What I found more interesting though, is the article the author linked to in the beginning: “Best Practices For Data Centers: Lessons Learned From Benchmarking 22 Data Centers”

This article has some excellent information, as the authors spent several years benchmarking 22 different data center buildings. From this benchmarking, it was found that data centers can be 40 times more energy intensive as normal office spaces. Using the benchmarks, they created a set of “best-practices” for data centers.

These “best-practices” include:

• Improved air management
• Emphasizing control and isolation of hot and cold air streams
• Correctly size central plants and ventilation systems to operate well at inception, and with ability to scale
• Optimized central chiller plants
• Central air handling units
• “Free Cooling”
• Alternative humidity control
• Improved UPS systems for battery backup
• High efficiency computer power supplies
• On-site generation with special chillers to cool waste heat
• Direct liquid cooling of racks and/or computers
• Lowering standby losses of standby generation systems

Overall I would say the article was well written and brings up some important points. Check it out if you have a moment.

Here is the HotHardware.com article on the WD Caviar and RE2 GreenPower 1TB HD’s

“WD’s GreenPower platform is the first 3.5-inch hard drive platform designed with power savings as the primary attribute. GreenPower drives from WD deliver average drive power savings of 4-5 watts over competitors’ drives, ultra-cool, quiet operation and solid performance.”

Here is an excerpt from the HotHardware article:

“With its high MTBF (mean time between failures) rating of 1.2 million hours and RAID-specific optimizations, the WD RE2-GP is aimed at the enterprise market, while the WD Caviar GP is marketed towards the consumer desktop segment. A very interesting feature of the GreenPower drives is IntelliPower, which is a “fine-tuned balance of spin speed, transfer rate, and caching algorithms designed to deliver both significant power savings and solid performance” according to WD. What this means to potential users is that WD isn’t telling us the exact spindle speed of these drives. We know that they are likely spinning at a speed between 5400 and 7200 RPM and that each GreenPower model may use a different, invariable RPM. So, while WD made power the priority with the GreenPower platform, it did so without disregarding solid performance, a wise choice in our opinion.”

I would have to agree with HotHardware and Western Digital on this one. In an enterprise solution, we cannot degrade performance too much. Hopefully they will continue to build on solid performance with still more power efficient products.

Here are the published power consumption numbers:

WD Caviar GP 1TB (MODEL: WD10EACS)

Read/Write: 7.50 Watts
Idle: 4.0 Watts
Standby: 0.97 Watts

WD RE2-GP 1TB (MODEL: WD1000FYPS)

Read/Write: 7.40 Watts
Idle: 4.0 Watts
Standby: 0.97 Watts

HotHardware did a very nice test with this hardware. The specs on the test system:

“When testing the WD Caviar GP and RE2-GP terabyte SATA hard drives, we used an Intel Core 2 Duo E6400 (2.13GHz) on an Abit Fatal1ty FP-IN9 SLI motherboard. We also used 2GB of Corsair DDR2 (TWIN2X1024A-5400UL) and a 120GB Maxtor SATA hard drive as the main system drive. The 1TB WD Caviar GP and 1TB RE2-GP were compared to a 1TB Seagate Barracuda ES.2, 750GB Seagate 7200.10, a 750GB Western Digital Caviar SE16, a 500GB Western Digital Caviar SE16, and a 74GB Western Digital Raptor.”

HD Tach 3.0.1.0 and HD Tune 2.54 were used for the performance data, and HotHardware had some very positive things to say:

“Right off the bat, we are impressed by both of these GreenPower drives, especially considering their sub-7200 RPM spindle speeds. Thanks to big platters and some internal tweaking by WD, the Caviar GP and the RE2-GP put up Average Read and Average Write scores that are very respectable when compared to the 7200 RPM and 10k RPM drives listed in the charts. There must be some additional performance tweaks in the RE2-GP as it manages to edge out its Caviar cousin.”

“The 1TB GreenPower drives can’t compete with the 1TB 7200 RPM Seagate Barracuda ES.2 in HD Tune, but they do manage to hang quite well with (and sometimes beat) the other drives in our testing. The random access times of the Caviar GP and RE2-GP were the two slowest of the bunch, but we are still impressed by these energy-conscious drives so far. We should also note that the RE2-GP once again nudges past the Caviar GP.”

And last but not least is the info I am interested in. HotHardware used a P3 Kill-A-Watt meter to get their readings, just like I used for the workstation and server data.

“We used the P3 Kill-A-Watt to measure total system power consumption at idle and at load (which consisted of running some intense disk benchmarks).”

