Real-time computing. Real-time news.
Blog: The Essence of 'On-Demand'
Keeping tabs on the trends making IT more agile and more efficient, from automation to virtualization.
The Essence of 'On-Demand' | Main Blog Index
August 07, 2008
Say Goodbye to GRIDtoday (In Name, at Least)
I’m sure regular readers have noticed a steady evolution in the content of GRIDtoday over the past couple of years. We began covering what is now considered traditional grid and distributed computing technologies, but then extended our reach to encompass emerging technologies like data caching engines and grid-based application platforms or fabrics. Around this time last year, the increasingly service-oriented IT world latched onto the concept of utility computing , which quickly gave way to “the cloud.” Needless to say, we’ve done our fair share of coverage on this front, as well. Heck, we’ve even ventured far enough from our high-performance roots to tackle subjects like process automation, x86 virtualization and resource mapping.
What do all of these things have in common? Like the business benefits promised by grid computing back in the day (you know, way back in 2004), these technologies and trends offer users flexibility, adaptability, maximum utilization and, perhaps most importantly, access to resources when and where they are needed.
But they are not grid computing. As complementary as virtualization and grid computing might be, no one would call them one and the same. While cloud computing and grid computing share a common architecture, they are not necessarily interchangeable. And unlike grid computing, which seems to have settled into its niche powering CPU-intensive, and often time-sensitive, tasks like film rendering and electronic trading -- a fine niche to be in, no doubt -- enterprises are leveraging this new batch of technologies company-wide. When it comes to cloud computing, there are those among us who predict it will become the dominant computing paradigm within the next decade.
It is with these considerations in mind that Tabor Communications has decided to rebrand and refocus GRIDtoday. We are very excited to announce that in the very near future, GRIDtoday will metamorphose into On-Demand Enterprise (www.on-demandenterprise.com), or ODE, a publication whose editorial scope is not limited by the architecture embedded into its title. As should be clear from its name (and from the discussions I’ve had with some folks), ODE still will cover enterprise use of grid computing, but the publication also will cover cloud computing, virtualization (from server to network to I/O), data management, SOA, automation, managed hosting, and any other cutting-edge technologies that let businesses optimize their datacenters and get the most from their IT expenditures. Essentially, the content will closely resemble that which currently populates GRIDtoday, but the publication is designed from the ground up to address this new ecosystem.
Amazon CTO Werner Vogels likes to talk about how Amazon’s suite of Web services lets companies free themselves from the “undifferentiated heavy lifting” of IT management and spend more time and money actually improving the business. Really, that’s what On-Demand Enterprise is all about, as well. It’s about the technologies that, in one way or another, give companies the peace of mind that whatever they need -- power, data, machines, storage, applications, you name it -- will be there when they need it, at the scale they need it, and with as little effort as possible on the user’s side. Some call it real-time, some call it utility, and others still call it on-demand, but I think everyone calls it the future of computing.
Along with the new name and expanded editorial scope, the ODE Web site will offer an improved user experience via continuous publishing, blogs, polls, RSS feeds, user comments, top headlines from around the Web, etc. Although the site will be at the center of the ODE ecosystem, we will continue to keep busy readers up to date through daily and weekly e-mail newsletters highlighting updated news, feature and blog content on the site.
Like I said, we’re all very excited for ODE’s launch, but I have to admit I’m a little sad that ODE will not be covering scientific computing on a regular basis. Some research-driven news still will find its way into the publication, but most will not. GRIDtoday made its name in this space, and I have met some of the brightest computer science minds on the planet covering this space, but nixing this coverage area was necessary in order to achieve a tenable editorial focus. Obviously, there still is very cool and very important work being done in this space, and the world is not without news sources on it. HPCwire, our sister publication, will be picking up where GRIDtoday left off (to some degree, at least), and the world’s scientific grid and e-science projects remain the central focus of International Science Grid this Week, which receives funding from numerous government agencies throughout the United States and Europe.
See, it’s Even about More Than Computing … Kind of
A prime, albeit extreme, example of the reasons behind our change can be found coming out of -- where else -- Amazon. The company announced last week a Web-based user interface for its Mechanical Turk service, part of the Amazon Web Services portfolio, designed to make using the service easier for non-developers. I’ll get to that part shortly, but first I want to talk about how a service like Mechanical Turk fits into the greater on-demand computing landscape.
Essentially a service granting users the ability to access an on-demand, scalable workforce from around the globe, Mechanical Turk, to me, is the perfect complement to EC2. As flexible as EC2 can be in the right hands, there are jobs that its VMs simply cannot handle, jobs that require human judgment. However, just because it involves humans instead of easily provisioned machines instances, that doesn’t mean Mechanical Turk is more laborious or requires a greater commitment. “We definitely designed it on the same principles [as the other Amazon Web Services], which are it’s highly scalable and it’s on-demand. There’s no upfront commitment that you’re going to put so much work into the system or require so many workers or anything like that,” explained Sharon Chiarella, vice president of Mechanical Turk.
