Posted on 04/27/2015 6:20:06 AM PDT by ShadowAce
CoreOS and Joyent's SmartOS/Triton have worked to redefine, in radically different ways, what an OS needs to be to run applications at scale in the cloud.
Now another candidate is set to join the ranks of maverick cloud OSes: OSv, open source, hypervisor-optimized, and "designed to run an application stack without getting in the way.">
OSv runs existing Linux applications, but is not itself Linux; it's entirely new and written from the ground up in C/C++. It can run on a slew of hypervisors and virtual machine systems, or in cloud environments like Amazon EC2 or Google Compute Engine. The company claims significant performance gains for apps running on OSv, saying the boost comes from design choices that rely on the hypervisor, rather than separate user and kernel address spaces, for keeping elements isolated.
Instead of using Docker containers, OSv uses its own application-image system called Capstan. Apps packaged with Capstan are said to be "only 12-20MB larger than [the] application itself" and are complete virtual images that can run on any hypervisor that supports OSv.
The people heading up Cloudius, the company sponsoring the project, have been heavily involved in the business of virtualization for some time now. Cloudius CTO Avi Kivity and CEO Dor Laor both helped create the original incarnation of the KVM hypervisor at Qumranet, and later helped maintain that project for Red Hat after the latter company acquired it.
The company has already packaged several applications as virtual appliances for OSv: Memcached (and a flash-storage optimized variant called Flashcache), Redis, and Cassandra. The company claims Java and its frameworks, along with Hadoop and NoSQL, are "being optimized and integrated to run on top of OSv."
OSv is meant to be a boon to management as well. It exposes its most crucial management interfaces as REST API endpoints and provides a mechanism for passing configuration data to an instance when it's spun up. Support for monitoring OSv instances through New Relic is also included, along with Jolokia for directly managing Java apps by way of JSON and REST.
Some of these early-stage decisions are savvy -- such as, providing close integration for enterprise Java applications. But entirely eschewing existing container technologies might come off as foolhardy, even for early adopters who haven't yet committed to containers themselves. One factor that might aid OSv in the short run is a converter that repackages Docker containers to work with OSv for the sake of convenience, in much the same way images for one variety of VM can be converted to run on another.
Users can download and run instances of OSv on their own hardware or on a cloud server, and the source is available under the BSD license. The company's also trying to get early adopters to sign up for a private beta program, where more bleeding-edge versions of the code will be provided.
I do not understand this cloud stuff. I just bought a new video editing program where 3 GB of cloud storage is included. I’m scratching my head, wondering just why would I let anyone have my personal videos, financial data, business workings, or whatever? Once someone else has it, you’ve lost physical ownership.
None of this would be secure once it leaves my own hard drive, which is seldom connected to the net.
I wonder the same thing all the time. Unless its a private cloud, where you (or your company) has physical control over if.
It’s all about convenience. If you store files on ‘the cloud’ (which is just a server someone else owns), you can access the files from anywhere you have internet. It’s pretty much the same as if you emailed the files to yourself, you can re-download them anywhere you want. Also, you don’t have to have your own large amounts of storage yourself.
However, I fail to see the point in it outside of work/sharing files with other people. I have no reason to put anything private or sensitive under someone else’s control. Not a good idea. I’ve heard stories about some cloud servers mentioning in their ToS that they have the right to your files and anything contained in them.
Interesting. Seems that corporate apps will likely go this way, the return of the mainframe, eh?
It’s all cyclic. we are now heading into another thin-client era. After that, we’ll come back to a think client era.
I had security drummed into me as a kid in the army (AR 380-5). I understand the convenience part, which is why I find it convenient to buy those little memory sticks. I own several, and can take data anywhere I want. I just do not understand todays crowd, not caring what they give to someone else to store for them, or what they say on places like Facebook or email for that matter.
The Cloud as computation resource is another story. If you were the only one using the computing resources of the Cloud and the Cloud had access to 1,000,000 processors then you could theoretically solve whatever problem you're trying to solve in one-millionth the time.
However, if millions of people are accessing millions of processors on the Cloud, or billions of people are accessing billions of processors on the Cloud then everyone has access, on average, to just one processor and there is no real speed-up.
My fear with the Cloud is that every time you make a simple request for information it will send thousands of processors going off to collect, dissect, analyze, and categorize your request in order to figure out exactly what type of ad to send to your browser.
Right now the early adopters are making all sorts of claims of how they are able to access millions of processors and analyze weather data, etc. But what happens when millions of bots from ad companies, or the Russian mob, or millennials are using all that processing power to dissect your buying habits or predict stock trends or make fantasy football trades?
I'm thinking that with Gresham's Law, over time the largest portion of computing power accessible via the Cloud will be for frivolous or noxious reasons.
Yup, I’m the same way. Everything is on my local machine, my backup external at home, or on usb drives I have. I have a box account, and I use it for work files that are too big for email (AV business, we have big ppts and videos, with corporate limiting email attachments to 25Mb each). But that’s it.
I can see the practicality for non-secure stuff like family vacation/wedding/etc photos, or the like. Stuff you want to share easily and don’t care who sees it.
Ideally, you would buy your own server, have it at home, and can then access it from anywhere yourself. But that requires a bit of $$ and a bunch of know-how that most people don’t care to learn.
Interesting idea. And I would think a natural evolutionary one. This could start a new technology trend in OSs.
That sounds like a lot, to me.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.