Fabrice Bellard is one of my heros. Humble, talented, and prolific - see http://bellard.org/
Not less importantly, most of it is open source, and in fact, much of it making today's open source tick as well as it does:
QEMU formed the basis on which VirtualBox, Xen and KVM were built. They would have been possible without it - but it was the shoulders-of-giants[1] they could step on.
FFMPEG makes every video related website or service in the world tick. (Again, same shoulders-of-giants principle)
And he's done so much more.
Thanks, Mr. Bellard.
[1] Yes, I'm aware that Newton was mocking Hooke. I'm using it in the commonly used sense.
Thanks for pointing that out. I've seen several of these projects as they've come across the tech scene over the years, but I had no clue they were all done by the same guy.
However, being able to start these projects, get them far enough that they gain traction (usually completely on his own), manage them until enough contributors have gathered so that the project is self sustaining -- and then leave them until the next project -- is a non-trivial task, and in my opinion not less impressive than Linus' handling of the linux kernel.
He had a share of public fame with jslinux, a lot of people realized how much he meant to all of us. Also he competes for the fastest pi program, he aces both ends of the spectrum. Impressive.
Bellard's work on computing Pi is possibly even more impressive.
In '97 he developed a new algorithm for calculating Pi 43% faster than the previous best[1].
In 2009 he broke the record for the largest calculated value of Pi (2700 billion digits) using a program he developed on a single desktop computer. Prior to that every previous record since '95 had been calculated on a supercomputer. Notably the immediate previous record used 640 nodes of the T2K Open Supercomputer (Appro Xtreme-X3 Server) at the Center for Computational Sciences, University of Tsukuba[2]
It's true that Bellard took around 130 days and the super computer did it in a day - but nevertheless the difference in pure speed is amazing.
Edit: I almost forgot his International Obfuscated C Code Contest winning Obfuscated Tiny C Compiler. A C compiler than can compile itself in 2048 bytes of C source excluding the ';', '{', '}' and space characters.
The RDMA migration has some impressive[0] performance promises. I can't wait to try it out, current (well, ok, wheezy) kvm migration pause times aren't so hot.
There are a lot of guides online if you search for how to set up KVM (which uses QEMU). Linux distros also provide documentation specific to their distro:
What distro are you most comfortable with? We run KVM on Debian in a few offices, with the VMs doing mail, web, and DNS. It's been about a year and it's been rock solid.
Would anybody be able to link me to some information to help me understand where qemu fits in to, say KVM? I'm very familiar with various VMware technologies and am trying to get a handle on other platforms, like KVM. Since all the tools seem to use libvirt to abstract things, I don't have a clear understanding of where the boundaries of responsibility are.
I can see that the process names that represent my VMs are "qemu", and I use qemu* commands to say, manipulate disks. Maybe it is all qemu, and KVM is the part that makes it accessible to non-privileged users?
Fabrice Bellard is one of my heros. Humble, talented, and prolific - see http://bellard.org/
Not less importantly, most of it is open source, and in fact, much of it making today's open source tick as well as it does:
QEMU formed the basis on which VirtualBox, Xen and KVM were built. They would have been possible without it - but it was the shoulders-of-giants[1] they could step on.
FFMPEG makes every video related website or service in the world tick. (Again, same shoulders-of-giants principle)
And he's done so much more.
Thanks, Mr. Bellard.
[1] Yes, I'm aware that Newton was mocking Hooke. I'm using it in the commonly used sense.
Thanks for pointing that out. I've seen several of these projects as they've come across the tech scene over the years, but I had no clue they were all done by the same guy.
Incredible.
They were started by the same guy. Bellard hasn't worked on ffmpeg for almost a decade. I suspect the situation is pretty similar for QEMU.
Projects this big are the work of lots of people.
True.
However, being able to start these projects, get them far enough that they gain traction (usually completely on his own), manage them until enough contributors have gathered so that the project is self sustaining -- and then leave them until the next project -- is a non-trivial task, and in my opinion not less impressive than Linus' handling of the linux kernel.
He had a share of public fame with jslinux, a lot of people realized how much he meant to all of us. Also he competes for the fastest pi program, he aces both ends of the spectrum. Impressive.
BTW, has jslinux been open-sourced yet? It wasn't last time it made news.
I have no idea, I only remember people trying to reverse it https://github.com/levskaya/jslinux-deobfuscated
What happened to that software-based LTE station thing for PC's of his?
He turned it into a company.
Amarisoft: http://www.amarisoft.com/
Bellard's work on computing Pi is possibly even more impressive.
In '97 he developed a new algorithm for calculating Pi 43% faster than the previous best[1].
In 2009 he broke the record for the largest calculated value of Pi (2700 billion digits) using a program he developed on a single desktop computer. Prior to that every previous record since '95 had been calculated on a supercomputer. Notably the immediate previous record used 640 nodes of the T2K Open Supercomputer (Appro Xtreme-X3 Server) at the Center for Computational Sciences, University of Tsukuba[2]
It's true that Bellard took around 130 days and the super computer did it in a day - but nevertheless the difference in pure speed is amazing.
Edit: I almost forgot his International Obfuscated C Code Contest winning Obfuscated Tiny C Compiler. A C compiler than can compile itself in 2048 bytes of C source excluding the ';', '{', '}' and space characters.
[1] http://bellard.org/pi/pi_bin/pi_bin.html
[2] http://www.hpcs.is.tsukuba.ac.jp/~daisuke/pi.html
The RDMA migration has some impressive[0] performance promises. I can't wait to try it out, current (well, ok, wheezy) kvm migration pause times aren't so hot.
[0] http://wiki.qemu.org/Features/RDMALiveMigration#Performance
Michael Hines has done a fantastic job getting this series merged. It's a rather invasive change and I am amazed at how quickly it was merged.
I'm thinking on virtualizing some from servers (web front and database) which are physicall now. Do you have any experiences to share?
virtualization is a pretty small niche, not sure anyone actually does it. /s
There are a lot of guides online if you search for how to set up KVM (which uses QEMU). Linux distros also provide documentation specific to their distro:
https://access.redhat.com/site/documentation/en-US/Red_Hat_E...
https://fedoraproject.org/wiki/Getting_started_with_virtuali...
https://help.ubuntu.com/community/KVM
If you need help, ask on #qemu on irc.oftc.net or #kvm on chat.freenode.net.
What distro are you most comfortable with? We run KVM on Debian in a few offices, with the VMs doing mail, web, and DNS. It's been about a year and it's been rock solid.
Would anybody be able to link me to some information to help me understand where qemu fits in to, say KVM? I'm very familiar with various VMware technologies and am trying to get a handle on other platforms, like KVM. Since all the tools seem to use libvirt to abstract things, I don't have a clear understanding of where the boundaries of responsibility are.
I can see that the process names that represent my VMs are "qemu", and I use qemu* commands to say, manipulate disks. Maybe it is all qemu, and KVM is the part that makes it accessible to non-privileged users?