The uprise of P2P, now also as a digital currency in Bitcoin and other crypto currencies, led me to an interesting idea. A while back I was thinking about a P2P distributed VM system that could act as a shared compute cloud between peers, however, until now I dismissed it as too complex to implement. But maybe it’s not!
Reading about how Bitcoin works, I realised there are very complex things already implemented and maybe this idea could become a reality some time in the future.
P2P seems to dominate in a lot of areas nowadays and also for the foreseeable future. It is not just an idea, it is a philosophy. Sharing files on the Internet for free may be controversial, but it has dominated so much that it is basically unstoppable. Democracy, although maybe not the perfect system seems to be overtaking the planet and the masses will probably (hopefully) prevail in the end. Although, I am getting off-topic now.
In p2p cloud computing everyone would contribute their computing power to provide a low cost or even free network of compute power to host websites/apps and any other computing needs.
Imagine everyone could take advantage of everyone’s spare compute power! Computers use a lot of energy and that way we could take advantage of the otherwise wasted CPU cycles. When your computer is idle it is not doing nothing. The CPU continuously cycles through its clock. Sure there are now ways to save energy by turning on power management, but I personally never take advantage of that and usually have it turned off and I believe many others do the same.
Power management features usually cause bugs and other problems and the best they can do is throttle down your CPU frequency to another set rate. A CPU runs at a constant cycle usually expressed in Hz, Mega Hz or now more likely Giga Hz even on your phone. That means if a cycle is not used it is basically wasted. With a P2P Network of distributing load these cycles could be put to good use.
In Unix/Linux you can give a process a high nice value if you want to prioritise other things over this process. The operating system has nice values from -20 to +19. The default is 0. So, if unspecified, a process gets a nice value of 0. If you want your process to be nicer and favour other processes you can give it a higher nice value. The P2P cloud could take advantage of this feature and run at nice value of +19. Then, everything else would run at a higher priority by default and the P2P Cloud wouldn’t interfere with the user’s processes or even on an existing cloud or maybe more likely on physical servers, because the nature of the cloud already takes advantage of spreading the work.
This P2P Cloud of computers could work similarly to Amazon or the Google Cloud, however being decentralised and free to low-cost for users of it. The p2p cloud OS could be based on Free and Open Source Software__ and use strong encryption so that VMs on the _p2p cloud are completely encrypted and controlled only be their “owners”. Owners of VMs would pay by either providing some or all of their spare CPU cycles, let it be their phones, laptops or desktop computers. The owners could share their CPU cycles and with that “buy” a VM. There would need to be measures put in place of the p2p cloud os that controls and enforces that owners do not abuse the service, but maybe there are ways to ensure that resources are shared fairly. Maybe it will need to be based on trust of the users, but maybe there is a way to actually force users to have a ratio like in BitTorrent.
Network speeds are also fast and usually one’s Internet connection is not 100% utilised. P2P Cloud could also utilise people’s Internet connection to the fullest. There would also need to be ways to figure out the load on the network and optimising its use. Maybe users could choose to throttle the P2P Cloud OS to only use a limited amount of resources, which would at the same time limit their VMs capacities.
Until recently, I thought that there are some serious problems with this idea. For example, if hosting web sites a VM or node would need an IP address and ultimately a domain name. DNS requires there to be more or less static addresses which can be reached reliably. Amazon and Google solve this problem with virtual IP addresses. Maybe it is possible to map virtual IP addresses to real ones and have the communication run on non-default ports.
As with everything, the devil is in the detail. But maybe this will get you thinking and maybe you can solve some of these problems.
If you think this is a good idea, please leave a comment as this could become a valuable resource.
I am a professional programmer and believe this is possible. However, it is a challenging problem and would need to be worked out and implemented on a fairly low level, I believe, to perform well enough.
I look forward to any comments. :-)
comments powered by Disqus