Google had a bad week in cloud computing, with serious downtime in Gmail, Blogger and Spreadsheet. Back in July it was Amazon that was embarrassed with their S3 outage. If you measure on total downtime, cloud computing still looks good compared to traditional hosting or in-house data centers. But that glosses over the psychological and market confidence issues, when a problem hits everybody at the same time. In contrast, when was the last time you heard about a massive Skype outage? Maybe it is time to look more seriously at P2P?
Well, actually about one year ago Skype did have a problem. But it was minor in comparison in terms of impact compared to the Google and Amazon outages. Skype claims over 9m people online right now, so this is major validation for P2P scalability and reliability.
This week we also saw the launch of Wuala, a P2P Cloud Storage solution (our review here).
Earlier, we reviewed a P2P approach to search (Faroo) and a P2P approach to video sharing (Metaaso).
With the exception of Skype, these are all tiny little start-ups. Interestingly, they have all originated outside America:
Skype - telephony - Estonia
Wuala - storage - Switzerland
Faroo - search - Germany
Metaaso - video - India.
This geographic origin may not be coincidental.
You need $ gazillions to be a Cloud Computing Platform. Those server farms cost a lot. Skimping, or misjudging demand, leads to outages, slow response and other confidence-killers. This is a game for the big boys - Google, Microsoft, Yahoo, Amazon, AT&T, Sun. These are all American firms, with access to plenty of capital. Disruptive innovation usually comes from start-ups that are starved for capital. You replace capital with technical innovation. That was true for Google, Microsoft, Yahoo, Amazon, AT&T, Sun as well when they started.
That is why I have believed for some time that P2P is the next big disruptive technology at the infrastructure level.
Disruptive technology sometimes needs support from big companies as well. Fortunately for P2P, 3 very big companies would benefit greatly from more use of P2P as infrastructure - Microsoft, Apple and Intel. It's a great way to mop up those underutilized desktop CPU cycles. And attack the cloud computing incumbents.
Historically, P2P start-ups have tended to focus on music sharing and have been hurt by legal issues, but they have been fine technically. Skype makes P2P respectable and proves that scalability does not have to be an issue. Skype is taking on one the biggest and most entrenched industries in the world and millions of people increasingly rely on Skype as a mission critical alternative to landlines or cellphones.
P2P infrastructure could play very well behind the enterprise firewall. It reduces CIO security fears about too many cloud based apps outside the firewall. This is important for P2P start-ups. They would need a lot of capital to go to market entirely with a consumer/SOHO offering. If they can get enterprise adoption at the same time, then they can accelerate cash flow and reduce need for funding.
Watch the P2P space. It's the next big wave of innovation at the infrastructure level.
Comments
Subscribe to comments for this post OR Subscribe to comments for all Read/WriteWeb posts
While cloud computing service has been demanded to fullfil certain service level, what about P2P. Technologically, even a small startup can come up with a cloud computing based service. But to meet a certain service level, this is where money will talk. This is where the big boys have the advantages.
Is there any P2P service meeting some sort or service level agreement? Regular P2P sharing doesn't have such thing, does it?
Posted by: Akhmad Fathonih | August 14, 2008 10:36 PM
I'm not 100% sure you used the term cloud enough in this post. Also I'm not sure it's even worth mentioning considering virtually anything on the internet satisfies the various definitions of cloud computing.
P2P itself is an example of cloud computing. You're here, stuff's happening over there, and you don't have to do anything more than work within a user interface.
It's also worth remembering that P2P and perhaps most notably Skype are also not infallible. Remember back in August last year when a Windows update lead to Skype being knocked offline? The growing trend by ISPs to block or shape P2P traffic puts a pretty significant question mark over it's viability, even though there's legitimate uses and countless potential with the technology.
So anyway P2P is really just part of the cloud and failures are just part of life. The very rare incidences of failure that the major players suffer from hardly outweigh the benefit of being able to say cloud every few sentences and 'outsource' scalability.
Posted by: Ben | August 14, 2008 11:08 PM
Akhmad, Service Level Agreements (SLA( are not a magic answer. What do you actually get when a vendor fails to meet SLA standards? Some discount on future fees - maybe - but certainly not any consequential loss (lost time and business).
Ben, sure if it ain't all on our PC or a server we directly control it is all "cloud". Maybe I should have said "server farm failures are serious". The real issue is the number of people that are simultaneously affected by failure. P2P seems to score better on that score.
Of course P2P is not infallible. The post mentions and links to the Skype problem one year ago. But it has the potential a) to be more scalable b) to be dramatically lower cost c) to have fewer points of catastrophic failure.