“When we compared the 1TB Seagate Barracuda ES.2’s specifications to these two 1TB GreenPower drives, we noticed that the 1TB ES.2 requires about four more Watts of power for both idle and seek modes. With this information in mind, we weren’t surprised to see how the power consumption results with the 1TB Caviar GP installed compared to those of the 1TB ES.2. The system pulled four less Watts during load and five less Watts at idle with the Caviar GP installed versus the ES.2. What surprised us was that the system used even less power when the 1TB RE2-GP was tested.

We realize that in the big scheme of things 5-10 Watts may not matter to many people, but you have to keep in mind that the RE2-GP is intended for large data servers. The power savings, and resulting cost savings, can really add up in a data center with hundreds of drives running 24/7 for a whole year.”

For you skeptics, the caviar GP has a 3year warranty, and the RE2-GP has a 5year.

WD has a money savings calculator on its web site to show how much deploying RE2-GP drives in a server environment can save.

HotHardware.com has this to say on the pricing:

“The 1TB Caviar GP currently costs about $230, and the 1TB RE2-GP runs about $250, according to our PriceGrabber search engine at the time of this writing. These prices are fair, as they fall in line with or are lower than the prices of other 1TB drives on the market.”

The HotHardware.com article referenced is:

WD Caviar & RE2 GreenPower 1TB HDs

Sun Netra T1/105:

• 1 * 440mhz SPARC64
• 512MB RAM
• 1 * 18GB Seagate ST318405LC

Dell PowerConnect 3024
• Managed 24port switch
• 1 port lit

Cisco Catalyst C3512XL
• Managed 12port switch
• 0 ports lit

Sun Ultra 80:
• 5 * 450mhz SPARC64
• 4GB RAM
• 2 * 73GB Seagate SCSI Ultra160

SGI Octane:
• 2 * 300mhz MIPS R12000
• 2G RAM
• 1 * 9GB (?) SCSI Ultra160
• 2 * 73GB Seagate SCSI Ultra160

Dell PowerEdge 2650:
• 2 * 2.2ghz XEON
• 1.5GB RAM
• 2 * 18GB U160/320 SCSI drives

Dell PowerEdge 2850:
• 2 * 3.2ghz XEON
• 2GB RAM
• 2 * 73GB U320 SCSI drives

Sun Netra T1/105:
• 1 * 440mhz SPARC64
• 512MB RAM
• 1 * 18GB Seagate ST318405LC

Dell PowerConnect 3024
• Managed 24port switch
• 1 port lit

Cisco Catalyst C3512XL
• Managed 12port switch
• 0 ports lit

Sun Ultra 80:
• 5 * 450mhz SPARC64
• 4GB RAM
• 2 * 73GB Seagate SCSI Ultra160

SGI Octane:
• 2 * 300mhz MIPS R12000
• 2G RAM
• 1 * 9GB (?) SCSI Ultra160
• 2 * 73GB Seagate SCSI Ultra160

Dell PowerEdge 2650:
• 2 * 2.2ghz XEON
• 1.5GB RAM
• 2 * 18GB U160/320 SCSI drives

Dell PowerEdge 2850:
• 2 * 3.2ghz XEON
• 2GB RAM
• 2 * 73GB U320 SCSI drives

Intel announced “Sossaman” on Tuesday. This is a “low power” dual-core Xeon based 32-bit processor. Because of the 32-bit catch, HP won’t use them. IBM, however, is ramping up production. This new chip uses a max of 31 watts of power compared with 110 to 165 watts for other Xeon models.

We will see how these go up against AMD’s Opteron’s.

Several studies have been conducted showing that when idle, Opteron-based systems tend to use about 40 percent less power than similarly equipped Woodcrest-based systems.

AnandTech’s Low Power Server CPU Shoot Out has some very good info.

Also check out this InfoWorld article saying that AMD is more power-efficient than Intel.

If you are interested, Neal Nelson& Associates has published a white paper on the subject of standardizing test methodology to measure and report the power consumption and power efficiency of modern server systems.

More on this later.

The Dell Optiplex GX150 had a peak during bootup of .63Amps (~69Watts). I collected some data while the box was booting (in Amps):

.57 .53 .58 .55 .57 .40 .38 .33 .41 .52 .58 .4 .5 .41 .5 .52 .4 .5 .61 .4 .61 .63 .53 .61 .63 .44 .43 .36

The Dell Dimension 4700 was quite a bit more power hungry. Peak was 1.58Amps, and it stabilized at 1.1Amp. On this one I was able to leave the meter connected for a while. From 8am-4pm Saturday, it used 0.72 KWh (not counting the monitor). From 8am Saturday to 8am Sunday, it used 2.1 KWh.

When in “Standby” mode, the Dimension 4700 draws .04Amps.

