the content almost unusable for part
of your users.
Regards
Michal
More information about the NANOG
mailing list