Posted by: Bernard Lunn
|
August 15, 2008 1:38 AM
Down times of Amazon and Google's cloud computing platforms have been an issue lately, but I believe the benefits far outweigh the disadvantages. The best thing cloud computing offers is "power to the people", it lowers the entry barrier for start-ups leaving the innovators to do what they do best... innovate, instead of worrying about infrastructure issues. Imagine how better off Twitter would have been if they didn't have to worry about down time and instead invested their energies in bettering the twitter user experience.
I've covered the benefits of cloud computing to application developers in my blog
http://sachendra.wordpress.com/2008/06/23/cloud-computing-offers-power-to-the-people/
Posted by: Sachendra
|
August 15, 2008 4:00 AM
The problem with using P2P to bypass the "$gazillions [a startup needs] to be a Cloud Computing Platform" is that it is just passing those costs onto another infrastructure player, the network provider (telco, ISP, etc), none of whom are going to eat those costs forever. You can criticize carriers for talking about implementing usage tiers or throttling BitTorrent but they are doing just what you say Cloud Providers must do -- managing costs.
Seems the solution is hybrid solutions that work offline but provide synch/backup -- in the spirit of POP/iMAP email solutions.
Hmm...hybrid desktop/cloud....that's Microsoft's model for the future, no?
Posted by: Rich | August 15, 2008 7:03 AM
This is an interesting article and timely too. However, re-evaluating cloud computing and thinking of p2p as an alternative is only being discussed becuase of Google's recent outages. Had Gmail and other services not gone down, I doubt this post would have been written.
Downtime is part of the game. Sure, it sucks when your service goes down for everyone, but as you stated "If you measure on total downtime, cloud computing still looks good compared to traditional hosting."
Cloud computing offers so many advantages, best in class hosting, backup, data centers, and on...
For some, p2p may work well. It works well for Skype and bit torrent, but what happens when a major ISP blocks or throttles that program's ability to use p2p?
I get it, people are upset becuase they couldn't access a service. I couldn't access Gmail at work either, but what are the alternatives? I still think cloud computing is better than p2p or in-house hosting.
Greg
beYOU.tv
Posted by: Gregory Schnese | August 15, 2008 7:26 AM
While the companies you mention may be getting a bit of press at the moment, they are not the first to test the waters in these particular market segments and none of them are doing anything on the technical side that has not already been tried. There were a slew of companies in the dot-com era who tried and failed at these exact same segments. Skype is the only exception and it appears that they were just as unsound as the others but happened to have more media hype to compensate for the lack of a sustainable business plan. Perhaps the reason that these new deadpool candidates are coming from outside the US is that US investors and entrepreneurs actually learned a few lessons back then and know what doesn't actually work...
P2P is hard, and the economic benefits are mostly eaten up in increased complexity and fault-tolerance costs. Because it is harder to actually coordinate and provision these ephemeral resources they are actually far more fragile and subject to disruption than centralized clouds. If Amazon's AWS suffers and outage they can examine the causes and deploy more resources to either make sure it does not happen again or deal with losing some chunk of the resource base and still providing the service; once any of these p2p systems suffers a minor failure its users will bail and lead to a cascading diminishment of the resource base to the point where the service will completely disappear.
Posted by: evgen | August 15, 2008 9:02 AM
I agree with Bernard on this one.
Peer to Peer collaboration works pretty well in most organizations in real life. I think the whole "give this packet to seven random strangers" model of p2p solutions that we have seen in the past isn't as interesting (or as important) as content discovery and contribution directly with people you know. P2P and centralized storage complement each other - HTTP is a peer to peer protocol, and it invented the cloud.
Disclosure: P2P is an important design elememt of our enterprise software product - and yeah, we're a tiny little startup (but one from the US)
Posted by: Gordon Taylor | August 15, 2008 11:18 AM
It's not that cloud computing (or distributed computing) is flawed, but it actually needs to continue to become more distributed. Instead of relying on one service provider, a website or data store should be distributed across many services. Perhaps this is closer to a P2P model, in that it would have no single point of failure.
Posted by: mike | August 15, 2008 10:02 PM
Great article, I quoted this and did abit of an extra writeup on http://struct3.com/has-the-cloud-failed-before-prime-time/
Based on some discussion going on at Y Combinator News- I am rethinking the whole P2P thing as a viable solution now. Its like one said, when your server goes down- you know where to look, when Amazon is down, you know who to call, when something is wrong in P2P. You are shit out of luck...
Marco
Posted by: Marco Kotrotsos | August 16, 2008 2:11 AM
No one's paying attention to p2p - time to make up some bullshit about cloud failures?
Posted by: website designs | August 19, 2008 12:26 PM