The Dell 1704FPTt LCD monitor that I tested had interesting results as well. When I powered it on, it was sucking .5Amp which quickly dropped to .01Amp with the monitor in “Standby” mode. When the computer hooked to the monitor was turned on, the monitor started to “wake up” and hit a peak of .38Amps. The power consumption stabilized at around this point and was about .37-.37Amps for the duration of the monitoring.

I was kind of surprised that the LCD took as much power as the 1Ghz PentiumIII Optiplex, but this is one of the models that was purchased shortly after “The Great LCD Boom,” so power consumption probably wasn’t at its best.

When showing these numbers to someone else, I was asked how much this was, in Watts. You can easily do the conversion yourself using the formula: Watts = Current * Voltage. In this case my voltage was ~110V. The only reason I don’t have the numbers from the Kill-A-Watt meter, is because I was trying to accurately get as many data points as possible while the machines were booting, and I did not want to switch between amps/watts as this would have messed me up.

Hopefully these numbers are useful to someone, if only to show how much juice these things are sucking. We have over 20 of them here in the office, and who knows how many are in service throughout the world. With any luck, someone will see these numbers and decide to go with a low power solution.

They go very in depth into every aspect of building a more efficient system. Mostly geared towards gaming but also good server info in there. I’ll come back and expand on this some more later, just wanted to throw this out there.

Of course I was stumbling around on Google the other day, and I found the Linutop. This thing is cool. The cheaper slower one uses less than 5 watts! It still has 256MB of RAM and an AMD Geode LX700 (x86) processor. The storage is a little small at 1GB using an external USB key, but if this is being used as a workstation, you shouldn’t need to download mp3’s and movies all day

They come with Linux and various applications preinstalled. Here is what their site says:

“The Linutop operating system is stored on an internal flash memory and cannot be altered by virus or mishap.

Initial state is recovered at each restart.

 Linutop is delivered with free standard ready-to-use Linux software: Firefox, Open Office, VLC Media player… and can be easily customized with additional software.”

The more robust model is new, and has 512MB of RAM (expandable to 1GB), an AMD Geode LX800 (x86). The storage is still 1GB external USB, but this model still uses less than 8 watts of electricity, and there are no moving parts in either of these things.

Sound and video are built in, as is ethernet and USB (4x USB 2.0 ports). They come with a strong aluminum case, and a 1 year warranty extendable to 3 years.

My favorite part aside from the no moving parts aspect, is the bracket they have to mount them on the back of a standard LCD monitor. Go to the Linutop website and check it out. They are a nice small computer ( 14 x 14 x 3.5 cm (5.5×5.5×1.38 in)) that won’t use a lot of power.

I know this probably sounds like a sales pitch, but I am not affiliated with these people in any way. In fact, here are the things I’m not impressed with:

• The processors are a bit slow for anything intense
• The warranty is a bit short at only a year, for something that seems so durable and has no moving parts
• They are expensive especially with the dollar sucking so hard
• Not very much storage if you did want to use it as a home computer (external USB drive could be employed)

The benefits seem to be in greater number than the drawbacks, but I’m rather poor so the cost and processor speed are the real deal killers for right now. Otherwise I’d get one to play with. I’d like to get some numbers with a Kill-a-Watt and one of these hooked up to a 14″ LCD used for a standard 8hr shift of decent load. Depending on electricity costs and desire for efficiency, it might be a cost effective solution.

Either way, check it out and see what you think. That’s why I put this stuff here.

“Saturday, December 29, 2007

Greener Computing Tips from Lifehacker

Lifehacker have come up with a list of 10 straightforward but effective energy saving tweaks all computer users can try.

Source:hippyshopper.com/2007/07/greener_computi.html

Posted by Green PC at 1:36 AM  “

From echannelline.com:

“15 August, 2007
By Paul Weinberg

A new “green,” 20-port 10-Gigabit Ethernet switch from Blade Network Technologies is being installed on Hewlett-Packard, IBM and NEC blade servers. William Terrill, associate senior research analyst at Info-Tech Research Group, described the consumption of only 25 to 30 watts on the new Blade switch “as dramatic,” compared to competing internal blade Ethernet switches from Cisco

“With an external switch that has the same number of ten gigabit Ethernet ports you are probably talking about 100 watts, minimum. Even the new Fujitsu, external switch which has 20 ports, ten gigs, generates about 200 watts.”

Terrill noted that Cisco “owns” the network infrastructure, which has forced the upstart Blade to focus on internal blade Ethernet switch technology.

“Blade believes blades are where most people in the data centre are going to be for their data center technology.”

Blade has positioned itself as a vendor of 10-Gigabit Ethernet switches at a time when one gig Ethernet is running to “the upper limits of the performance,” stated Terrill.

The 10-Gigabit from Blade is the first Ethernet switch that offers both connectivity to the processor blades within blade servers and 10-gig capabilities for the external network, he continued.

