Crowdfunding Campaign Seeks a Fully Open Source Alternative to Citrix XenServer (kickstarter.com) 66
"Free/libre and 100% community backed version of XenServer," promises a new Kickstarter page, adding that "Our first prototype (and proof of concept) is already functional."
Currently, XenServer is a turnkey virtualization platform, distributed as a distribution (based on CentOS). It comes with a feature rich toolstack, called XAPI. The vast majority of XenServer code is Open Source.
But since XenServer 7.3, Citrix removed a lot of features from it. The goal of XCP-ng is to make a fully community backed version of XenServer, without any feature restrictions. We also aim to create a real ecosystem, not depending on one company only. Simple equation: the more we are, the healthier is the environment.
The campaign reached its fundraising goal within a few hours, reports long-time Slashdot reader NoOnesMessiah, and within three days they'd already raised four times the needed amount and began unlocking their stretch goals.
But since XenServer 7.3, Citrix removed a lot of features from it. The goal of XCP-ng is to make a fully community backed version of XenServer, without any feature restrictions. We also aim to create a real ecosystem, not depending on one company only. Simple equation: the more we are, the healthier is the environment.
The campaign reached its fundraising goal within a few hours, reports long-time Slashdot reader NoOnesMessiah, and within three days they'd already raised four times the needed amount and began unlocking their stretch goals.
Nice (Score:5, Insightful)
Re:FOSS must learn to organize and collaborate (Score:4, Insightful)
Remember that when you're next in a hospital and need heart surgery. I don't know about you but I'd rather have someone specialized to the task.
Software development requires a specific skill set, time, and energy that not everyone has. Despite all that bullshit Bill Gates et al are spewing about everyone learning to code, not everyone can code. Even if they had the talent, they may not have the time to learn it on top of whatever else they're doing.
Your argument is breathtakingly ignorant, and a perfect example of the self-important attitude that keeps Linux and most other OSS projects from going mainstream.
Re: (Score:3)
To continue your analogy... Would you rather have a heart surgeon who learned his craft from a secret society using methods which have not been openly peer reviewed OR would you like a heart surgeon who studies all of the open literature on heart surgery and learns the best practices from his/her peers.
Re: (Score:3)
That's actually not the comparison I was making. I was trying to point out not everyone who uses software, has the time AND the energy AND the skill to also maintain software. A heart surgeon, for example, cares about medicine. They care about saving lives. It is a demanding job that already requires a lot of time and energy and skill to master, and that doesn't leave much room for software development.
But according to the GP, because they don't ALSO know how to do software development, said heart surge
Re: (Score:2)
To your point, yes, medical knowledge is based on generations of doctors each standing on the shoulders of the giants before them. But the analogy still breaks down badly because if we tried to continue it, you would have small groups of heart surgeons divvying themselves up into different camps who all proclaim that their software is the best and insist on re-inventing the same medical procedures over and over again because of NIH syndrome.
Your "different camps" are just like proprietary software developed by different companies who don't share trade secrets.
Open source heart surgery publishes results for everyone to learn from.
Re: (Score:2)
Except in Open Source we'd have 5 different camps not doing heart surgery but squabbling over which scalpels to use and designing new scalpels that don't work in standard autoclaves because of their 3d printed ergonomic handles and the actual practice of the surgery would be the exact same as it is now.
Re: (Score:2)
Re: FOSS must learn to organize and collaborate (Score:1)
Because everyone always hires the best for a specific task. Ever hear of Therac-25 maybe you should look it up.
Closed source/open source it does not matter.
For the open source advocates: Tell me when was the last time you actually looked over every line of code in an open source system? For how long did you monitor all updates after? How many systems were included? Who is responsible for all of that? How do you know you can trust them? If someone paid to have the software analysed and a bug that they could
Re: (Score:2)
Nothing is perfect, open is still a few steps ahead of closed because the opportunities are available to you even if you choose not to take them, and everyone has an equal level of access to the code.
Every problem that *can* happen to open source, can and does happen to closed source code as well, but closed source has an additional set of problems too.
Re: (Score:2)
But even with open source, the average user (who is not a programmer) still has no idea what their systems are doing.
