STH Global Server Farm
- Thread starter Dk3
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
The more I think about this the less I like a community share system. Security for STH is bad because you don't control the infrastructure. Security for the hosters is bad because you cannot control what is going on.
I'd be OK loaning a VM for STH to use but more like Patrick and the mods to use only for STH purposes.
Here's another idea for what to do with the capacity --- host a Linux-Bench image pre-packaged and have the image available through mirrors.
I'd even say it'd be slick to have images with common components. Like if you wanted to do a web server benchmark having the load generating nodes in a VM image hosted on STH. That would be useful to me at least.
I'd be OK loaning a VM for STH to use but more like Patrick and the mods to use only for STH purposes.
Here's another idea for what to do with the capacity --- host a Linux-Bench image pre-packaged and have the image available through mirrors.
I'd even say it'd be slick to have images with common components. Like if you wanted to do a web server benchmark having the load generating nodes in a VM image hosted on STH. That would be useful to me at least.
Great idea. Especially the last downtime give a reminder that this is very important.
Security wise this is problematic.
If the database is backed up, there's a chance that someone with access to the backup can extract credentials, mail addresses and other potentially sensitive information.
It's an interesting idea, but it seems to me that there are too many things that can go wrong.
Severe hypervisor vulnerabilities (like XSA-108) is just one thing, who knows what else one could do? (ARP spoofing on local network to redirect traffic comes to mind).
Execution is of no issue if we are only talking about a distributed filesystem of some sort. Plus like Dk3 noted. It's wrong to assume we can make multi-gb sized backups on an hourly base though.
A quick search suggests something like the PAST distrubted filesystem, (DHT), with K replicas in either direction of the chain. Source
No idea if there are implementations that directly follow the DHT approach. (Don't mention Btsync
)
What about if we did this differently? Why doesn't ServeTheHome become ServeTheCloud? Like what if we got a full rack somewhere. Then colocated boxes there in a shared cabinet?
Alternatively what if Patrick did this for us and then we could rent out boxes? Instead of having our own cloud labs, maybe we could have access to a colocated physical infrastructure?
I know places like Rackspace and SoftLayer do bare metal, but this is like bare infrastructure?
Alternatively what if Patrick did this for us and then we could rent out boxes? Instead of having our own cloud labs, maybe we could have access to a colocated physical infrastructure?
I know places like Rackspace and SoftLayer do bare metal, but this is like bare infrastructure?
I like the colocation idea but with OVH and similar companies offering servers for as low as £4 a month and quite meaty dual socket systems for £40 a month it kinda shoots colocation in the arm unless you want a huge deployment.
Maybe I'm alone on that opinion though.
Maybe I'm alone on that opinion though.
I work at a small colo and could squeeze a small hyper-v vm into a hyper-v server somewhere...as far as doing a shared rack, if that's something that's desirable we could possibly try and do something?... would have to talk to the owner... but I love this place, always great info here especially for the C6100 craze and the WD white labels