One gig may have reached its level in terms of performance in the transferring of data among blades.

“One gig, two or three years ago was overkill. But now with virtualization and the new multi-core processing, you are beginning to reach the limits of one gig Ethernet for a lot of activities, a lot of applications.”

This potential overloading of the network with one gig represents a direct result of virtualization where there is a major push to keep those processors on the blade servers running at 80 per cent utilization, Terrill stated.

“With multiple applications accessing the network simultaneously, one gig could be a major bottleneck.”

Terrill also predicted a lengthy transition for the full adoption of the new green blades in existing and new HP, IBM and NEC blade servers.

Meanwhile, Vikram Mehta, president and CEO of Blade, suggested that much of the discussion around power savings and increased cooling has been centered on servers.

But, he maintained, “It is not just servers. There is a whole bunch of networking equipment as well that adds to that problem.”

Blade is also aggressively pricing its 20-port 10-Gigabit Ethernet switch at $9,500 (U.S.) or $475 per port.

The company has also set up a new channel program to encourage channel partners to sell its software upgrades for such applications as manageability, performance, security and various new functions under server virtualization to customers.

Mehta estimates there are more than 200,000 Blade Network Technologies blade server switches in use today among end user organizations.

“There is an opportunity for these channel partners to provide consulting, integration, network planning and design.”

Meanwhile, Terrill told eChannelLine that addition of NEC as new OEM customer for Blade is an opportunity for it to make gains in the lucrative Japanese market where “NEC is huge.”"

Sun claims this thing has all sorts of bells and whistles..

“At A Glance

• Record-breaking performance, unequaled energy efficiency, and space-saving compute density
• The best rack-optimized 1RU server for massively horizontally scaled web environments
• UltraSPARC T1 processor with CoolThreads technology offers six or eight cores with four threads per core
• Delivers more than five times the throughput per rack of Xeon-based servers running Java applications
• Over 2x throughput of a dual processor Xeon processor server with 5x the server efficiency measured by the new SWaP (Space, Watts and Performance) metric
• A range of Linux distributions have been certified for the UltraSPARC T1 processor. Please click here for more information.
• Available as a Sun System Pack, which combines the Sun Fire T1000 with the right services as one offering“

You can see the performance benchmarks here. They claim that “Sun Fire & Sun SPARC Enterprise Servers with CoolThreads technology deliver breakthrough levels of performance with dramatic power and space savings, as demonstrated by the raft of World Record benchmarks.” Sounds good to me if it is true.

There are 4 reviews on the site currently, scoring an overall 3.5 out of 5. Here are some bits of the reviews:

“By Ben, the Mortician from Los Angeles, CA on 8/22/2006

Best Uses: Application Servers, Web server, Systems Integration

Pros: High Performance, Highly Scalable, Amazing Ruby Performance, Good Value

Cons: Not for every Workload
Prooved to be an excellent platform for deployment of Ruby on Rails, Apache web server, and Tomcat.
The box performed excellently, as it’s threaded architecture consided well the process based model in which our application deploys with. It has the best price/performance of any of the other boxes we tested.

Also suprising, it provides great Netbackup thruput and handles threaded system integration apps (Splunk for example) very well..

That said - for certain apps (the database part of our application, image processing, etc.), it performed poorly. Hence why it runs on Opteron.

With this boxes, it’s all about picking the right tool for the right job. A philips versus a flat, pipe wrench versus a hammer…. you get the idea.”

“By Colofoxyus! fromMexico on 8/8/2006

Pros: Highly Scalable, High Performance, Good Value

Excelent! Cool Threads are the best!”

“By Jeff Schave […]from USA on 8/7/2006
Best Uses:Web serving, E-mail, DNS
Pros:Good Value, Small Footprint, Highly Scalable
Cons:No second power supply
The T1000 now ships with either 2×73 GB SAS disk drives or 1 80 GB SATA drive. SAS drives will offer much better performance for some types of applications, and the second disk drive will reduce a point of failure in this model.
About the only thing this box is missing is a second power supply.”

“By Thorleif Wiikfrom Berlin, Germany on 7/24/2006
Best Uses:Web server
Pros:High Performance, Small Footprint
Cons:Only one Hard Disk

The Sun T1000 might be a cool product. But why is there only one Hard Disk Slot?
It would be a great product with two hot pluggable 2,5 SAS drives like the sun Fire X4100.”

Sorry folks, but I think the last thing someone designing a low power server is going to include is more power supplies and more hard drive slots.

One of the things that makes me most happy on Sun’s site for this product though, is the neato Sun Fire T1000 energy use calculator that allows you to see how much energy different configurations of the Sun Fire T1000 will use. This includes different memory, processor, and hard drive configurations.

End Note: I’d buy a crate full if I had the cash