"green", solid state server / cluster / grid compu

Ecological issues around computing. This is an experimental forum.

Moderators: Ralf Hutter, Lawrence Lee

Post Reply
IsleVegan
Posts: 5
Joined: Tue May 15, 2007 4:16 am

"green", solid state server / cluster / grid compu

Post by IsleVegan » Tue May 15, 2007 4:45 am

I am interested in putting together a "green", high performance, fanless computer utilizing such components as solid state / industrial grade ide flash disk such as this:

http://www.transcendusa.com/Support/DLC ... =TS8GIFD25

or perhaps 2 or more 2GB variants of the above used with the built in raid controller on many of the AMD Opteron 2000 or 8000 compatible motherboards.

Would such a motherboard and processor combination work with any mCubed fanless heatsink cases and power supplies?

http://www.hifiatx.com/

As the sustainability committee student congress member at my community college I have been advocating energy efficient, solid state, RoHS compliant, fanless, open source solutions for expanding computing resources on campus.

Recently I was able to get approval to purchase one server and one thin client for a total maximum of $3,000 USD. For several months I have been advocating for a change in the paradigm of computing resources.

The objective is to create a technology demonstration pilot project at my community college to show how we can reduce heat output, energy and space requirements, noise, toxins from manufacturing (larger computers),

I like the client-server cluster idea, wherein there would be 1 high-end server and several lower end clients all connected as a cluster so that system-wide resources could be allocated dynamically as needed.

Any suggestions here - cluster / grid software for linux seem so complicated but I could probably get help from the Hawaii open source education foundation and the University of Hawaii tech folks.

Based on what I have read it seems that the Opteron gives the most performance per watt as well as a good bit of room to upgrade with the forthcoming 4 core drop-in replacements, direct connect and hyper transport technologies.

I may be wrong on some of what I have said, so please feel free to correct me as necessary.

The main objectives are good value, expandability and all of the aforementioned sustainability factors are a must-have.

What might you recommend to accomplish my objectives within 3k? Initially I'll use second hand monitor(s), keyboard(s) and mice or borrow as necessary and request further money from the student congress board once I can show the project to be a success.

Thin-client recommendations, comments and info are welcome as well.

I thank you for your time and attention.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Tue May 15, 2007 7:31 am

It's great to hear you're actively advocating awareness and changes to make IT greener at your college. You're obviously aware that green is more than energy efficiency. It's also....
  • Reduction of use of hazardous substances -- RoHC
  • eco-materials selection
  • ecological design for end-of-life (ID of different plastics, friendliness to shredding machines, no bad paints or coatings, easy disassembly, use of recycled parts, etc)
  • design for longer life
  • take back service? (at end of life)
  • eco-packaging
IMO, there's little about http://www.hifiatx.com/ that's environmentally friendly. It uses a massive amount of AL, which is probably not recycled AL. It's also packaged as a high end product -- which means there's lots of it.

Chances are, the most eco-friendly cases are generic ones made of light steel (4 times less energy to produce than aluminum, per weight) with minimal plastics and cheap (minimal) packaging. Assuming other things are equal, the lower the weight, the better.

Remember the oft-repeated mantra -- reduce, reuse, recycle -- in that order. I'd look for 2nd hand gear for everything, personally, and swap out only the components subject to wear & tear (if they are old): HDDs and fans, maybe the PSUs.

My guess is that a standard HDD actually has LESS embedded energy/raw materials than a solid-state HDD. RAM (and other silicon-based chips) are very high energy / raw materials intensive products.

IsleVegan
Posts: 5
Joined: Tue May 15, 2007 4:16 am

what about the energy requirements of operation

Post by IsleVegan » Tue May 15, 2007 2:50 pm

normally i'm the one pointing out the importance of the toxins released and energy requirements of manufacture but in this case i'm wondering if over the life of the product, might the much lower energy requirement to operate the tiny solid state drive become better with time while the standard hard drive continues to use so much more energy.

it's likely these computers will be left on a lot of the time, when it goes beyond my control and since it is intended to be a server.

further, my objective is a fanless system, so i'm wondering about the balance of energy from fan operation compared to a larger heatsink case. there is also the issue of failure. i've had more than one fan and hard drive fail but haven't had solid state components do that.

i understand that there is a limited write life for solid state disks but i would set it up so that users would write to their own usb devices or direct them to an online service.

this is one server that i'm proposing, to service as many thin clients as possible.

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Tue May 15, 2007 8:18 pm

normally i'm the one pointing out the importance of the toxins released and energy requirements of manufacture but in this case i'm wondering if over the life of the product, might the much lower energy requirement to operate the tiny solid state drive become better with time while the standard hard drive continues to use so much more energy.
I strongly doubt it.

If you go for a notebook drive, whose capacity is now up to something like 250GB, the power draw is typically 1W, and 3W at max load. I think this compares favorably with solid state drives. There are 7200rpm 2.5" drives that will do this.

