MAY 23

One other thing that would help with Data Center energy consumption is if some of the heat to piezo to electricity generators (http://www.sciencedaily.com/re...225026.htm)
were hooked up to the waste heat coming out of data centers. Data centers have to constantly worry about waste heat and expend a huge amount of energy on AC. This would hit both the heat sink, and energy consumption birds with one stone

Real-time data compression reduces the amount of data written to storage devices and thus reduces CPU, disk, memory and network utlilization on the storage system. It can do this through its patent-pending algorithms that allow write and read operations from any location within the file while avoiding the need to decompress the whole file.

This smells more of snake oil than anything else. Aside from the impact on disk access times that the compression would have, it will also require more hardware to handle the compression/decompression of the data. This sounds just like whole disk compression, which generally has wiped out any gains as it has been performed by the CPU and so increases CPU cycles.

Wholelly Crappolla Batman lets invest all of the precious data that mankind has into this patent pending technology

Sorry, I don't buy the PR hype. You need computing power to compress and decompress the data which will eat up power. You can save on storage space, but in most data centers, the disks spin down when not in use anyway.

Wow. Someone didn't think this through very well. Compression = more CPU time = more electricity. Duh.

There's only one place where compression can lead to net energy benefits - and that's in data transmission. Especially for HTTP traffic. Turn on mod_gzip and your HTTP data will shrink dramatically, at limited expense of CPU power (especially if you use a hardware accelerator). This results in a benefit because, face it, when you visit a website the data often has to travel thousands of miles. Every bit saved is in effect a bit of energy saved on a dozen routers, hundreds of switches and many other devices.

"Compression = more CPU time = more electricity. Duh"

It's much more complicated than that. Obviously, data compression requires a net increase in CPU power draw, but you are gaining back all the power required for many HDDs and raid controllers that you would have replaced as the data would be compressed. In that case, it would be fairly straightforward to figure out the NET reduction in total energy expended after gathering all the data on processor power usage, HDD power usage, how many units of storage could be removed, etc.


Now, using normal system CPUs to calculate the compression/decompression algorithms is very inefficient. A much better idea would be to
add low-power, custom designed chips that implement the compression/decompression algorithms in hardware to the systems. This could be done DIRECTLY within PCIexpress RAID disk controllers that connect the servers to the harddrives. They would be doing real-time compression and decompression every-time a file is read or written, and at much lower power and much higher efficiency than the primary system CPU.
If you coupled this idea with smarter compression algorithms like those mentioned in the article that can access pieces of files without decompressing the entire file, this type of system could potentially create incredible gains in data-center power efficiency while keeping the same data capacity and use levels. At the same time, it could save a lot of money and service more users than a traditional solution. win - win - win.

Anyone know of something like this on the market? I wish I had the time and money to look into this on a serious level.

... drink deep, or taste not the Pierian spring.

"Compression = more CPU time = more electricity. Duh"

Go back to your gaming console, you clueless little troll. Leave the computers for the big kids. I'm surprised you were even able to navigate to this page.

I happen to agree with the troll Vashti.

Compression also requires higher quality hard drives in order to maintain data integrity. A one bit error in an uncompressed data sokum is neither here nor there, but a one bit error in a compressed sokum could be a disaster.

Rubbish! I manage an 80,000 square foot datacenter. Our annual energy bill is around $6 million. That doesn't account for the 900 gallons of Diesel we burn every hour we are without utility power. I can tell you assuredly that lack of data compression is not the problem. Who are they kidding? Virtualization will do a lot more for lowering energy usage in the datacenter than data compression.

Datacenter capacity planning all boils down to "watts per square foot". Newer datacenters frequently have a power density of 200 watts per square foot. Older datacenters can handle from 75 to 125 watts per square foot. More important than being able to supply that much energy to the floor is being able to remove the resultant heat. Our facility has over 4,000 tons of cooling capacity which accounts for about half of our utility bill. Lighting 80,000 square feet of datacenter costs another $50,000 a year. SoCal Edison is very expensive, even for industrial customers.

Data compression will help by requiring fewer or smaller storage arrays. Virtualization helps even more by eleminating entire servers. Unless you're a Yahoo or other gigantic data warehouse with rack after rack after rack of EMCs or NetApps, CPU, memory, routing and switching generally consume more energy than storage.

To say that data compression alone can cut datacenter energy consumption by 95 percent is hogwash! To be very generous, I would suggest perhaps a 10 percent industry-wide savings.

I agree with Bram, compression would need to be an end-to-end affair in order to truly maximize energy savings.

I believe some of you people need to think bigger. I used to work for a tier 1 storage vendors and we had hundreds of clients with over 1 petabyte of active storage and a 33% growth rate. Imagine a refrigerator size cabinet full of spinning disk consuming 7 kilowatts an hour, 24x7. Now imagine an air-conditioned room with over 15 of those cabinets, adding another 5 cabinets every year. Now for the really scary part, we had numerous clients with over 20 petabytes! Yes, 300 cabinets and adding another 100 every year. You do the math.

These are the target for this compression technology. If they can retire just half of those cabinets, it’ll provide an enormous power saving/CO2 reduction.

& it makes a LOT of assumptions about the data being stored. Try getting 90% compression on a JPEG image database or a PDF document management system. All this assumes all the data is plain old text, aside from the other fairly fundamental problems pointed out.

Sounds like someone left a market speaking advertising executive alone with the copy for too long.

Broadband over Narrowband (BoN)compresses H.264 three hundred percent and still looks great. A DVD quality movie is 600 MB. The NY Times article "Charging by the Byte to Curb Internet Traffic" of 6/15/08 tells why this can save many households hundreds of dollars per year.

As big a flop as Hans Reiser is, his filesystem actually managed to improve performance while using encryption and compression, and it's native to the FS.

Throw that on a box and smoke it. ;)