World Of Gadget

As hardware and bandwidth costs continue to drop, the biggest expense associated with large data centers has been their power draw; it’s estimated that 1.5 percent of US energy production now goes to feed them. Although there have been a variety of solutions for increasing the energy efficiency of these facilities, a new paper describes another approach to cutting power costs: shifting loads to wherever the power is cheaper.

The study was performed by academics from MIT and Carnegie Mellon, who teamed up with Akamai, a company that runs a content distribution network that caches local copies of frequently requested files near major network hubs. By obtaining data from the requests that pour into Akamai’s facilities and comparing it to spot energy prices from around the country, the authors were able to determine that it’s possible to save the money paid for powering the data centers, though not the electricity itself.

The basic idea is that many companies, especially those heavily invested in IT, have multiple data centers scattered around the country. Thanks to various distributed file systems, each of these will typically have access to all the information needed to respond to a given user’s request. As such, it should be possible to redirect a request to wherever power is the cheapest at a cost in transit time that may be measured in milliseconds if network congestion isn’t bad. Servers in locations where energy costs are high can be allowed to idle, at which point they consume 60 percent or less than they do while under load.

The authors obtained market prices for electricity from a number of locations in the US, which allowed them to confirm that prices varied over time, and that the variations weren’t correlated between locations, meaning that the system was dynamic enough that shifting loads made economic sense. They then stepped through a series of actual load information from Akamai, and modeled datacenter electricity usage with a model that was produced by Google. With a baseline set, they then explored whether money could be saved under various conditions.

Distance played a large role in determining the economics. So, for example, the savings showed big jumps at two specific distance thresholds: when Boston-area traffic could be rerouted to the DC area, and again when it could be sent as far as Chicago. The ability to take advantage of real-time energy prices also played a big role. Prices rose quickly as the time lag to respond stretched from one to three hours, and then continued to rise more gradually out to 12 hours.

But, at the moment, bandwidth costs provide the biggest barrier to optimization. Bandwidth still costs a bit more than power, although its price has been dropping while electricity rates have remained stable or risen, so the relative importance of bandwidth costs is expected to drop. Optimizing the distribution of load while keeping Akamai’s existing bandwidth policy in place cuts the potential savings in electricity costs down to about one-third of what they would be otherwise. Optimizing for both bandwidth and power costs simultaneously, however, had a large impact.

All told, the answer you get depends dramatically upon the assumptions involved. In a worst-case scenario, where the process is optimized for bandwidth costs and the distances loads are shifted are kept short, the savings may be as small as five percent, which is pretty minor unless you happen to be Google or Microsoft. Under the ideal conditions, the cost savings could clear 50 percent—in comparison, moving an entire facility to some place with lower-priced power would only lead to savings of 65 percent.

The authors point out that most companies with profiles like Akamai already have load-balancing software in place that responds to things like congestion, bandwidth prices, and reliability issues; adding energy prices doesn’t represent a big leap. They also describe a whole host of ways that the dynamics could be changed. For example, at colocalation facilities, racks are generally charged based on the amount of power that they could consume, rather than the actual power provided. As demand-response pricing and smart grids become more common, the savings for shifting a load large enough to help a utility avoid a brownout could be substantial. It would also be possible to send loads to where the air is cool, saving money by actually using less power for running air conditioners.

And, ultimately, this may be the sort of technique that enables intermittent renewable sources to be brought into the grid with less disruption. Shifting the load to where the sun’s shining or the wind’s blowing will probably be far easier than producing the infrastructure needed to send power from renewable generating facilities to where the data centers are.

Although the US and other nations currently produce ethanol from the sugars and starches of crops like sugar cane and corn, ethanol isn’t a good match for our existing fuel infrastructure—and this form of production runs the risk of putting energy in competition with food production for resources like land and water. As a result, attention has shifted to figuring out how to produce a new generation of biofuels from different sources that more closely approximate the diesel and gasoline in use today.

Yesterday’s edition of Science contained a perspective on the prospects for these next-generation biofuels. Although it’s very short, only a page long, it contains an excellent list of references to very current publications. The diversity of approaches it covers highlights how many options there are to produce different fuels, each with its own advantages and drawbacks. In the following sections, we’ll discuss different methods of producing biofuels; although the text presents them as alternatives, it’s important to emphasize that there are significant overlaps among them, and more than one technology might emerge for widespread use.