However, unlike Amazon Web Services’ poster child, EC2, Mechanical Turk now gives lay users the ability to harness its unique type of power. Originally, says Chiarella, the service was very much geared toward developers, but Amazon found “there was a demand for this service in the general business community, where they didn’t necessarily have a huge development team that they could allocate to do this.” Now, general users can access a Web-based interface that allows them to easily enter a large amount of work, review it and get results, quickly, says Chiarella. More details on the new GUI and tools are available in the announcement.
For the advanced user, though, the developer tools still are available, and they do allow for some interesting uses. For example, Chiarella explained, the developer API allows you to integrate Mechanical Turk directly into your application. “Let’s suppose you have an automated tool where someone submits a new ad -- you’re an ad-serving engine -- and you want to make sure that ad doesn’t contain any explicit images, or alcohol, firearms or tobacco, because they’re all regulated,” she hypothesized. “You can very easily take that image link and submit it through the API and get it out to Mechanical Turk very quickly, get the results back, and that can directly integrate into your application. Really, it just gives business the choice of do they need direct integration into their application, or do they want the ease of using a Web UI.”
Given the number of complaints I’ve heard about the skill level required to get started with EC2, I asked Chiarella if a user-friendly GUI is something Amazon is considering for its other products. As I expected, she declined to comment for the other services, but did note that Amazon is big on listening to its customers. If this is true, and if customers keep on Amazon about easing the EC2 experience, that would be a big deal in the cloud computing realm -- where many consider a Web interface part and parcel of the paradigm.
Like the rest of its Web services suite, Mechanical Turk was designed to tackle specific problems within Amazon. In this case, it was to keep the product catalog as clean as possible, without duplicates, misinformation, bad pictures, etc. Amazon saw an opportunity to use the Internet to harness an outsourced workforce and, voila, Mechanical Turk came into existence. The big use cases actually haven’t changed much. According to Chiarella, data management tasks like cleaning and tagging catalog data still are popular, as are editorial tasks like transcribing audio clips, determining whether Web posting are positive or negative, and using judgment to decide whether content is within a site’s guidelines. Requesters also submit research tasks seeking information on a wide range of topics.
The question that remains for me, though, is whether EC2 might cannibalize Mechanical Turk as artificial intelligence continues to improve. I didn’t ask, but I assume that the human-reliant service can’t offer SLAs comparable to what software can offer. If, for whatever reason, there aren’t workers willing or available to perform the tasks, I guess they don’t get done. But if AI advances to the point where algorithms can match (or close to) the human brain for Mechanical Turk’s current tasks, performing them on EC2 will all but guarantee they get done.
Chiarella doesn’t rule out the possibility, but warns against questioning the human brain. She contends that there always will be tasks best suited for humans. It makes sense, actually. I mean, humans have to write the algorithms in the first place, right?
“I’m a technologist at heart, and I do believe that technology will continue to advance,” Chiarella told me. “That said, I think the processing power of the human brain is awesome, and we definitely have found, so far, that there are a number of really varied tasks that humans can do much more quickly and efficiently and accurately today than an algorithm can do.
In some ways, Mechanical Turk provides the last mile to your algorithm. You might have an algorithm that can clean out your catalog 95 percent, but that last 5 percent is really, really impossible to do with your algorithm, and Mechanical Turk workers can complete that for you.”
Elsewhere in the Issue
Although I haven't left myself room to comment on it, I don't want the news of our evolution and the Mechanical Turk analysis to outshine the rest of a jam-packed, truly awesome issue. Be sure to check out all three features, in which we chat with Lucasfilm's director of IT operations, delve deeper into the Yahoo-HP-Intel cloud test bed, and explain what Citrix's OVF-based Project Kensho means to the future of virtualization. As for the big announcements this week, here are some of the highlights: "IBM Spends $400M on New Cloud Computing Centers"; "Mozilla Deploys Vertica, Pentaho for Large-Scale Analytics"; "Mobile Operator Picks Xeround to Improve User Experience"; "OpenNebula Open Sources Virtualization/Cloud Management"; "DataSynapse Offers Dynamic Service Management for VMware"; "VMware Enhances Virtual Datacenter Automation"; and "Intel, Facebook Sign Infrastructure Solutions Agreement."
-----
Comments about GRIDtoday are welcomed and encouraged. Write to me, Derrick Harris, at editor@gridtoday.com.
Posted by Derrick Harris - August 7 @ 8:30PM
(Digg, Technorati, more)
