New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Are you seeing any particular issues?
No, I was searching Google for kvm providers in UK just for fun, and found some which own data centers, so I would like to see if these dc's are multihomed or not.
Bgp.he.net
which ASN ?
I don't have a specific asn. How can I see at bgp.he.net if the provider DC is multihomed?
paste an ip in
google "< provider > asn"
example:
http://bgp.he.net/AS62240#_graph4
Though you cannot put too much trust in that - If space is announced with a lot of prepends on one ISP it is essentially singlehomed with "only" a backup-on-fault system.
@Willim not necessarily as, for example, you have no idea how the outbound routing looks like and there is no way of knowing externally unless the provider provides a looking glass. Outbound routing is much more important in a hosting scenario.
That's stil multihomed.
@william, this is single homed:
Yes, thanks, i know how a multhihomed network looks like
@William that's Anycasted
Not only (1 /24 is unicast (UK) and ~600 announced /48 are unicast v6, while a /23 is anycast and a /48 v6 is as well)
So two different network policies on one ASN? ?
600 /48? You should be arrested for that deaggregation.
@patrick7 in their case it might be technically justified.
Not my problem in the end - The v6 routing table will grow far more anyway as /48 is the default PI/Direct Allocation size.
Anyway, my ~600 /48s are not global - Only sent to HE, which is fine with what and how i do it (and their routers handle it fine) - 90% of the routes are not further distributed as HE sends out a /43 announce (/48s are filtered) per location for global access and my other upstreams do only get the /43s announce, for worst case scenario (and some idiotic eyeball networks with filters larger than /43) i also announce it as /32 and do my own routing between locations (tunnels).
600 is also not a large amount in total numbers (and by end of year it's easily 1000+) - i use 2-3 /43 (64-96 /48) per location currently and assign /47s to enduser tunnels (with 1 /48 routed and another for future use).
More than 2 - Anycast for R4, Anycast for me, Unicast for me, Unicast for a few customers, Unicast for my tunnelbroker and downstream BGP for Tunnels with BGP.
I can use one ASN for everything like that, from external PoV it makes not really any difference (and RIPE approves that to save ASNs even though with 4byte this is now irrelevant).
Technically it works, however it's not standard. Some good networks do BGP optimisation based on external factors, like latency and packet loss. Some do it on per AS basis. Heaving it set up this way undermines those metrics, and hence hurts potential benefits that could otherwise be achieved.
Nevertheless, nothing to make a fuss about, but while ASNs are essentially free, why not make it one routing policy per ASN as it was supposed to be ?
OVH, Softlayer and Leaseweb also use one ASN globally, as do Level3, Cogent, Telecom Italia...pretty much any Tier1...
I also dont need to care much about networks that do optimization by AS as these are mainly DCs (and then only the ones with the overpriced FCP) while my customers are 95%+ end-user ISPs that barely do any optimization at all. Worst case i can still contact the ISP if there are issues.
I could get a bunch more ASNs from RIPE (and i have some more, i.e. 61163) but it is easier from management/software PoV to use one globally, the other RIRs partly bill by ASN so it would cause me much higher costs to use ARIN/APNIC ASNs for their respective footprint.