Cellulose or lipids: The first question is what you want the initial fuel stock to be. Cellulose, a polymerized sugar, is found in combination with proteins and other compounds; together, they provide plants with their structure. It’s available in abundance, and all sources of cellulose are more or less functionally equivalent, so a cellulose processing facility could work with a huge range of sources: agricultural waste, wood waste from lumber processing, leftover construction material, and dedicated biofuel crops like switchgrass. A recent USDA report suggests that the US has an annual supply of a billion tons of cellulose to work with.

The problem is that cellulose doesn’t break down easily or quickly, and its component parts—sugars and proteins—lend themselves most readily to the production of ethanol. The alternative is to go with something like lipids, which are long-chained hydrocarbons that are chemically more similar to fuels. 

As with cellulose, a lipid is a lipid, regardless of the source. Unfortunately, there are far fewer sources, since plants don’t generally contain an excess of them. In a few cases, lipid-processing facilities have been built to operate using the waste from meat packing plants, but a larger source of lipids would require the development of an algae that grows well in waste or salt water.

Industrial or biological: Once the material itself is isolated, it generally needs processing before it’s ready for use as fuel. Here, working with lipids has huge advantages, as they can generally be fed into the same sort of reactions that are used during oil refining. 

Cellulose is another matter. Several industrial processes can quickly break it down into more chemically-flexible components, but all of them require elevated temperatures and pressures, and thus a substantial energy input.

Once broken down, catalysts can rearrange the simpler components into usable fuels. But there are significant barriers to using the processes which we’ve already developed, designed to handle relatively pure mixtures. It’s difficult to fully eliminate water from cellulose, the plant material contains an unwanted mix of nitrogen and sulfur compounds, and the pH of the reaction material can be difficult to control. Any one of these issues can interfere with the action of an essential catalyst, so finding a combination that provides high energy payback and works well with cellulose isn’t a simple thing.

The alternative is to go biological, as there are organisms that make their living by digesting cellulose. Unfortunately, as anyone knows who has ever seen downed trees rot, the process is extremely slow. There are ways to speed it up, of course, but there’s a very clear trade-off between speed and energy input. Once it’s digested, researchers still face the challenge of finding a way to convert its component sugars to something more useful than ethanol.

Existing or engineered organisms: The ideal situation, of course, would be to have an organism that does everything for us, from digesting ethanol to spitting out some sort of lipid derivative that can be fed directly into fuel tanks. Life, however, is rarely that simple; evolution tends to frown upon organisms that release energetically valuable chemicals into their environments, and we’d be growing any biofuel organisms in the sorts of numbers that allow evolution plenty of opportunity.

That said, there are some organisms that happen to do some very convenient things on their own, like a bacterial strain that stores energy in lipids to such a degree that nearly 80 percent of its dry mass is in lipids. Unfortunately, we know almost nothing else about the strain, including the details of metabolism and whether they can be engineered in the same way we handle more common bacteria, like E. coli. And some form of engineering would almost certainly be needed if we’re going to have them growing on cellulose.

The alternative approach would be to take an organism we do know well, like E. coli, and engineer it so that it could digest cellulose and directly produce a chemical that could be easily converted to fuel. (A review of microbial fuels lists a variety of intermediate-chain hydrocarbons and alcohols that might work.) A sufficiently hydrophobic liquid, once outside the cell, should separate from the liquid the bacteria are growing in, allowing easy harvesting. But, at the moment, we only have limited experience with re-engineering an organism’s metabolism to do something that might not do said organism much good.

Because of all the complex trade-offs, it’s likely that there won’t be a one-size-fits-all biofuel solution, but the energy available to us (based on the USDA report) is substantial and could dramatically cut our imports of fossil fuels—up to 30 percent, based on a report that predates the recently tightened fuel economy standards. 

Fortunately, it appears that the market is taking the advice of the National Academies and developing a portfolio of approaches. The Science perspective notes that a handful of companies already claim they’re on track to begin producing non-ethanol biofuels on the scale of millions of gallons within the next five years, and more will join them within the next decade. Although none of them are giving away the precise details of what they’re up to, it’s clear that they are taking a variety of different approaches to producing that fuel. Gas has never looked greener.

