James

This is a great question and seems to be on the minds of many of the customer with whom I interact. The discussions used to be as simple as making sure products were Energy Star certified (a simple checklist item), but now they are more about real energy savings (work output per watt, etc) and being able to characterize the relative energy consumed by similar products. The discussion has further gone into recycling (while not strictly related to energy savings, it is part of a company’s “green” story), materials, manufacturing processes, etc.

There are several ways a company can help decrease their IT group’s energy consumption. My experience is in the computing industry.

First, take the efforts of processor manufacturers seriously. The current trends of shrinking die sizes and throttling individual cores are relevant and do save measurable power. Maybe not is a single system, but definitely relevant for a data center.

Second, take a close look at virtualization if relevant to your environment. Many system components sit idle (or relatively so) but still consume power. Granted, virtualization will increase utilization (and hence power consumption) of a single system, but in many cases that is less than the power consumption of multiple systems that are only partially utilized.

Third, explore power saving technologies that may not have been specifically developed for power savings. Case in point: solid state hard drives. More expensive? Yes. Smaller than rotating drives? Sure. However – the current draw of most any SSD is substantially lower than that of a rotating disk. I’m not sure most portable users actually need all 160 GB of their hard drive, so a smaller drive that consumes less power (and hence extends battery life) may be worth exploring.

Fourth, storage consolidation. The same reasoning applies with storage consolidation as does with server consolidation.

Fifth, remote management. A remote management product can save substantial energy by not requiring the IT staff to visit onsite if they need to troubleshoot an issue. Although not directly related to data center energy consumption, it still factors in to the total energy consumed by the IT department.

posted April 24, 2008

This is a great question. I think about the IT infrastructure as 3 main components: Servers, Storage, and Backups. The common thread to reduce power consumption is by consolidating and maximizing your physical resources.

Servers - For servers, the trend is to consolidate several physical servers into a few virtualized servers using VMware, Virtual Iron, XenSource, and the pending Microsoft Hyper-V for Server 2008. Though the features and costs of these packages all vary, the end result is same - a few physical servers running at high utilization hosting virtual servers instead of many physical servers running at low utilization. The cost savings include a reduced physical footprint (less room in the data center), reduced cooling costs, and reduced power consumption while actually increasing the uptime and reliability of the servers in the process in most cases. The ROI is typically tremendous.

Storage - Many of the same arguments apply to storage. Centralization is the name of the game. By installing a few centralized storage pools and then sharing these common resources you're able to significantly reduce the number of spinning disks in the data center, dramatically reducing power consumption, and complexity while increasing overall system stability (less moving parts is always better). What's better is that many of the past barriers of centralized storage (cost, complexity) have been eliminated with the adoption of iSCSI - it's fully mature and very easy to setup and manage.

Backups - Power consumption has long been an argument for backups. The 2 most common backup mediums are disk and tape. The goal for backups is to provide a balance between performance and overall cost for the best Total Cost of Ownership. A very good strategy is to combine disk and tape keeping the last full backup and daily differentials ready for fast restore on nearline disk storage, while keeping copies of your data on tape for offsite disaster recovery or archiving. The main advantage to tape is it's portability and zero power consumption while being stored. Again the centralized theme appears. It is much more cost effective to centralize several backup systems in 1 main backup system in the data center to reduce the complexity (less moving parts) and power consumption. There are some very innovative solutions available to share resources and centralize your backups even across heterogeneous systems - combining your backups for Windows, a VMware setup, and your AS/400 into 1 system for example.

In summary, the root answer is to centralize, centralize, centralize and optimize the utilization of all running resources - processors, disk space, backup space, etc.

posted April 24, 2008

James,

all the major hardware vendors are moving in the direction of lower power consumption as well as reduce the occupation of space in a data centre.

I went to some forums and they were very similar among each other, "green computing" was the keyword!

To decrease power consumption you should swap almost all your IT infrastructure, with the side effect of reducing also the power required for cooling.

