The last two t3n editions focused on Hosting Hosting for Reseller and Managed Server. This time we highlight the topic “Root Server”. While in a managed server, the hoster takes over a large part of the administrative tasks, it is responsible for a root server itself for the smooth operation. This means more effort, but also provides greater flexibility and also saves costs. We show what to look for when choosing a root server and compare the offers of many hosters.
The name “Root Server” comes from the Linux / Unix environment. Who is “root” has unlimited rights on a system – but also the duty to take care of everything carefully, to solve any problems yourself and to know what he is doing there. There are many in this product segment, some of which differ significantly in terms of price and performance.
The basic system
In the first place is the decision to the desired or required hardware. Depending on the individual needs, some providers use some “old” but cheaper systems, for example based on the Intel Pentium 4 processor. Other providers rely on modern multi-core or multi-processor systems. If you want to run virtual machines on the server, you should look for a modern CPU with hardware virtualization features that are required for current solutions such as KVM or Hyper-V. But even today this is already possible in the “standard segment”, as our comparison table shows. The included memory will usually be two to four gigabytes and up depending on the purpose.
The hard disk space is realized in most systems by local hard disks. Here, the use of at least two redundant hard disks is recommended, in order to prevent one of the most frequent causes of hardware problems and, if possible, to avoid a new installation or restore from a backup, including the associated downtime and work. This is now “standard”, shows the comparison table. When using a RAID1, ie the mirrored storage of data on both hard drives, of course, only the space of a single hard drive is available. Whether the hard disks are operated in a software or hardware RAID depends on the personal preferences as well as the required throughput. Throughput is usually higher for hardware RAID systems, because the controller handles the work here and also has access to a separate cache memory (usually 128 to 256 MB). In some data centers, it is also possible to rent storage space on central SAN / NAS systems (eg via iSCSI).
If you prefer to use hardware from a particular manufacturer, for example, if you have had good or bad driver or longevity experience with certain controller models, then this is another criterion in our comparison. In general, however, “noname” systems are quite solid and use components of well-known manufacturers.
Traffic and connection
The volume of traffic contained in the selected package should be sufficiently large. Depending on the application, quite different volumes of “a few gigabytes” (for a simple website or a mail server) up to several terabytes (streaming or download server) can come together every month. It is also important to pay attention to the cost of traffic overruns, which can be disproportionately expensive, especially for some lock offers. Offers that contain an unlimited volume of flatrates are often linked to unclear fair use conditions or linked to hidden restrictions that limit “unlimited” to “usual / average usage” by other customers. In fact, unlimited offers are rare today – or in a higher price segment. If “unlimited” means “as much as possible”, it may happen that you share, for example, with 20 to 50 additional server customers a shared 100MBit cable – slow transfers or even packet losses are unfortunately inevitable here. What would be bearable for the cheapest possible download server is unsustainable in a business or streaming server.
The often boldly advertised information on “available external bandwidth” are usually only limited meaningful and to be enjoyed with caution. Not always “more” really means “better”. Several GBit are advertised here, although this only corresponds to the sum of theoretical bandwidths at a certain point in the network and the bandwidths are not consistently available. In the case of cheap hosting offers, in particular the connection to the rack or to the server is often the decisive bottleneck. It makes more sense here to inquire with the provider about the guaranteed bandwidth available for one’s own server – and this all the way to the worldwide Internet.
To operate multiple domains / websites on one server, it is sufficient to use a single IP address (virtual hosting). However, if different services are used, for example, multiple independent SSL sites, you will need separate addresses for this. In order not to unnecessarily “consume” many addresses, providers generally only allocate a small number of IP addresses to customers by default; others are usually available on request or at an additional cost. While RIPE policies do not allow you to charge for IP addresses per se, they do appear to be a “maintenance fee” or similar term in practice.
If you would like to use IPv6 on your server, you should inquire in advance whether the hoster already has sufficient knowledge, offers IPv6 in his backbone and whether worldwide accessibility via suitable upstreams is ensured. For some providers, IPv6 usage is already possible, though sometimes only on demand or explicitly as an experimental service.