Evaluating Data Centre Pricing – Detailed Factors
In a previous article, I outlined a broad-spectrum view of some of the major pricing components of a typical response to an RFP for colocation data centre services.
In this piece, I’ll discuss some specifics relating to some technical and highly-variable factors in pricing, such as power consumption.
What you’ll see in a proposal
Unfortunately, there is no standard response format that dictates how a potential provider will encapsulate their proposal. For example, typically, power will be mentioned but in terms of the ‘how’, it’s impossible to say.
The provider’s discussion may be highly granular and detailed or it might be a single line item referenced in the proposal. You may have a number of options whereby you can influence those costs or you may not.
Some basics
To begin, it’s perhaps necessary to brush-up a little on school-level science. It’ll help put some of the provider’s possible terminology into context.
Specifically:
- the Watt – a unit of power used to describe energy transfer (technically 1 joule per second);
- a Watt is also calculated (in this context) by amps x volts (so, W=AxV);
- a KiloWatt (kW) is simply 1,000 Watts (sometimes written as W(103) so 5,000 Watts is 5×103 Watts or more commonly – 5kW).
This matters because the Watt is sometimes, incorrectly, confused as a measure of energy. It isn’t and as stated earlier, is rather a measure of energy transfer. That means it is a very common way of measuring the energy transfer required in order to achieve something.
In the context of co-location proposals, the provider may use it to indicate just how much electrical supply they are making available and at what cost, to your equipment. It can also be described colloquially in some proposals as “the power” being delivered to you.
While on the subject, you may see some of these measures referred to also:
- Amps (amperes) – a measure of the electrical current coming over the power cables. A lot of IT equipment will be rated based upon the amps consumed. This is imperative as a sum of the total amps required by your installation will be a measure of required power to service it;
- Volts – this is the unit of potential difference between two electrical points and therefore the electromagnetic ‘pressure’ (analogy) that drives the flow of power;
- Ohms– this is the unit of resistance that slows electrical current flow.
Other factors
There are some other terms you may see included and priced:
- Circuit(s) – these are the distribution points for electricity, including various circuit breakers. Note that a lot happens in the data centre with power before it reaches a circuit including UPS (Uninterrupted Power Supply), PDU (Power Distribution Unit) and Power Panels. I won’t be saying more here on anything that happens before the circuit;
- Whips – an unfortunate name but this is the lengthy flexible cable which often is used to connect the power panel circuit to your equipment. It’s sometimes described as a ‘flexible circuit’;
- Cords – the individual power cables that connect your equipment to a whip.
What this all means
Potential providers typically will price based upon a configuration. That is roughly what they expect your equipment to look like both in terms of its nature and how it is stored.
The storage element is almost a premises issue. In other words, the more equipment you have, the more racks and cabinets you will need to store it and that means the more equipment you might need to rent. In addition, it will also affect your total floorspace requirements.
The effects of those things on your costs are clear.
What your configuration ‘looks like’ in terms of what it comprises, is very different and more complex. Clearly in general terms, the more equipment you have and of certain types, the higher your requirement for electrical capacity is likely to be. The higher that is, the higher your costs may be.
This can result in some complex design issues.
For example, some equipment is more power-hungry than others. Looking at the efficiency of your devices is therefore important.
You might also look again at exactly how much of your IT infrastructure is co-located because it doesn’t have to be all-or-nothing. You may choose to keep some low-criticality but power-hungry items in your existing premises where possibly your electrical supply costs are more attractive.
Some client organisations choose to do their own design analyses and hand it to the data centre potential providers on a ‘price this’ basis. That’s fine if you have the required topographical design skills and/or external consultancy to help. If you do not, that might be risky and you could lose opportunities for optimisation and cost reduction.
Keep in mind that working through the power consumption and circuit requirements of your installation is not something to be undertaken lightly on a DIY basis.
Another common approach is to give a generic topography of your existing infrastructure and applications architecture to providers and ask them to propose a design and costed proposal.