I think this is accurate to a large extent, but I'd similarly argue that licensing is less relevant to such users as well; free-as-in-beer matters more to such users than free-as-in-freedom. It's possible that Xenserver 7.2 will continue to live for quite some time; bare metal hypervisors generally need updating less often than guest OSes (well, except Hyper-V). Xenserver is, for the most part, in competition with Proxmox, the free version of ESXi, the free version of Hyper-V Server, and of late, Docker in
Re: (Score:2)
The timing of the Spectre / Meltdown bug and the release of the new Xen Server 7.3 is certainly not fortuitious, along with the fact that they said they would not release the patch for 7.2 and lower.
They were probably informed in advance of the bugs and had the time to plan this mi
Comment removed (Score:5, Insightful)
Re: (Score:1)
They didn't build it. They're standing on the shoulders of giants already and just wanting more of the pie. XenServer is not a shiny shiny new product made from scratch by Citrix. It's based on open source. They're just exploiting loopholes in the licensing by closed-sourcing what they can to attempt to drive valuation. Embrace. Extend. Extinguish.
Re: (Score:1)
They didn't build it. They're standing on the shoulders of giants already and just wanting more of the pie. XenServer is not a shiny shiny new product made from scratch by Citrix. It's based on open source. They're just exploiting loopholes in the licensing by closed-sourcing what they can to attempt to drive valuation. Embrace. Extend. Extinguish.
I think you don't know the history of Xen.
"They" that built Xen are the University of Cambridge alumni that created Xen and later founded XenSource, that company which Citrix paid $500 million for.
The shoulders of giants you refer to have been employees of Citrix since 2007 and paid by Citrix to continue development of Xen both as the enterprise product and the open source product.
Re: (Score:2)
AMD GPUs are popular with Open-source rigs, especially since the OSS driver performance is on par or better than the closed-source one even on recent GPUs. but several things play agains AMD becoming more popular on OSS rigs.
GPUs are no longer used only for gaming. For instance AMD GPUs are more efficient at crypto-mining, meaning that there is a premium on them and stocks have been low. This is great news for AMD by the way but not so much for gamers. Meanwhile Nvidia has been investing a great deal in GP
Removes a single point of failure--or e$ploitation (Score:1)
Bravo!
Why not KVM? (Score:2)
Pardon my ignorance on the topic but what can this offer that Linux's KVM cannot?
Re: (Score:3)
This page [linux-kvm.org] has KVM's "ToDo" list. A good number of items on that list are supported by Xen. In addition, KVM will not play well [if at all], with older CPUs made prior to extensions enabling virtualization.
KVM also doesn't work with Intel's Atom CPUs unless extensions are available.
Re: (Score:3, Interesting)
KVM also doesn't work with Intel's Atom CPUs unless extensions are available.
It doesn't? In my last company we used an Atom C2000 and we used KVM/Libvirt to run VMs on it using Ubuntu 14.04. In fact we had one design win that depended on it.
You may be thinking of the feature (I forget the code name) that lets you virtual-ize PCI devices. It couldn't do that so you had to rely on the linux kernel bridge or OpenVSwitch.
Re: (Score:2)
I've run KVM and OpenStack (using KVM) on Atom based systems it definitly works.
Why not oVirt? (Score:2)
"Build a UI that controls all the VM's with some kind of web front end.
One of the tricks is going to viewing remote systems. There is a remote desktop variant that encodes displays as h264 or similar. That should be viewable in a web page. Better yet, if you can somehow tie into existing paravirtual drivers in common distributions, though there are some details there to work out."
You mean like oVirt (www.ovirt.org), which has all of these (or equivalent features) and more, including live migration, HA s
Re: (Score:2)
There are a variety of systems out there which do this...
Proxmox for one, SolusVM, and many more.
Re: (Score:2)
KVM will not play well [if at all], with older CPUs made prior to extensions enabling virtualization.
KVM is a module for doing virtualization in qemu. It requires VT (or equivalent) but you can still use qemu without KVM. However, for the target market, that's not a drawback. Statistically nobody is running KVM or Xen on a processor so old it doesn't have virtualization extensions.
Re: (Score:2)
Virtualization extensions are not exactly new, they've been around for more than 10 years now... It's highly unlikely that anyone will be deploying such old processors on a hypervisor system.