With regard to fans... a single medium or low speed fan will not draw more than 2~3W; usually less. It's hardly an issue. No contest between that and a massive aluminum case for environmental issues. Also, my bet for longevity would be on a fan cooled system over a fanless one. I speak from experience. A single quiet slow speed fan can easily drop CPU and VRM temperatures by over 10C. This temp. difference in 24/7 use over years will mostly definitely have an impact. And when a good fan is run at 1/2 its rated 12V rpm, well, you can certainly count on it lasting a lot longer than usual.

Also, the reality of energy consumption during the use phase vs energy used during manufacture is quite surprising for most people: For desktops, >75% of the energy (for the product's entire life cycle) occurs in manufacture. For a server, it will be less -- perhaps 40%? 50%?

Simple fact: Electricity consumption during the use phases of the desktop is typically only about 15-20% of the total energy in its life cycle. For a server, it might be double or triple that.... but the differences in energy consumption between fanned and fanless, and SS drive vs. notebook drive are extremely small no matter how you look at it.

People are rarely exposed to the noise of a server, which needs to be in a secure location. You can make it quite quiet without going fanless, and the end result will not only be much less costly, but more reliable and far more ecological.

IsleVegan
Posts: 5
Joined: Tue May 15, 2007 4:16 am

Can you provide some documentation?

Post by IsleVegan » Thu May 17, 2007 3:23 am

I would love to go with the best choice and it would help my case to have some definitive, reproducible, documented / peer reviewed evidence for these claims.

These are college students and college admin, so we look at these kind of things.

Mahalo for your time and attention!

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: Can you provide some documentation?

Post by MikeC » Thu May 17, 2007 7:00 am

IsleVegan wrote:I would love to go with the best choice and it would help my case to have some definitive, reproducible, documented / peer reviewed evidence for these claims.
All the info I cited comes from professionals who work in environmental issues around electronics. It's all published stuff... though not necessarily in the public sphere. A lot of it came from the 2007 IEEE International Symposium on Electronics and the Environment in Orlando, FL, which I attended last week; most of this has not yet been released to the public in any form. Some of it came from the book, Computers and the Environment, a collection of studies. I quote directly from this book:
ch.3 -- Environmental Impacts in the Production of Personal Computers by Eric Williams wrote:But whereas nearly all of a refrigerator's lifetime fossile fuel consumption (96%) is due to the electricity produced to run it and a small amount for production, for a computer the situation is reversed; a relatively small proportion (25%) is for use, while most (75%) is for production. In contrast with other appliances such as refrigerators, where improving the electricity efficiency during operation is the main strategy to reducing energy use over the life cycle, for computers it is very important to save energy in the production phase and increase the life spanof the use phase.
On the site where the book link is, you will find many informative articles, data and links.

torrent11
Posts: 8
Joined: Tue Mar 27, 2007 1:18 am

Post by torrent11 » Thu May 17, 2007 10:08 am

I'd like to re-echo the point about using second hand gear, saves money too...

As others have pointed out in other posts these days it's the software that seems to be the performance bottleneck, look at how you can use efficient software to improve the performance of existing systems.

Cheers

Steve

IsleVegan
Posts: 5
Joined: Tue May 15, 2007 4:16 am

reusing hardware

Post by IsleVegan » Thu May 17, 2007 7:09 pm

I plan to use an existing case and perhaps in the future have a case built out of reused wood / expanded metal / mesh or similar materials by one of the community college metal / wood working departments on the island.

What kinds of software solutions might you recommend to improve performance? I'm not familiar with the threads / discussions you mentioned.

How difficult is it to create a system that cannot write to the drive units in normal use, except when an admin needs to do an OS update...? I was thinking that there might need to be a virtual memory / temp / log storage drive unit but never tried these things.

Once I choose the hardware I will take more time to explore the software options, since there is less than 15 days remaining in this fiscal year.

Thanks again for all the thoughtful responses.

scdr
Posts: 336
Joined: Thu Oct 21, 2004 4:49 pm
Location: Upper left hand corner, USA

Post by scdr » Sun May 20, 2007 12:21 am

How difficult is it to create a system that cannot write to the drive units in normal use, except when an admin needs to do an OS update...? I was thinking that there might need to be a virtual memory / temp / log storage drive unit but never tried these things.
You might want to take a look at Puppy linux for ideas on this.
http://www.puppylinux.org
File system designed to be mostly read only, and to minimize writes
to storage device. (e.g. run off of CD-R, writing back to CD-R or flash drive.)
For overview of how they do it:
http://puppylinux.org/docs/?Getting_Sta ... t_of_Puppy

(Sort of like using Virtual PC with a base drive image and another
image file storing changes.)

Puppy is mainly aimed at having a small image (running mostly in RAM);
if it doesn't do what you want, you might be able to borrow some of the file system ideas.


If you haven't already, it might help to outline what this system is supposed to do.
(Hard to suggest hardware/software without knowing what the job is.
Or maybe I missed it in reading the thread.)

Post Reply