Video game sales date for July has come in from the NPD Group, and the industry continues to sink. “The US video games industry declined for the fifth consecutive month, bringing year-to-date sales to $8.16 billion, down 14 percent from the same time period last year,” wrote NPD analyst Anita Frazier. “In order for the industry to come in flat or slightly up for the total year, the back five months of the year have to come in 11 percent (or more) higher than the last five months of last year.”

She points out that we do have surefire hits like Halo: ODST and The Beatles: Rock Band coming later this year, but for now, the industry is treading water. Let’s see who’s doing it most efficiently.

Nintendo

Sales of Nintendo hardware are slowing down, but with both the Wii and the Nintendo DS enjoying such a huge lead, the company has plenty of time until it feels Microsoft or Sony hardware breathing down its neck. The Wii sold 252,500 units in July and the Nintendo DS sold 538,900 units. In July 2008, the Wii sold almost 300,000 more units than it did this July; has the momentum finally slowed?

Not when it comes to software… even if that software often comes bundled with hardware. Wii Sports Resort came in first place with 508,200 units sold, along with the MotionPlus. Wii Fit with the Balance Board came in fourth place with 164,300 units sold, and then Mario Kart with the wheel on the Wii, Mario Kart on the DS, and Pokemon Platinum version took the fifth, sixth, and seventh places. New Super Mario Bros. came in at ninth place with 101,800 units sold, and the EA Sports Active Bundle took the tenth place with 96,800 sold.

Nintendo hardware may see slipping sales year over year, but that just means the company is coming in at number one by a slimmer margin. Taking seven of the top ten slots with mostly catalog titles and first-party games also shows that Nintendo continues to enjoy massive success.

Microsoft

Microsoft just about matched 2008’s sales, with 202,900 Xbox 360 units sold in July. The second-best-selling game was also on the 360 as NCAA Football 10 sold 376,500 units. (That’s 139,100 more units than the game sold on the PS3.) The only other 360 game in the top ten was Fight Night Round 4, which took the number eight slot with 116,400 units sold.

“The Xbox 360 is the only console system showing a unit sales increase year-to-date,” Frazier pointed out, and in fact the 360 came within 50,000 units of catching up to the Wii.

Sony

Sony has very little to brag about in the NPD report. The PS3 sold 121,800 units, the PSP sold 122,800 units, and the PS2 sold 108,000 units. Each piece of hardware is well behind its competitors. The only game that Sony charted in the top ten was the PS3 version of NCAA Football 10, which took the number three slot with 237,400 sold.

That’s relatively good news, if you don’t notice how easily the 360 trounced those sales with its own port of the game.

Not much good news for July

Sales are down, the software chart is dominated by Wii and DS titles that have been out for years or that continue existing franchises, and Sony needs to do something to energize gamers. Perhaps… and we’re just throwing out ideas here… a redesigned PS3 along with a price drop?

It’s going to be quite a challenge to get these numbers out of the gutter, but the end of 2009 still has some heavy-hitters coming, including Call of Duty: Modern Warfare 2.

Microsoft has announced some big news concerning the next major revision of Office for Mac: it is currently on track to ship late next year, and it will replace the much-maligned Entourage with a bona fide version of Outlook for Mac. 

In the meantime, Microsoft’s Mac Business Unit has a final version of Entourage Web Services Edition ready for those who need better Exchange support now, and they will be packaging it in a new “Business Edition” of Office for Mac 2008 starting next month.

Seriously, it’s Outlook. For real.

You read the news right: Microsoft is finally making a real, actual, official, true, and correct version of Outlook for Mac OS X. This new application, which Mac BU general manager Eric Wilfrid told us he has been using for nearly a year, is built from the ground up to fully support Exchange’s latest protocols. Written in Cocoa to take advantage of the latest OS X technologies, Outlook for Mac will also have a completely redesigned database that will be compatible with Spotlight and Time Machine.

“Because it’s designed for the Mac, it will be different from Outlook for Windows,” Wilfrid said. “But it will be fully compatible. For instance, Outlook for Mac will support Information Rights Management, which controls access to sensitive documents.”

As a sort of stepping stone towards a full Outlook client for Mac OS X, the Mac BU has released Entourage Web Services Edition. This version of Entourage is architected to connect to Exchange 2007, and it provides much faster data synchronization and improved e-mail and calendar support. Users of Office 2008 with SP2 applied can download Entourage Web Services Edition today, but bear in mind it requires Exchange 2007 SP1 RU4 or later. It’s definitely worth checking with your IT department to see if your company’s Exchange server is compatible.