I have to say you should be very careful of moving in that direction without a business case (but as you are a CFO this should be obvious), the cost required to swap the infrastructure could never reach the break even!

Also from a financial point of view, new HW is CAPEX, while power consumption/cooling maintenance are OPEX and in some cases are threated in different ways.

Luca

posted April 24, 2008

In a data center environment I have found that the greatest reduction in power utilization comes from centralizing servers through the use of virtualization technologies such as Xen or VMware. With power costs going up in the data center, there may be no better reason to do this. But there are many other benefits to using virtualization, including more effective use of overall computing resources, cost savings due to fewer new servers being needed, increased efficiency of administration, and so forth.

In the office environment, the single biggest savings will likely be by enabling power management features on your desktop machines. Most desktops are not used or needed 24x7 so why leave them running? Enabling power management may reduce your desktop computing power requirements by 50% or more.

These two steps by themselves may provide enough of a reduction to make further, perhaps more dramatic measures, less necessary.

posted April 24, 2008

James,

In the era of green initiatives it seems like many are trying their best to pitch in. In some cases a firm might not have the finances to make the necessary green upgrades internally to make an impact with regards to their power consumption. If this is the case they should look at outsourcing to a firm that is taking real steps to going green. My firm (Mercury Technologies) is a partner of Fortiva for email archiving services. Refer to the last section in the link:

http://blog.fortiva.com/

If your going to pass the energy buck, pass it to someone who's taking real steps to decrease power consumption and carbon output. If you want to discuss other options on how outsourcing can decrease power consumption let me know.

james.taylor@mercurytech.com

Good Luck, Go Green

posted April 24, 2008

I have been asking my IT managers and CIO for the last six months if they they see or are responsible for the power bill. So far not a single one is.

Responsibility for power costs needs to be moved to the IT manager running the data center. The IT manager will have the cost saving necessary to purchase the available tools.

Is data center power in CIO P&L?

As a group we need to lobby our congress to give tax benefits for server reduction.

Thank you for asking this question

Clive from http://www.focustsi.com

posted April 24, 2008

James,

An organization should begin by assessing their current archive management strategy. The goal is to understand the types of corporate data that is being generated, which data is persistent or static (a.k.a. “archivable”) and make sure that this data is being stored appropriately.


Most organizations have huge quantities of archive data sitting on traditional primary storage systems that drain disk, energy and IT
resources. Archive data by definition is unchanging and infrequently accessed, which means it does not belong on primary, or Tier One, disk. Instead, it belongs on lower-cost, or Tier Two, storage.


Once data has been classified, the next step is to get all the archive data off primary storage systems and onto an appropriate, less costly secondary archive tier. It is also important to put policies in place that ensure the data is moved from primary storage.

posted April 24, 2008

90s saw killer micro syndrome with throwing hardware at problems ... in lieu of scarce/expensive skills to understand/analyze problems. Going on two decades of the approach has resulted in large deployments of significantly underutilized hardware.

current buzzword is the 40+ yr old technology, virtualization ... which offers opportunity for significant consolidation w/o requiring significant additional skills/resources ... some scenarios of institutions going from 30,000 servers to 3,000 and similar reduction in datacenters.

the science center had done original virtual machine implementation, cp40 on 360/40 with custom modified virtual memory hardware. cp40 morphed into cp67 when standard 360/67 with virtual memory hardware became available. last week of jan68, 3 people came out from the science center to install cp67 at the univ.

Links:

• http://www.finextra.com/fullstory.asp?id=18404

posted April 24, 2008

James,

Having read a number of the responses to your question and working in the IT field, I can see that virtualization has become a major initiative throughout the IT community. Though it is true that through the proper setup and use of the technologies around this it becomes possible to not only decrease power use and space use, let us not forget that it also lowers the total amount of BTUs generated by these systems in climate controlled data centers. Lowering the amount of heat generated decreases the amount of time that your cooling systems need to run their compressors and fans, extending their overall life cycle and also significantly reducing the power costs involved with longer sustained run times. An additional item to keep in mind is that a data center with proper cable management and air-flow designs will help to keep your now more fully utilized servers from developing heat and fan failures.

posted April 25, 2008

There are many options to optimize power consumption and space cooling.in an IT structure.

