1. Audit everything in the datacentre
If you don’t know what you have, you don’t know what you are spending money on. It might sound like a broken record for many industry stalwarts, but you would be surprised how many times it needs to be repeated. And the datacentre is no exception.
One of the first things that needs to be done in order to identify those areas where cost savings can be made in the datacentre is an audit. This should be undertaken yearly, if not bi-yearly.
The two obvious areas to audit are the IT equipment and the facilities. Often they are managed by different departments, but in the end both need to be integrated to gain a holistic picture.
From the servers and UPS to the switches and cables, every piece of IT equipment must be logged. Similarly, from the crack unit and in-row cooling modules to the light bulbs and power supply, the entire facility must be audited. Why? Because they all consume energy (or at least contribute to it) and that equates to money spent. If you know what it is, what it does, and how it is being used then you can take steps to reduce its cost. By extension, this holistic approach also entails auditing the applications being run from the datacentre – both from virtual machines and physical servers – to see what equipment is being used.
The final part of the audit is to look at the processes and staffing levels. Do you or the customer have the optimal practices in place? Are you obtaining the highest levels of productivity out of headcount? The answers to these kinds of questions can lead to change for the better, but only if you know what needs to be changed.
2. Consolidate hardware – and don’t re-purpose it
If there's less of it, then you should be using less money to power it, or at least that’s how the theory goes. From consolidating servers to upping storage utilisation rates, virtualisation can play a big part in helping to reduce datacentre expenses.
“Virtualisation allows you to do more with less and is changing the face of datacentres,” NetApp marketing director, Roger Mannett, claimed.
“Innovative software technology is now available to maximise the utilisation of commodity storage hardware, improving utilisation rates and reducing the amount of storage required. Using virtualisation technology can reduce the amount of storage by up to 70 per cent in some cases.”
Although the cost-saving benefits of virtualisation have been well-documented and continue to be validated by market examples, there are anecdotes of organisations cancelling out gains by repurposing equipment for things like disaster recovery (DR).
“Reducing the number of devices in the datacentre is a key one and we are seeing clients reduce their consumption by 30 to 40 per cent in terms of raw power usage for the servers, but that also has a similar impact for cooling,” HP mission critical facilities practice lead, Mark Toner, explained.
“We have seen some clients make significant savings just by consolidating and virtualising. The important point when you do that is you have to follow through and get rid of that legacy technology.”
3. Consolidate applications
The next step to take after consolidating the hardware is to migrate towards shared services and consolidating applications. This will allow an organisation to get more leverage not only from the hardware infrastructure, but also the software and applications.
While some departments may not want to share and care alike, sometimes executive decisions are in order to achieve the cost-saving goals being set. By matching the equipment consolidation with a similar approach to applications, significant benefits can be derived from less management complexity and fewer licensing expenses.
“It is a bit like balancing your books – obviously you have to make a little effort to identify things you can’t put your finger on in the audit,” Sun Microsystems chief technologist, Angus MacDonald, said. “At the end of the day, it gives you a low-cost saving. The only expense involved is the time used to identify those sorts of things.” ---P---
4. Turn stuff off
Do those lights really need to be on all the time? Probably not. The same philosophy can often be taken with the IT equipment – especially for organisations primarily involved in computational operations.
“The lowest hanging fruit with IT equipment is power management. If processors aren’t being used, then turn them off or put them into idle mode,” Emerson director of marketing, Peter Spiteri, said. “You will save about 80 per cent on the power on that server, which is really important because some equipment comes into play when you are doing backups, or whatever. But after that, it doesn’t need to be left in full power mode.
“The really important part is to know what sort of function you are doing, because if you are Google you don’t want to turn power management on – you want them switched on to respond as quickly as possible for search. But if you are CSIRO, where you are doing a bunch of stuff and then nothing for three hours, to wake it up takes a few seconds but once it is up and stays up it incurs costs for almost nothing. You need to apply these approaches appropriately to get those cost savings.”
If you have already done your audit, then this part of the process should be easier – at least that was the case for Sun Microsystems when it did its own datacentre consolidation.
“We took the decision that if we couldn’t identify what a server did, we’d turn it off,” Sun’s MacDonald said. “We did that and at the end of three months nobody came to us and complained when we then decommissioned the server. The amazing thing was at the end of three months we had removed about 8 per cent of the hardware. We didn’t get too much flak from people.”