KVM works fine on many models of Atom, i've used it on a few. The really lowend models are again not the sort of systems you'd actually be using something like this on.
Re: (Score:2)
Xen supports para-virtualization, which was built into the original GPL version of Xen and remains useful to approach "bare-metal" speed for the virtual machines. Much of the need for this has been reduced through the development of "docker", which can be treated much like a Xen based para-virtualized VM with instances of even lighter weight.
The very active CentOS Xen community has, as I've observed, been much larger and much more active than KVM in dealing with new server and guest environments. My informa
Re: (Score:1)
The virtio drivers provide paravirtualization for KVM. I don't know how they compare with Xen but it used to be quicker than VirtualBox.
Re: (Score:2)
Are those working well? I'd not tried them the last time I worked with KVM.
Re: (Score:2)
Works very well, supported out of the box on linux for many years and available as an installable driver package for windows, there is now support for virtio in various bsd flavors too.
Re: (Score:1)
Paravirtualization is really only useful when the host CPU doesn't have the hardware necessary for virtualization. Since that is almost no x86 CPUs these days, the value of paravirtualization is low.
In addition, KVM supports Virtio drivers for network and disk drives, reducing the value of the paravirtualization approach even more.
Re: (Score:1)
"Paravirtualization is really only useful when the host CPU doesn't have the hardware necessary for virtualization."
That's not true. Paravirtualization can provide better performance in either case. From the virtio wiki [libvirt.org]
Re: (Score:2)
Excuse me, but "para-virtualization" is an optional feature and an effective performance enhancement for virtualization: Every virtualization technology supports hardware virtualization.
Re: (Score:2)
"Paravirtualization is really only useful when the host CPU doesn't have the hardware necessary for virtualization."
That's not true. Paravirtualization can provide better performance in either case.
Only in the case where your hardware doesn't support VT-d and SR-IOV (other hardware features built for IO virtualisation).
Re: (Score:2)
Xen supports para-virtualization, which was built into the original GPL version of Xen and remains useful to approach "bare-metal" speed for the virtual machines.
No, para-virtualisation is no longer useful. PV is dead, kiilled by Spectre.
Much of the need for this has been reduced through the development of "docker", which can be treated much like a Xen based para-virtualized VM with instances of even lighter weight.
WHAT???? You can't compare PV with docker (no virtualisation, just process containment via namespaces). HV with accelerated IO drivers has displaced PV.
The very active CentOS Xen community has, as I've observed, been much larger and much more active than KVM in dealing with new server and guest environments.
What???? That might seem to be the case because very few people use Xen, so the CentOS Xen community it the only place there is any open-source concerted effort around Xen. But, that is because all the concerted open-source virtualisation effort is happening on Linux/KVM.
My information may be out of date: this may also just mean KVM worked well since then. But I've seen a number of clients simply give up on KVM and just switch to Xen or Citrix Xen. due to unexpected limitations and the time necessary to spend valuable engineering time tuning their own virtualization servers. It's no longer on my recommended product list.
KVM isn't a
Re: (Score:2)
XenServer has a full set of tools and a comprehensive GUI for doing anything and everything with the host and VMs. And this is the free version. The Paid version is very generously priced and provides a few more really nice features.
The last time I tried using KVM, there wasn't a single decent management app for KVM that didn't also cost an absurd amount of money and still couldn't do everything that XenCenter/XenServer could do without extreme fiddling. Whether that's still the the case, I'm not sure.
Re: (Score:2)
XenServer has a full set of tools and a comprehensive GUI for doing anything and everything with the host and VMs. And this is the free version. The Paid version is very generously priced and provides a few more really nice features.
The last time I tried using KVM, there wasn't a single decent management app for KVM that didn't also cost an absurd amount of money and still couldn't do everything that XenCenter/XenServer could do without extreme fiddling. Whether that's still the the case, I'm not sure.
When was that? 4+ years ago? oVirt [ovirt.org], the open-source project that Red Hat Virtualisation is built on, supports all the features (and many many more) that (than) XenServer has in the paid version. In fact, it has all the features that RHV has, except the branding and commercial support.
oVirt has been viable as a complete open-source solution since about the 3.3 release [ovirt.org] which was in Sept 2013. 3.0 was the first release to ditch the old .Net-based UI that came from Qumranet that was still shipped in RHEV 2.x.