The easy answer is to first ask you local utility to come out and do an audit. Most of them will do it for free as they too are being driven to reduce electric demand.

The next option is to invite 2 or 3 larger commerical mechanical contractors to evaluate your HVAC and powe usage and request they bid a performance contract.

posted April 25, 2008

as has been said virtualization is probably the best solution in most cases. blade servers also come to mind. thin clients are probably going to make a comeback with the advent of virtual desktops (beyond citrix).

cloud computing may be all the rage next also. it may not decrease the overall power consumption but for your local installation it may help. something similar to offsite or remote infrastructure.

posted April 25, 2008

There are many opportunities for optimizing data center energy use. If you treat the entire room as an energy system, it becomes easier to see the opportunities. Many of these can achieve savings with less cost or impact than changing servers or IT strategies. This is definately not a one-size-fits-all situation and one never wants to jeopardize the data center operations. However, here are a few possibilities:

Challenge the operating conditions. Some data centers I've been in are uncomfortably cool a good part of the time. Yet, when I've asked, I am told that the equipment is typically rated for operation at up to 90 F. Be clear on requirements for equipment vs. comfort cooling for staff. Every degree that the room temperature is raised represents some measurable savings. Similarly, research the humidity requirements. I have found reheat coils in ducts serving data centers. It would be very rare that reheat would be needed to prevent condensation since some reheat occurs naturally in the duct work. Also, improving air circulation can eliminate hot spots that might cause people to turn down the thermostat for the whole room.

Can the heat be recovered by using it to warm adjacent spaces? Some systems recirculate most of the air within the data center. By installing some duct work (with fire dampers and duct smoke detectors) and controls this heat can be recovered.

Uninterruptible power supplies generate heat. They should not be inside the server room. Place them in a warehouse or other uncooled or minimally cooled area.

Lighting is adding heat to the area. Data centers typically run 24/7 but that doesn't mean the lights need to. Remember, if you shut off 50 watts of light you are saving over 100 watts of energy. That is 50 watts for the light plus 50 watts to remove that heat plus the efficiency factor for the cooling. Leave a safe level of night lighting on but put the rest of the lighting on presence sensors of times.

Remember that idea of the total savings being the savings from turning off a device or moving it to an uncooled space being the total of the device and the reduced cooling load when you do the cost justifications.

Look at the equipment providing the cooling especially rooftop DX units. They can be real energy hogs, especially if they have been in use for a long time. Take a careful look at things like economizers, air leaks, and rain soaked insulation.

Most of these things can be checked by a facilities engineer or an HVAC engineer. They can evaluate the financial impact and make recommendations as well as execute any resulting project.

posted April 25, 2008

This is a topic which was of interest to me long before it became big news focus.

It ties in with other cost factors. In other words, if you fix one problem, that causes more cost some place else, you have not accomplished anything to help the bottom line.

A collegue of mine was IT manager ad a government agency with more than 100 PCs networked. He ran a test with different patterns of behavior for different groups of PCs, by department.

He found that PCs that were powered down when people went to lunch, took a break, etc. wore out faster, needed parts replaced more often, than any other scenario. Plus there was loss of productivity when people returned to their work area, waiting on the PC to boot up again.

The second highest cost area, was in PCs that were left running 24x7 ... stuff has an effective life span, that is much longer if the PCs are powered down outside the work day.

The longest life span, and lowest cost of maintenance were the PCs that people shut down at end of their work day, rebooted when return to work.

I have an automated message to co-workers that goes out 2 pm every Friday to remind them that their PC need not be powered up all weekend, that it will not need as much maintenance if it is powered down when they not at work.