From that, you may get a proposition that involves many of the above components and probably many other factors. You might see prices segmented by:
- price per kW/hour for your entire establishment co-location based on sizing estimates plus a percentage contingency – sometimes called “all in”;
- prices per circuit or per rack (or conceivably cage for larger installations);
- metered cost per kW hour actually consumed – less common for smaller to medium-sized installations;
- etc.
As I can’t tell you exactly what elements will form the basis of your proposal, I can’t be specific as to what you should be doing in response. I can though help with some general questions you should be asking by type of pricing in the proposal.
All in pricing
This has intuitive attractions such as:
- single price/cost line item;
- all inclusive (usually);
- easier cost forecasting due to fixed price per month;
- should offer a degree of protection against unexpected electrical supply price increases.
It comes with some issues though:
- if you over-specify, you’ll be paying for more than you’re consuming. You could easily purchase more circuits than you need too.
Ask some of the following:
- does the provider guarantee the kWs they’re selling you on 100% loading and availability? Getting it for only say 80% of the time might be a problem during the other 20%;
- are they able to guarantee your maximum use is factored into all other clients’ maximum use and that their site could cope in such circumstances;
- does it provide dual A/B circuits at that level for both? If you buy 5kW but a problem means you need to default to the other circuit and it can only offer 2kW, then you may have trouble;
- where they measure their kW delivery. At your rack or cage is optimal. Further up the power supply chain (e.g. power panels) might mean you’re getting less due to resistance over Whips etc. No point in paying for a level you’re not getting.
Price per rack (or cage/cabinet)
This is, by definition, a slightly more complex arrangement.
The basic principle is clear but it essentially turns upon how many circuits per rack you need and of what type (usually the number of Amps).
The attractions are easy to see:
- typically economic (apparently – but verify);
- easy to understand in principle.
They need to be balanced against:
- the risk of over-purchasing;
- rapid expansion can be difficult to quickly accommodate.
There are some important questions to ask:
- exactly how much power can be delivered as a maximum to a rack – remember, that’s not always the same as the figure you’ve purchased;
- is cooling guaranteed for the maximum power load;
- could the circuits handle a maximum equipment load on the rack?
Metered (pay for the power you use)
This is a system most of us are familiar with from our domestic arrangements and therefore it’s easy to get to grips with.
The plus points are clear:
- essentially, each month (or agreed billing period) you’ll get an invoice for the power you’ve used;
- you may therefore have a degree of dynamic control over your costs incurred.
You will though need to be aware of:
- your cost base will be harder to forecast as it is usage based;
- your base costs may vary if supply costs do;
- your costs will be affected by the PUE (Power Utilisation Efficiency) of the installation. This needs to be found and assimilated if you’re to avoid paying for the provider’s inefficiencies but it is also an area of complexity you may wish to avoid.
Key questions:
- how do they charge for the electricity and how does that cost relate to what you have been billed;
- will there be an extra charge for cooling;
- where is the meter in the circuit? Typically, you want it right at your rack so you know the charge is the power you’ve consumed and not power being ‘lost’ further up the supply chain.
Red Warning Lights
Basically, you’ll need to be on the lookout for hidden charges or perhaps charges that are not assessed correctly in terms of their future impact on your cost base.
There is nothing particularly technical about these and common sense should be the order of the day. Things such as ‘management fees’, ‘fuel charges’, ‘support staff rates’ and ‘area maintenance charges’ can all be lurking in odd corners of the proposal.
Even if you think these are reasonable, make sure they are also constrained and controlled and you can’t end up being charged whatever the provider wants.
Summary
Ultimately, you need to understand the cost per kW as a bottom line. That can take a little mathematics.
Remember, Volts x Amps = kW. So, to see what you’re getting for a given installation, using sample figures:
- 200V x 20Amps = 4,000Watts. Divide by 1000 gives 4kW;
- Assuming only 80% of that can be delivered by electrical regs = 4×0.8= 3.2kW available.
Based upon that figure and dividing by the quoted power price, you’ll get a good cross-comparable figure to look at between proposals. Do remember to factor in things such as cooling costs and other charges which might change your perspective.