Before the questions
These are the questions people actually ask while choosing a configuration. The BGP answers hold anywhere; the SKNK Cloud answers reflect what we publish today.
Questions & answers
Do I need BGP to host a website or an API?
No — and most sites never will. While the provider handles routing, one working public address carries a website or an API happily. BGP earns its keep in specific situations: you exchange routes yourself, you announce an authorised prefix, or you genuinely want to learn routing. Without one of those, start with the ordinary configuration and keep the complexity you would have bought.
What is the actual difference between BGP VPS and Standard?
Network options. Standard runs conventional networking with no BGP session; BGP VPS carries one session per tier, extendable to two, and the additional IPv6 /48 exists only there. Compute, storage, native IPv6 and optional public IPv4 appear on both sides. Whether the extra session earns its place depends on whether your design manages routing itself.
Does a BGP session require my own public ASN?
No. The two ends may use public or private ASNs, and the ASN used inside the session can differ from the origin the public Internet sees. Whether you should get your own public ASN depends on how routing is delivered, who holds the authorisation, and what your upstream requires. Settle who originates publicly before you touch a ROA.
Does picking a BGP configuration come with an ASN or a prefix?
No. The BGP option buys the capability to run a session. An ASN, an address assignment and permission to announce are separate resources; none of them materialise because you ticked a box. We also run an ASN sponsorship service — it is a different thing from Cloud, so keep them apart in your head.
Who announces the additional Cloud IPv6 /48?
Under the published plan, we do — AS219361 announces it publicly and routes it to your VPS, which is why you need no ASN of your own for it. How much of the block you may use, and on what terms, follows the delivery agreement.
What is the difference between one /48 and one IPv6 address?
An address is a door plate; a /48 is the estate. For a documentation example: 2001:db8:1234::20 is one plate inside 2001:db8:1234::/48. Being handed one plate does not put the estate under your management.
Two BGP sessions means two independent lines?
No. Two sessions can land on the same router or the same uplink. The count says nothing about independence; the routers, the lines and the failure domains behind the sessions do. Ours start at one and go to two — and at two, this paragraph still applies. It is not redundancy insurance.
Default route or full table — what is the difference?
A default route is “everything else, ask the gateway”: it handles only destinations without a more specific match, and it is cheap. A full table turns your routing software into a live map of the Internet — policy can get very fine, and memory, CPU and update load climb with it. Which one a service feeds you is a question for the provider. One caution: neither term says anything about speed.
How much RAM do I need for BGP?
Answer three questions first: how many routes and paths you will keep, which software, and what else lives on the machine. A default route plus a few local prefixes and a full table with multiple paths are different leagues. My advice is unglamorous but works: measure memory and convergence on the software version you will actually run, then decide. Any universal “full table fits in X GB” figure you read online — treat it as folklore.
The session says Established. Why is the address still dead?
Because Established covers only the neighbour layer. Every ring below it can break: the prefix never exported, filtered upstream, traffic never arriving, the address unconfigured, the firewall blocking, the application listening elsewhere. Walk from network to application — local exports and upstream acceptance, then external route views and arriving packets, then host address, firewall and listener. Most of the time the answer is in the second half.
RPKI says Valid. So the service is reachable?
No. Valid means the claimed origin ASN and prefix length match a validated authorisation; it does not authenticate the sender. It does not look at the path, the filters, or your application, and it will not open your website for you. Judge reachability with three things together: origin validation, an external route view, and one actual request of your own.
If I move providers later, does the additional /48 come with me?
By default, no. A prefix routed with a service is not PI space in your name; when the service stops, the route stops. If you want to build something long-term on an address block, either hold your own resources or read the use, withdrawal and exit terms line by line before you commit anything to it.
Will BGP make my IP cleaner or my latency lower?
Neither happens by itself. BGP manages routes; a clean IP follows usage history and abuse records, and latency follows the actual path, interconnection and load. A new ASN or an extra session changes none of those causes. Anyone selling BGP as a way to “wash” an IP deserves extra scrutiny.
Can I use BGP for Anycast or high availability?
The protocol allows it — Anycast is several places announcing the same address. Deploying it is a different story: the prefix must be authorised, the routes accepted, the application health-checked, the traffic catchable when it moves, and stateful services need a plan for connections and replication. The boundary, stated plainly: protocol support does not mean one of our products already ships this.
Can I order or pay for SKNK Cloud today?
Not yet. Ordering, payment, signing and activation are all closed, and prices are not shown. What you can do: browse configurations, save a configuration link, copy a summary — none of it reserves capacity, allocates addresses or creates an order.