In the 1960s, Lawrence Roberts invented computer networking via data packets, which led directly to the development of ARPANet and the Internet . And now Roberts is trying to fix one of the Internet's biggest problems: network overload caused by peer-to-peer file transfers.
Not Al Gore.
(Credit: Rafe Needleman / CNET)At Structure 08, he laid out the problem: 5 percent of the Net's users are running P2P transfers taking up 80 percent of its capacity, which is dramatically limiting the available bandwidth available to everyone else. Roberts' company, Anagran, is able to detect which "flows" are P2P traffic, and reduce the bandwidth available to these communications when other users' systems want it. Roberts says that Anagran's technology even functions when P2P transfers are encrypted. I'm not going to pretend I understand exactly how this works, but it has something to do with keeping information about the flow of data between all computers connected through an ISP in memory in the Anagram appliance, and then leveling off traffic of P2P communications as needed--and throughput only, not latency. Judging by the reaction of the audience of infrastructure geeks sitting around me, Roberts is on to something. "He's the real deal right there," the guy next to me said at one point, pretty much gazing up at the stage in awe.
Roberts claims that the Anagran devices also ensure that high-priority traffic, like VOIP and video streams, can be guaranteed better performance.
Roberts was clear that he has no desire to punish P2P users, but rather he wants to make sure that they--and everyone else--get their fair share of bandwidth. That share, he believes, cannot be 80 percent of the Net's capacity, especially if the other 20 percent has to be allocated to the 95 percent of the Net's users who aren't using P2P.
You'll find Anagran bandwidth fairness boxes (also called FR-1000s) in university settings now, where the P2P file transfer problem is most acute. Anagran doesn't currently have any commercial ISP customers, but I'll bet that they're all looking at them.
Roberts has no position on the legality of content being transferred over P2P links. "Illegal or legal is not the issue at all." It's about fairness, he says: equal capacity for equal pay. What do you think?
See also: Baggage and bits: Overage fees have unintended consequences.
Anagran FR-1000: The shape of bandwidth to come.
(Credit: Anagran)
Parker Harris (center) of Salesforce.com, gets grilled by Om Malik (left) and Michael Copeland.
(Credit: Rafe Needleman/CNET Networks)SAN FRANCISCO--In consecutive talks here at Structure 08, we just heard from big brains at Salesforce.com and Microsoft on the topic of software vs. services. Surprisingly, these companies, which historically have espoused very different philosophies, are converging on a similar pitch: you can't build one with the other.
In Salesforce's case, co-founder and EVP Parker Harris said that early on, Salesforce's architects wanted to build a platform service for consumers, not just an application. But business logic prevailed: customers needed an application, not a platform.
"As a technologist, you want to build a platform, but you risk losing touch with what you're building it for," Harris said. "So when we started, we said we're going to build a service that's fast, simple, and right the first time."
And once Salesforce did expose its platform to developers, it found it couldn't think about its main app separately from this underlying architecture. "Software and infrastructure are not separate things. They're one thing."
Microsoft's Debra Chrapaty illustrates the point that Microsoft is a software-as-a-service company, too.
(Credit: Rafe Needleman/CNET Networks)For its part, Microsoft is no longer thinking only about applications. It continues to build apps with varying degrees of local and Web-based functionality, and because of that, Debra Chrapaty, vice president of global foundation services, keeps a close eye on the efficiency of apps. Because when scaled up into a data center, computational-efficiency matters.
The company needs to "make every kilowatt count," she said. Chrapaty also said that at least one of Microsoft's Virtual Earth servers, in Colorado, runs on wind power.
I find it curious that the two companies, Microsoft historically a pure software play, and Salesforce, a poster child for software as a service, are needing to solve the same problems today: building online applications and platforms that are reliable and open to their customers. And I especially liked that Microsoft is taking responsibility for the energy its apps use. I'd love to see that mindset spread through the industry.
Click here to see more stories from the Structure 08 conference and on cloud computing generally.
SAN FRANCISCO--Here at the Structure conference, everything is cloud, cloud, cloud. No one wants to own their own Web hardware anymore, it seems, and the company representatives speaking here are happy to provide the software and virtual services to replace the hardware.
One of those is GoGrid, which is shooting for the same cloud-computing market that Amazon.com is making a run at with its EC2, or Elastic Compute Cloud, service and related Web services.
The GoGrid pitch: We're cheaper. And easier.
GoGrid CEO John Keagy told me that, at volume, his services undercut Amazon's. He charges 8 cents a gigabyte-hour for compute services, compared to EC2's 10 cents. Also, data storage is associated with compute servers, and if a server goes offline, when it comes back, the storage will still be there.
At Structure on Wednesday, Amazon CTO Werner Vogels pitched "persistent storage" as a new offering from Amazon.
Keagy also said GoGrid has a graphical user interface-based control panel for its customers, allowing them to quickly set up their compute environment in a simpler manner than Amazon's service allows.
I can't do a hands-on with these two cloud services, but there are a few other points that I found interesting. First, GoGrid offers virtual Windows services, as well as Linux, and about 50 percent of its installations are for Windows processes. Some popular Web 2.0 services, like CommunityServer, are still Windows-only.
Also, GoGrid has never had a system-wide outage, as Amazon has. Keagy is realistic, though: "We're in beta. It will happen to us too." But, he says, with well-designed systems, recovery can be swift.
One thing GoGrid certainly doesn't have is Amazon's scale. Although the company is a division of the well-established ServePath, its single 20,000 square-foot facility can't hold a generator to Amazon's massive distributed infrastructure. Keagy did say he is building out distribution for GoGrid, using more of ServePath's locations.
Like the new Mosso cloud-based storage service, GoGrid is accessible through REST (representational state transfer) application programming interfaces.
Click here to see more stories from the Structure 08 conference and on cloud computing generally.- prev
- 1
- next
