Redundancy is a good thing in today’s digital economy. It ensures that vital infrastructures remain available at all times. When it comes to major issues such as traffic flows, having a redundant connection is vital. But does the same thing also apply for a medium-sized company’s connection to the datacenter? Redundancy costs money – so it pays to take a critical look at its usefulness and necessity.
When a company moves its IT environment (whether primary or not) to a datacenter, it often looks at the ‘tier’ level of the datacenter. That level says something about the extent to which the equipment at the datacenter is available. In particular, the main focus is on which services and amenities perform redundantly, in which the word ‘redundant’ means ‘double’ (in contrast with what the word actually means, i.e. ‘superfluous’). But when things are redundant at your datacenter, that doesn’t tell you anything about the quality of that redundancy, or about availability. For instance, a network connection may be doubled up over the same or intersecting fiber optic route; but if that optical fiber is damaged, for example by excavation work in a street somewhere along the line, then both the primary connection and the redundant connection drop out and the availability of your data drops to zero. Also, the redundant connection does not need to have the same speed, yet it can still be considered as redundant. These are all things that may give you a sense of false security.
How much downtime is acceptable?
The Uptime Institute categorises datacenters into four ‘tiers’, based on the Telecommunications Infrastructure Standard for datacenters, TIA-942. These tiers say something about the number of guarantees for the type of hardware used to ensure redundancy. For example, a tier-1 datacenter offers no redundancy, a tier-2 datacenter offers partial redundancy, a tier-3 datacenter offers N+1 and a tier-4 datacenter 2N+1 redundancy. This means that in a tier-3 datacenter, all facilities are operated with one extra system as back-up, for example 1 UPS + 1 extra. With a tier-4 datacenter, there is one additional level of security, because it features fully redundant facilities as standard, plus one extra is kept in reserve, making 2 UPS systems + 1. So, obviously, the uptime and availability in a tier-4 datacenter is higher than in a tier-1. Yet even for a tier-1, the availability requirement is 99.67 per cent per year (tier-4: 99.995 per cent per year). For many organisations, having virtually 100 per cent availability is not only unaffordable, but is also not necessary. It is more a matter of taking a critical look at the SLAs and guarantees that a datacenter offers and what the allowable maximum downtime per year may be. In so doing, it is important to look carefully at the datacenter’s certification. Only then will customers know for certain that the criteria and amenities in the datacenter will meet specific quality demands. Next, look carefully at the way the redundancy is provided (are the facilities comparable, but actually separate?) and then choose the datacenter you need (or divide the IT systems over 2 facilities).
To be able to make a properly informed choice, you should ask yourself a number of questions:
- What is the maximum allowable downtime per year?
- How is the redundancy implemented at the datacenter (number and capacity)?
- Are the redundant facilities physically separated?
- Is redundancy guaranteed during servicing?
- Is there any redundancy in the network providers?
- Is redundancy set out in the processes and procedures?
- Are there guaranteed SLAs with the system providers?
Make the right choice
Redundancy is essential for vital data traffic, but is based on a pricelist. For many organisations it is recommended to take a critical look at the actual need for redundant datacenter facilities. Often, a combination of your own computer space (or tier-2 datacenter) and a tier-3 datacenter will achieve the right level of reliability at an acceptable cost.[ssbp]