More sophisticated network servers, and non-PC computers, have the capability to schedule down time. Some companies need 24x7 access to their data. Others are tied to a particular work shift, with occasional need for access in the evening. You can analyse computer logs, find out what time of day, wee hours, no one needs the computer.

Ours automatically powers down at 2 am. If anyone is signed on & active after 1 am, they get a message alerting them to the impending shutdown.

Ours automatically reboots itself at 6 am, so it is ready for business when the morning crew arrives.

Do you know how much electricity your server is using, when it is just sitting idle, no one using it? There's also heat exhaust, air conditioning.

Newer computer models tend to be more energy efficient than older ones, but there are trade-offs associated with any upgrade.

I believe that human productivity is more critical than saving money on computer resources provided to them. In other words if your people can get 3 times as much work done in the same time period, this helps the bottom line a heck of a lot more, than anything saved by skimping on computer resources supplied to them to get that job done.

Thus, there needs to be a focus on computer organization, computer security, how people get data inter-communicated.

Do an ABC analysis on the amount of hassle people have to go through to get their jobs done, for the purpose of eliminating the hassles, so they can get more work done in the same time period. Is there anything, the people are doing, that the computers can do for them?

posted April 26, 2008

Hi James,

I truly believe that at the end the question is not how to decrease power consumption but how to avoid increasing it year after year, at always larger rates.

From that point of view, I think that infrastructure changes in the direction of Gartner's Real Time Infrastructure or Extreme Transaction Processing with the underlying decoupling between applications, services and hardware, be it servers, networks or whatever, should help to optimize current resources and avoid uncontrolled growth.

posted April 29, 2008

James one way of significantly reducing your energy consumption in the Data Centre is to optimise the cooling that you have. Many DC's have up to 60% of their expensive conditioned air bypassing the servers. This is an expensive waste of energy. By recapturing this cold air with simple techniques such as sealing cable openings and blanking off open areas in racks it is possible to make significant savings. In addition to cost savings additional cooling capacity can be acheived. A 1% reduction in ambient room temperature will save 5% of energy costs. I have a number of white papers on this topic if you are interested.

posted May 1, 2008

Excellent question. We basically asked that same question about 18 months ago. It wasn't feasable for us to continue to expand our data center and add more servers. We went with virtualization technologies, VMWare and we took a long hard look at our SQL servers and found that not only could we combine more processes on to a single SQL server, but that we could do more with less.
With VMWare, we've taken the attitude that someone has to prove to us that they need their own hardware, otherwise they will get a VMWare client.
We've recently begun to look at several technologies that 'dial down' the power consumption during non-peak hours. Most server vendors are offering something like that today. VMWare has the ability to VMotion clients to another host and power down a host if it's set up that way.
As far as HVAC needs, 62 degrees is a lot more expensive than 74 degrees, take a look at how your data center gets cooled, sometimes just routing the heat out is often more effiencient than cooling more.

Links:

• http://www.vmware.com

posted May 1, 2008

First, it depends on the footprint. As always, different application based on requirements.

Virtualization and consolidation are easy gains. Often overlooked is maximizing efficiency. Move from farms and clusters to clouds and grids with power intelligence. Dynamic system resource allocation based on need can automatically power off things you don't need running and bring them back when you need them.

If you're operating in a computer room/data center that was built in the 90s or early 2k, you're probably using grossly inefficient CRACs that are bleeding dollars. Toss in some blanking panels and look to replace CRACs with in-row coolers and move to dense blade platforms to cut out excessive power supply bleed... in most facilities that can have significant ROI over 5-10yrs... even more since you can up the ambient room temp.

Understand your budget for the next 5yrs and work backwards. Consulting with experts from the likes of APC and Liebert, who have awesome professional services, can help you set a goal based on your unique situation. They'll also help develop a plan by which you can execute. Then all you need is a competent staff that can deliver.

