Infrastructure & Networking Glossary

This glossary explains the terms dedicated servers, colocation, cloud compute and network capacity. These terms are used on server pages, server quotes and server offers. A number of terms on this page are referring to HostServer services. These HostServer services are linked from this page. The definitions on this page are only explaining the terms and NOT the services. The definitions are as accurate and neutral as possible to use them for server offers, server contracts and server discussions.

Autonomous System (AS / ASN) AS / ASN

An Autonomous System is a network, or a group of networks, that is run under a single, clearly defined routing policy on the public internet. Each one is identified by an Autonomous System Number, or ASN, which other networks use to recognise it in global routing. Large hosting providers, carriers and enterprises operate their own AS so they can control how their traffic enters and leaves their network. Learn more.

BGP (Border Gateway Protocol) BGP

Border Gateway Protocol (BGP) is the routing protocol of the Internet. BGP is used by each network to announce parts of the IP address space that it can reach (i.e. the parts of the address space that it is responsible for) and the routes (i.e. paths) that it uses to reach them. When traffic flows between two Autonomous Systems (AS) on the Internet it generally finds a route which goes through several networks. BGP allows the network operator to influence this traffic and also allows them to announce their own address space. Learn more.

BYOIP / BYOASN BYOIP

BYOIP is short for "Bring Your Own IP" and is used to announce IP address space that you already have on a provider's network, rather than using addresses that your hosting company supplies. BYOASN (Bring Your Own Autonomous System Number) is similar, in that it allows you to announce the AS Number that you already have on a provider's network. This way, you can keep using the addressing and the routing identity that you currently use, while hosting on someone else's network. This does not harm your own reputation, does not break up your continuous online presence and you still remain in full control. Learn more.

Peering

Peering are two networks that peer together (i.e. connect to each other) to exchange traffic to and from users of both networks. This peering can be of the settlement-free kind (i.e. neither pays the other), or of the paid kind. Direct peering improves the path of traffic to and from users of the peering networks, by reducing latency and avoiding transit through the upstream carriers of one of the peering networks. Learn more.

Internet Exchange Point (IXP) IXP

An Internet Exchange Point (IXP) is a physical location where all networks meet. It consists of switches of high performance which are connected by all the networks. A network connects once to an IXP via a port. Then all other networks can be reached from this port. IXP's concentrate all the interconnections for a region in a few cities which is the reason why traffic between local networks is kept local and thus is very fast. Learn more.

Unmetered vs metered bandwidth

Unmetered bandwidth hosting is a type of web hosting that is billed based on a fixed port speed for a flat monthly fee. There is no extra charge for the amount of data transferred to and from your servers. In contrast, Metered bandwidth is billed based on usage, this can be the total amount of data transferred or a sample of the data transferred over a time frame. Heavy traffic websites are best hosted on an unmetered bandwidth hosting plan, whereas low traffic sites can sometimes be more cost effective to host on a metered bandwidth hosting plan. Learn more.

Bare metal server

A bare metal server is a single physical machine dedicated to one tenant, with no virtualisation layer between the operating system and the hardware. Because nothing is shared with other customers, the full capacity of the CPU, memory, storage and network belongs to that one workload, which gives consistent performance and strong isolation. Learn more.

Dedicated server

A dedicated server is a physical server that a provider owns, maintains and rents to a single customer. The customer rents a complete physical machine and has full control over it. He does not have to purchase the server itself and the provider takes care of the maintenance. Dedicated servers are also called bare metal hosting. The term is used for single physical servers hosted for single customers. Learn more.

Colocation

Colocation is a service where you place your own server hardware in a provider's data centre. The provider supplies the rack space, power, cooling, physical security and network connectivity, while you keep ownership of the equipment. It lets organisations run their own hardware in a professional facility without building and operating a data centre themselves. Learn more.

Cloud compute

Cloud compute hosts virtual machines on physical servers and lets you provision and remove them as needed. Individual instances can be scaled up and down on demand. A hypervisor partitions each physical server into separate instances, so capacity can be added quickly to meet demand, and you are charged only while an instance is in use. This is very different from a single physical server dedicated to one tenant. Cloud compute trades that single-tenant isolation for elasticity and fast provisioning. Learn more.

10G / 40G / 100G

10G, 40G and 100G are terms that refer to Network Interface and uplink speeds in Gigabits per second. In simple terms a 10G port can handle 10 Gigabits of information per second. Thus 40G is 4 times faster than 10G, or 4 x 10G = 40G. Similarly, 100G is 10 times faster than 10G, or 10 x 10G = 100G. The higher speed is required for applications such as streaming, backup and many storage clusters that move very large amounts of information. Learn more.

Latency / RTT

For this article we will be referring to latency as the time it takes for information to travel from one end of a network to the other. Typically this would be expressed in milliseconds, and we will be referring to the round-trip time (RTT) - the time it takes for information to travel from one end of the network to the other and back to the source again. For interactive workloads, such as virtual desktop infrastructure, reducing latency is critical to providing an acceptable user experience. The key to achieving low latency is to place servers as close as possible to users and ensure that the network paths are of the highest quality. Learn more.

DDoS protection

A Distributed Denial of Service attack tries to overwhelm a server or network with traffic from many sources so that legitimate users cannot get through. DDoS protection detects this hostile traffic and filters or absorbs it before it reaches the target, often by scrubbing the traffic upstream, so the service stays available during an attack. Learn more.

Tier 1 network

A Tier 1 network or Internet Service Provider (ISP) is a large network which connects to every other network on the Internet without purchasing transit from another provider. These networks are typically considered to be part of the Internet's backbone, and most other networks purchase transit from these, or peer with them in order to access the Internet in question. Learn more.

Uplink

The uplink for a server, switch or rack is the connection from the server, switch or rack to the next level in the network, ie to the Internet. The speed of the uplink connection is usually specified in gigabits per second. A server, switch or rack can have several uplinks. These uplinks are normally configured to be redundant, so that if one uplink fails the connection to the Internet will not fail. Learn more.

Rack / U (rack unit)

A rack is the typical casing for servers and network devices in data centers. The capacity of a rack is expressed in so-called rack units (U). 1U corresponds to a height of 1.75 inches. Servers are typically described with 1U or 2U, which corresponds to the space they require in the rack. A full rack has 42U and more. This is how data center space is planned and sold. Learn more.