We're currently evaluating web hosting providers in the APAC region and one of the criteria that we are currently considering is the availability of routes going to the web hosting provider.
In this regard, I would like to ask for your idea regarding this. Is it safe to conclude that the web hosting provider's available routes would would depend on the peers who are advertising their AS / network? (i.e if web hosting provider claims that they are peering with telco a, b, c but as seen from a third party looking glass, only C is seen advertising the web hosting provider network that would mean web hosting provider is effectively utilizing c as their upstream??)
That you dont see that route on a lookingglass doesnt mean much. Only Could tell you they dont transit there.
Its all depending what you definiëren with available routes.
If i peer with all ISP's in a specific area and your looking glass isnt licated there does that mean its bad? You need to know much more. If your customers are local there its even prefered.
Its never that black/white ...its depending on your needs!
Thanks for the prompt response. Actually our requirement is to find a webhosting provider whose routes are widely advertised locally and regionally. This is why I thought of using bgp as a basis studying the availability of routes of the hosting provider.
You wont see those local peerings unless all those providers have looking glasses. So thats not gonna work out in this case. You will only see who they transit with...
Ok. Thanks for the information So that would mean that to answer my question, I would need to determine the web hosting provider who has the most number of peers and most number of transit providers?
Ok. Thanks for the information So that would mean that to answer my question, I would need to determine the web hosting provider who has the most number of peers and most number of transit providers?
You wont see those local peerings unless all those providers have looking glasses. So thats not gonna work out in this case. You will only see who they transit with...
I cant answer that, only you can. If that hosting provider has many peers but your customers are not behind them its of no use.
But generally a hosting provider with many local peers, either public or private -might have- ok connections
For example if they private peer with 10 mbps and the other one has 10 gbps public peering...
There isnt a real answer to your question. Its based on your own buisiness needs and decisions on those.
Ok. Thanks for the information So that would mean that to answer my
question, I would need to determine the web hosting provider who has the
most number of peers and most number of transit providers?
what i found usefull is to check the autnum objects in whois, as many document
their peerings and transits there.
robtex has also some of this info, in a webinterface...
also helpful was peeringdb - you can lookup indvidual ASs without logging in
like this
it may give you an indication as to which exchanges your (potential) provider
is present at - though not all providers have a / maintain their peeringdb
record.
BGP is a distance-vector protocol. In other words, when BGP is
advertised on a link from one AS to another, only the "best" distance
to a particular route is offered.
If A and B both advertise the web hosting provider's (WHP's) route to
D and A's distance is better then D won't advertise B's WHP route to A
but it will advertise A's WHP route to B. Thus a looking glass at B
will see both routes in the BGP RIB, but a looking glass at A will
only see A's route. With more AS's between A and B than just D, it's
possible (even likely) that neither A nor B will see the others' route
during normal operation.
Should WHP's connection to A drop, D will find that B's distance is
now better and B's route to WHP will be newly advertised to A. A, then
having no other connection to WHP, will accept the route via D to B,
and pass it onward. When we talk about the BGP table "converging"
after a change, this is the process we're talking about.
Even if A and B are directly connected, if WHP sets its initial
"distance" via B worse than via A, B will decide that A has a better
distance to WHP and won't advertise its own version of the route to
anyone else at all. And that's before you consider local prefs,
communities and other mechanisms for fine-tuning route propagation.
And, even if A, B and C have multiple routes in the BGP RIB, generally
only one of those routes will be selected for the packet forwarding
FIB. So, a traceroute from B to WHP may travel via A even though B is
directly connected to WHP.
Long story short, if WHP is connected to A, B and C then A, B and C
should each see their own direct route to WHP in the BGP RIB, but
there's no guarantee that anybody else will see more than one of the
three at any given time.
Ok. Thanks for the information So that would mean that to answer my
question, I would need to determine the web hosting provider who has
the most number of peers and most number of transit providers?
if i was choosing a hosting provider, many other considerations would
come before this
Is it safe to conclude that the web hosting provider's available routes would would depend on the peers who are advertising their AS / network? (i.e if web hosting provider claims that they are peering with telco a, b, c but as seen from a third party looking glass, only C is seen advertising the web hosting provider network that would mean web hosting provider is effectively utilizing c as their upstream??)
Modulo a lot of nit-picking caveats, this would indicate that they are purchasing transit from C, while they may or may not be peering with A and B. If a customer of A or B is able to reach them in a single AS-hop, but A and B are not advertising a route to C to looking-glasses or their own peers or transit providers, then A and B are peers, but not transit providers, to the web host.
I would need to determine the web hosting provider who has the most number of peers and most number of transit providers?
A large number of transit providers has been shown, both theoretically and experimentally, to _decrease_ uptime, because of greater route-convergence times when there are more parallel paths to a destination. Your mileage may vary, but the optimum number is usually somewhere around three transit providers. The number of peers, though, and more to the point, the number of routes acquired through peering, is an excellent measure of how large and how long an ISP (in this case a web-hoster) has been in business. That shouldn't be your only measure of quality, however. It may be that reliability of power, competence of remote-hands, and flexibility to accommodate your needs are more important than how the packets get delivered.
peer - exchange routes with a neighbor (BGP/OSPF/ISIS/EGP/Static).
transit - your neighbor agrees to send your routes to -their- neighbors.
peering you can control, transit is controlled by a third party.
so Jack, you could pay Sprint for transit (they propogte your routes elsewhere) and
then insist on no-export for the routes you give them.. -IF- Sprint honours your
no-export, then your just peering, regardless of what you are paying for.
That's different from who the provider peers with. We (AS1312) don't
peer with much of anybody, but I *hope* our routes are pretty widely
advertised. Anybody *not* seeing routes for AS1312? (And yes, most
of the routes end up going through one or another of MATP's PoP's).
So what failure mode are you trying to protect against by finding a provider
with a lot of routes? I suspect there's probably a better metric to deal
with the actual concern...
So that would mean that to answer my question, I would need to determine the web hosting provider who has the most number of peers and most number of transit providers?
So that would mean that to answer my question, I would need to determine the web hosting provider who has the most number of peers and most number of transit providers?