Rarely will you find a resource with extensive experience at both IT Operations management and facility management that is worth a salt and not a contractor running around the country doing this at a premium. If you have one of each and need a bridge, find a vendor you can trust and let them help you tailor a solution.

posted May 1, 2008

This question is not only pertaining to IT Companies but all Corporate house need to review their Energy Consumption. There are technologies available that not only bring down the Energy Consumption but also help maintain a more congenial and healthy environment for the occupants.
For a typical Commercial Building, 30% of the power consumption is due to the HVAC( Heating Ventilation & Air Conditioning)products.
In a typical IT Park, where the work stations of the employees are quite densely located, it becomes imperative to provide sufficient amount of Fresh Air. According to international norms(ASHRAE) and our National Building Code an office space must have minimum 17CFM ( Cubic Feet per Minute) of fresh air for each person. Most Air Conditioning provides only recirculated air. This is extremely unhealthy and causes BRI( Building Related Illnesses) such as Asthama, Legionarries Diseases, Lung Cancer etc.due to Indoor Pollutants like VOCs, Radons, asbestos etc. Myth is that outdoor Air is more Polluted but its been proved that Indoor Air inside airconditioned environment can be 10-100 times more polluted. This happens because Air Conditioning works on a closed loop of Air. Its the same recirculated air that we breathe inside AC environment and the Concentration of the pollutants like VOCs from the paints, Formaldehydes from the furnitures,Radons from the floor, VOCs from the Printers, Photostat machines etc..keeps increasing.The occupants land up breathing all that. Now to bring in freash air it would mean increasing the Load to the AC but if we use Treated Fresh Air or Enrergy Recovery Ventilators, the Air coming inside would be PreConditionined using the enegry of the conditioned ,polluted exhuast air. These devices can bring down the fresh air load by 65%- 78% which is a huge saving in the capital and running cost of the HVAC system.

Links:

• http://www.drirotors.com

posted May 1, 2008

Companies manufacturing end product should consider electronic components which operate on low power and consume very little power to make the batteries life longer and safe the green environment.
Designers should consider designing projects towards low power consumption, i.e. switch off components or reduce the running speed when not used, then only wake up when needed and sleep again.
Semiconductor manufactures should also consider their manufacturing and production processes to benefit the green environment (be careful with waste) and low power consumption.
All these combined together would consume less power in IT infrastructure, hence save energy costs and safe the green environment

posted May 2, 2008

I think Al MacIntyre brings up a host a valid and correct points, and the overall answer in my opinion is to use a virtual service with thin clients for your users (where possible).

If a thin client suffers from any kind of problem, it can be swapped within minutes and productivity doesn't suffer. Also, thin clients + monitor consume 50% less power than a typical PC.

My company, Nasstar plc, provide a Hosted Desktop service, and it's provided our clients with the answer to decreasing their own power consumption at the same time as aiding with employee productivity (and therefore improved morale).

If you've got 100 companies all using their own server hardware locally, then outsourcing it to a single SaaS provider's Data Centre, then you've dramatically reduced the overall carbon footprint between those companies.

posted May 2, 2008

All are good answers but the absolute breakthrough will come when we get rid of the CPU clocks. Modern day Data Centers have a zillion CPUs each zipping line a sprinter on amphetamines at Mutli Giga Hertz speeds spending natural resources providing virtual results.

All these hertz need energy vast amounts of it, managing CPU speed is a measure but doing away with it using clockless designs will be a quantum leap forward.


There are a few institutions working on this right now like ARM and FORTH , check out http://www.ics.forth.gr/carv/aspida.

To take it a step further just think that every electronic device today like phones,ovens, hard disks , toys etc. uses one or more CPUs ...

posted May 2, 2008

Virtualisation is the obvious hot topic as many have quoted already.

Other developments such as Intel’s VPro chip allow for machines to be on for less time when not being used.

I would also say though that going back to basics such as making sure user / employees power down screens and don’t leave them on standby can really help with power consumption.

Also, if you really want to get deep into this debate then it can be argued that giving users the ability to work at home can reduce the travelling they have to do and thus reduce their cost to the environment.

posted May 2, 2008

Companies can dramatically reduce their power footprint through virtualization. Server, storage and desktop virtualization technologies offer companies more flexibility while reducing power consumption by as much as 60-70 percent.