Re: (Score:2)
Quite possibly. One doesn't up and change their entire VM infrastructure at the drop of a hat, so there's little incentive to keep up on up and coming technologies unless you're explicitly looking to make a change. Or if something comes up that so newsworthy that it manages to punch through all the noise.
oVirt has definitely improved a lot since I last looked. However, one of my biggest complains still stands. You have to configure a frightening amount by hand without any tools to automate. I am leary
Re: Why not KVM? (Score:2)
oVirt has definitely improved a lot since I last looked. However, one of my biggest complains still stands. You have to configure a frightening amount by hand without any tools to automate.
Wrong. You can script setup of the entire solution, ujt the exact methods may differ (depending on how you deploy the manager, as a VM or stand-alone server, and previously also whether you installed the nodes from image or on top of RHEL (answer file or kickstart).
New in the current release are provided ansible roles for setting up varoois components (I havan't dug into this though).
Considering that it explicitly targets RHEL/Centos7 and only that, there should be no reason why there isn't a whole bunch of best practise automation mechanisms to set up key elements like hooking up the ISCSI (or NFS) backend.
Anything for setting up the manager (including as a VM on the first host) can be scripted via engine_setup or hosted_engine_s
Re: (Score:2)
Wrong. You can script setup of the entire solution, ujt the exact methods may differ (depending on how you deploy the manager, as a VM or stand-alone server, and previously also whether you installed the nodes from image or on top of RHEL (answer file or kickstart).
Key part bolded. Again, (and this is based on a very cursory viewing of the admin docs) this isn't provided by them. You have to write your own scripts and set up your own automation. That means you now have to learn to manage and use not only oVirt, but Ansible and other tools as well.
The primary target of the commercial versipn is to replace VMWare. But that doesn't always mean expensive (see hyper-converged).
Okay, fair enough. Enterprises with elaborate VMWare setups are going to already have the resources to set something like oVirt up.
Unfortunately that's a handicap for smaller companies that are just getting started with t
Re: (Score:2)
Wrong. You can script setup of the entire solution, ujt the exact methods may differ (depending on how you deploy the manager, as a VM or stand-alone server, and previously also whether you installed the nodes from image or on top of RHEL (answer file or kickstart).
Key part bolded. Again, (and this is based on a very cursory viewing of the admin docs) this isn't provided by them. You have to write your own scripts and set up your own automation. That means you now have to learn to manage and use not only oVirt, but Ansible and other tools as well.
It's a bit hidden (but hinted at) in the RHV documentation (but the options are in the engine-setup --help, and it always tells you it is writing answer files). See this page which has a section on answer files [ovirt.org].
Ansible is an additional option for shops that already have it, not required for scripting installation.
Okay, fair enough. Enterprises with elaborate VMWare setups are going to already have the resources to set something like oVirt up.
Anyone who hasn't got the resources to run a hyper-converged (IOW, software-defined storage cluster on the same hosts that run VMs, instead of expensive storage appliances that have sufficient redu
Need to have (Score:2)
Re: (Score:2)
I agree this would be a good thing... but I’m not sure Kickstarter is the way to go for something where you’d ideally want broad corporate backing, given Kickstarters have a habit of just disappearing with your money.
Of course I’m assuming people want this professionally. If it’s really the case that people want it for their home hobby servers, then Kickstarter is fine.
That name though... (Score:3)
Really no one could come up with a better name than "XCP-ng"?
Re: (Score:3)
Re: (Score:3)
raised four times the needed amount (Score:2)
Sorry I don't follow a lot of kickstarter campaigns, but of the ones I have heard about it seems like the ones that over raise are more likely to fail. I think it is a matter of their eyes are bigger than their ability to execute. I will be interested to see if this one can keep the
Re: (Score:1)
I'd feel the same way if they hadn't already released a working proof-of-concept. It's based on XenServer 7.2 while what they'll be working on will be based on 7.3. A good number of the stretch goals have been done by users on the side for one purpose or another at times or by folks inside the Xen project or involved with XenOrchestra as proofs-of-concept as well. For example, software RAID is fairly common in small lab settings for XenServer installations, one I've used myself.