RE: GRC Supplementary Developments Proposal
I do not support running Virtual machines on volunteer computers. They give their spare computational power in hope that it will be useful, and wasted on running a hyper-visor.
I do support changing standard for WUs (applications). The current boinc standard/api is just too clumsy even to the point that some projects opt to create wrapper apps.
Make the crunching capability of Gridsters more accessibe for one-off or commercial computation? Definitely. It takes far too long to set up boinc projects, fight the app cofig, plans, etc and get the project whitelisted.
It needs to be fast and straightforward. Candidates should be presented with, first a guide on how to make their app compatible with boinc/gridcoin and then a place to submit their app and WUs.
There will be some base, low reward for crunching those WUs, community can decide on more reward by whitelist voting or the candidate them-self can pay up if they either want more power or just reward+support grc.
Hi tomasbrod,
I agree VBox as Hypervisor is very wasteful when it comes to resources, so I am currently experimenting with VMWare Vapp implementation which is an "app level" hypervisor rather then os level.
I'll need to do a lot of tests, but so far this technology looks promising. Especially when it comes to inter-compatibility with other platforms as most of them support this. At the moment, it's industry-wide accepted standard.
Without allowing ease of migration for customers, we are unlikely to attract businesses. In my opinion, two key factors are in play:
It's unlikely i will be close to initial testing version of Vapp in less then 2 weeks based on documentation examined so far. The more I read, more problems starts to pop-out, but i guess that's the rock'n'roll we like to live with.
I was searching on the topic of how to compute in cloud, but I could not find relevant guide. What I have to do for my "thing" to become "vApp"?
In order to serve customers, their needs must be identified.
What they have?
What they want?
What they want to do themselves?, eg: installing the app, logging with ssh, filling database with data.
What they want from the provider (us)?, eg: running a bunch of VM images, providing a VPS, guaranteed up-time and reliability, serving downloads or running this program million of times (with different inputs) and collecting the results.
@tomasbrod, i am currently examining and making a list of what Google App Engine / Azure and Bluemix offers. Out of that list, i am going to separate, what's relatively easy to do, what's problematic, and what's impossible. I think it's a good starting point.
As for the Vapps, the standards differs from service to service, but you can look into it as "app level hypervisor" rather then os level. For example, you can deploy mysql on the OS, or as an Vapp. In second case, only the core requirements of an app are virtualized. To make it to the simplest level for the purpose of explanation, such virtualized system would not have anything other then what's required to run mysql. No ssh, not even bash. It's a way to save overhead that OS level virtualization comes with, as well allowing ease of migration. However that's just brief descriptions.