Finally, the Mac BU is rejiggering the versions of Office 2008 it is making available to consumers. The “Student and Home Edition” will stay the same, but a new “Business Edition” arrives on September 15. Packaged in a Serious Business Black box, this new edition will include Entourage Web Services Edition and Document Connection for Sharepoint compatibility, and it will sell for the same $399 as the “Standard Edition” did. Also included with the Business Edition are over 200 professionally designed, business-oriented templates and clip art, as well as seven hours worth of training at lynda.com.

According to Wilfrid, a recent survey showed that 90 percent of Office for Mac 2008 users would recommend the package to colleagues and friends—despite the fact that the loss of Visual Basic and a lack of complete Exchange support relegated Office for Mac 2008 to also-ran status in many business environments. 

But with Visual Basic set to return next year and the news of a full Outlook client coming to the suite, the Mac BU hopes to suggest that Office for Mac is serious business.

Nokia first unveiled its vision for a next-generation Internet Tablet device when it announced plans for Maemo 5—a new version of the Linux-based tablet operating system—at the Open Source in Mobile event last year in Berlin. Early plans indicated that the OS would be paired with a device that would include 3G connectivity and an OMAP3 processor. Ever since that announcement, mobile Linux enthusiasts have eagerly anticipated the arrival of an actual Maemo 5 device.

A Nokia product codenamed RX-51 recently cleared FCC approval and has now appeared in leaked photos that made their way onto the Internet; it appears to be the long-awaited Maemo 5 product. The source code of Maemo 5 provided some early clues about the RX-51 last year. Maemo fans have been poring through it and documenting various technical details that provide insight into the hardware.

The new pictures also seem to indicate that the device is a phone, and not just a Web tablet like its predecessors. This would be a major step forward for Maemo and could reflect a more Linux-centric mobile strategy for Nokia. The handset maker announced a project earlier this year called ofono, which seeks to build a complete telephony stack for mobile Linux platforms. The emergence of ofono and Nokia’s move to make Maemo a viable smartphone platform has raised a lot of questions about the company’s commitment to its other mobile OS, Symbian.

Nokia acquired Symbian last year; its phones ship with its own S60 variant of the OS, which is increasingly anachronistic compared to modern smartphone platforms. Nokia’s plan of opening the Symbian platform appeared to be part of a move to modernize and strengthen it. Several reports have emerged this year, including one in the past few days, saying that sources close to Nokia are suggesting that the handset giant sees Symbian as a dead end and is preparing to move entirely to Linux.

There are still a lot of unanswered questions about Nokia’s long-term platform strategy, but it’s still far too early to say whether Nokia is switching entirely to Linux. As we have recently described, the Symbian Foundation has launched a number of far-reaching initiatives that aim to address the platform’s weaknesses. The resulting user interface and development improvements could make it much more competitive. The announcement that Symbian is getting a mobile version of Microsoft Office also suggests a long-term commitment.

It’s also important to keep in mind that Nokia has a harmonization strategy: the Qt toolkit, which it obtained through an acquisition of Trolltech last year. Nokia is bringing Qt to S60 and plans to make Qt the foundation of the Maemo user experience in Harmattan, the version that will follow Maemo 5. This means that developers will be able to use Qt to build software that can run on both mobile platforms. This addresses the potential fragmentation issues and makes it practical for Nokia to continue using both platforms, targeting each to contexts in which it’s appropriate.

Maemo products have traditionally been developer-centric and somewhat experimental, but it’s likely that we will see the platform move into the regular consumer market as it matures and Nokia boosts its commitment. If it ends up being a viable and compelling replacement for Symbian, we could potentially see a transition over time. But Symbian is the world’s dominant smartphone platform today, and its considerable momentum means that there is still a strong business case for modernizing it and moving it forward alongside Maemo.

Nokia has recently partnered with Intel to build a somewhat mysterious new class of mobile devices. In addition to delivering Maemo on Internet Tablets, it’s likely that we will see Nokia pushing Maemo in whatever those products turn out to be, while Symbian remains the standard platform for the company’s more conventional smartphone handset lineup.