5. Turn up the heat
Are you serious? Turn up the heat? Yes, dead serious. The old belief that datacentres need to be run at under 20C no longer holds water with the temperature ratings built into newer IT equipment.
Something as trivial as turning up the temperature at which the datacentre runs can have a fairly dramatic impact on air conditioning costs. Obviously datacentre managers need to speak to their vendors about what the impacts of that are, but running a datacentre at 23C or 24C degrees as opposed to 19C or 20C can actually save quite a lot.
6. Check cooling
Inefficient cooling set-ups are one of the most common costs in the datacentre.
Simple things such as poor placement of equipment, failing to utilise hot and cold aisle arrangements and ignoring progressive developments like in-row cooling can significantly add to the cost of cooling a datacentre.
“Power costs are the second largest operating cost for a server room or datacentre,” APC country general manager, Gordon Makryllos, said. “Organisations need to be looking at their power and cooling solutions as the key area for cutting costs. Cutting your power costs by 30-50 per cent can save IT budgets and allow budget redeployment from non-value adding to increased investments in IT.”
As part of this, a sound first step is for datacentre managers be aware of the dynamics and characteristics of air flow in the datacentre to say where the best place to put equipment is.
“Closer to the crack units, a lot of people is assume is a good place to put your heat intensive equipment – it’s actually not because what you get is a very high velocity of the air near the crack units which is less effective for cooling,” Hitachi Data Systems manager datacentres and storage services, Tim Munn, said.
“Two-thirds of the way across the room away from the crack unit is probably a good spot because the velocity of the air is reduced and you get better cooling. “You put your high density in your most effective cooling area and you might put your communications gear close to the crack unit because it doesn’t require as much cooling.” ---P---
7. Put sensors in place
To get the right cooling in the right place, however, you need to know what is happening at different spots in the datacentre.
“You can put things like sensors in a rack that measure temperature. As simple as that sounds, it’s one of the things people miss. Quite often, IT managers won’t know what their power bill is because they don’t actually pay their bill,” NEC product manager datacentre and hosted solutions, Loren Wiener, said.
“There is software you can get that ties in with the sensors so it becomes integrated. If you have a PC at home, you can click on it and see how much your CPU usage is, how much your Internet usage is and all the rest of it. Companies don’t necessarily tend to do that with their datacentre space or their rack space. So that is a typical thing you can look at.”
The sensors also will often tie into the server, into all the hardware, and even into all door panels. There are copious places that can have a sensor on them that can measure everything from the actual temperature to the security and help you gauge what is happening.
8. Modular UPS
In a lot of cases, it could be better in terms of cost in the short- to medium-term to pull out any old UPS and buy a more efficient one. According to industry representatives, a UPS as old as five years might be up to 10 per cent less efficient than a new modular, scaleable UPS.
“If they can afford to spend money on a new one, they can save money over two to three years in running costs,” Eaton general manager power quality, Michael Mallia, said.
Furthermore, if a datacentre manager is looking to upgrade or deploy a new set of equipment, modular UPS are the way to go.
“If you are planning a new rack rollout or datacentre, you should size the UPS nearer to the initial load and then have it capable of being upgraded in capacity as you need it,” he said. “That way you are nearly always running near the optimal efficiency of the UPS. Apart from that, go for as high efficiency UPS as possible.”
9. Raise awareness of energy costs
Knowledge is power and if the entire organisation is aware of the cost of power in the datacentre and how it negatively impacts the bottom line, they are more likely to assist in reducing consumption where they can.
For example, in a computation-heavy organisation, if employees know they can contribute to cost-saving measures by informing the datacentre of when projects need servers to be up and running and when they can be decommissioned, they are more likely to do so. This is especially the case if the entire organisation is aware of who uses what, where and when.
10. Follow guidelines from groups like The Green Grid
With a membership list that includes some of the biggest names in the game, governments and environmental groups, The Green Grid is one place you can go to for legitimate advice on how to help reduce datacentre costs.
The Green Grid also provides free tools to help evaluate cooling and power issues. Although its existing online tool, which has been developed to help datacentre and facilities managers determine how much free cooling and free evaporative cooling is available for individual datacentres, is only available in the US and Canada, it will be coming down under in future.