We are working with an organization right now that is looking to replace over 700 desktops with solid state desktops, virtualize their server and storage environments... bringing a total savings of $250,000 year over year, and that is just the power savings.

Efficiency is the name of the game... doing more with less. I would highly recommend looking at DataCore, VMware, and Pano Logic for an immediate impact on your environment.

Additionally, I would recommend getting a datacenter efficiency assessment done. This will provide insight into other optimizations regarding the support infrastructure (such as HVAC, UPSs, etc...)

Lastly, I read an article recently where it talked about a company using the heat generated in the datacenter to heat parts of the building and melt ice on the side walks in the winter time. Innovation can also payoff nicely.

Hope this helps.

Links:

• http://www.vmware.com
• http://www.datacore.com
• http://www.panologic.com

posted May 2, 2008

James,

I would answer it from an IT perspective with business effectiveness. I believe this would answer the question appropriately. We are an IT services company. We are working with an energy management company in US who had problem in energy consumption. They got in touch with us and we offered a remote monitoring solution and a data center solution which has helped them overcome their problem.The IT infrastructure is the backbone of a company containing network elements like desktops, servers, storage devices etc., Monitoring and managing it properly will make your effective. You also get real time information.

Feel free to contact me for further information.

Thanks,
Murugan,
muruganandan.k@e4e.com
1-800-334-8312

Links:

• http://www.e4e.com

posted May 2, 2008

Industry is beginning to respond on this as well. I recently saw a review for a desktop PC from Lenovo - their Eco line. It was rated at 57w when running, 39w idle and 1w in standby if you wanted to do remote management off hours. Base model is about $900.

Links:

• http://www.crn.com/hardware/207400835

posted May 2, 2008

I have done a research on this filed, and this is how we will do it (I don't say it's a best way, just our way):

1. By introducing virtual servers we will decrease the number of machines. Actually, we will decrease from 17 servers to 4 - 1+1 on the primary location and 1+1 at the backup location
2. By introducing storage servers we are able to manage power accordignly to our need - when we need additional space, we just pitch a new disk, not before
3. By removing 15 servers from the game our demand on the cooling is much lower, so our climatisation devices are working on half power - one will triger only if the temperature raises, not before.
4. We have switched all our desktop computers with new one, and all our monitors with new, power efficient ones. Each computer consumes approx 35% less power, so it's a good saving.

posted May 6, 2008

The obvious answers are consolidation of servers, virtualization and low power Racks. I would suggest a different approach. If companies are serious about the environmental impact of their IT infrastructure, they can first create a baseline of the carbon foot print of their existing operations. There are many ways to do it , lots of outside consultants are there to help regarding this. This is the litmus test of the commitment as there are many ways the Carbon impact can be measured, for example it can take into account the impact of the commute of the employees to the office which will be a measure totally unrelated to the IT operations but will incur carbon cost. Once they get their base line carbon foot print which will be a comprehensive measure of the IT shops energy consumption, heat generated, etc, etc they can setup a carbon foot print reduction road map which will help reduce the power consumption.
Small changes in behavior have lasting impact. some of the direct measures can be replacing Cathode ray monitors with Low power LCD monitors, Changing desktops to Laptops, implementing enterprise wide group policies on power management (standby, hibernate and monitor off, etc). Moving server database and data ware house batch runs to off peak times abd so and so forth.

The bottom line is no one change is sufficient, there needs to be a comprehensive attempt to create a green zone which will produce the long term ROI for the IT organization.

posted May 6, 2008

There is relatively inexpensive technology that connects to the power source (panel or cut off). It not only reduces overall consumption, but can also provide harmonic filtration and surge protection. Payback is typically very attractive as well.

Peter Lytle is my business partner on energy matters, and I think you would find him very conversant in the technology and very helpful and responsive.

posted May 7, 2008