There are so many ways of classifying computer products. Servers are at the heart of the IT industry. Today I wanted to share with you my proposal for how all servers can be classified by their physical attributes. This is work in progress, so your comments will be very helpful to make this a useful tool for research purposes.
I’ve chosen 7 criteria which, althoiugh not mutually exclusive, provide a clear way to understand what each product is. These are:
- Server location – servers are installed in the customers data center or office location, a colocation data center, or in a cloud data center.
- Application basis – server applications are split into those running only on physical systems (i.e. those that run one at a time on top of an operating system), those running in a virtual machine and those running in containers such as docker.
- Element form factor – servers can be built with their components integrated into a single system (like a PC), with elements fitted to a proprietary chassis (asn blade systems are for instance), or fitted as elements into standard 42U racks.
- System scale – I’ve selected three price classes for the overall cost of the server – small (<$100k), medium ($100k-$1m) and large (>$1m).
- Storage type – whether integrated, or fitted as separate arrays servers arebuilt to handle three types of storage – file (as under most WIndows systems), block (as in most mainframes and Unix servers) and object (as in most high end cloud servers).
- CPU type – the three types of processor are x86 processors from Intel or AMD, proprietary (IBM z and Power, Oracle Sparc, Intel Itanium) and ARM (used in large systems such as HPE Moonshot and many self-built cloud systems).
- Operating system – servers can be split into those running Microsoft Windows, proprietary/Unix (z OS, iOS, Unix variants (IBM AIX, Oracle Solaris, HPE HP-UX) and Linux.
While there are other factors which could be used, I believe these criteria are enough to provide an overall classification.
So let’s look at a few servers to see how this schema works out.
Microsoft has released an on premise version of a server that can run its Azure cloud applications called Azure Stack Hub. Hardware vendors such as Dell EMC, HPE, Lenovo and Huawei provide their standard servers as the basis of these systems.
AWS has started offering Outpost to its customers – a version of the self-designed servers it has designed, built and installed in its own data centers. Unlike Azure Stack Hub, Outpost servers are entirely proprietary designs provide by AWS, rather than built ontop of baranded servers.
Dell PowerEdge servers come in many form factors from simple single-processor machines installed in offices to rack-mounted elements of large systems installed in big data centers, although they are not typically used by cloud service providers. As far as I know all PowerEdge servers are based on x86 processors.
IBM mainframes are still the servers of choice for many financial services, government, insurance and other industry sectors. They are all based on its proprietary z system processors; upgrade elements fit into a proprietary chassis provided by IBM.
A good classification schema would help me – and I hope others – in comparing one system with another and avoiding confusion. Let me know what you think. If this is useful we can work on storage and networking product classifications as well.