Virtualization – the killer ap

This year we have seen a lot being said about virtualization. The ingredients for a good virtualization story is that the technology should virtulize the CPU, memory, IO and storage. It is heartening to note that the open source community has all of these done via the inclusion of Xen in the kernel (for CPU, memory and IO), and with the Global File System that provides storage virtualization. With a fully open source stack, the world’s the limit in what we can achieve. To me, this is the evolutional jump in computing that is not unlike the jump from a single tasking OS (aka MS-DOS) to a multitasking OS (like OS/2, sorta Windows, and the Unix bretheren) on commodity hardware and now to multi-OS-tasking, again on commodity hardware.

With the 4 parts to virtulization, the key trump card is the ability to do live migration of the guest OSes. I demoed this to over 500 Red Hat partners in six cities in the last 30 days (Seoul, Mumbai, New Delhi, Chennai, Bangalore and Kolkata) using two Dell laptops (host #1 and host #2), a 100MB switch and with a video server running on one guest on host #1, and getting that guest to migrate live over to host #2, all while streaming the video to a client.

Live migration is now fully workable, and so it is stange to read that:
“So we are making the following changes, and postponing these features to a future release of Windows Server virtualization:
· No Live migration
· No hot-add resources (storage, networking, memory, processor)
· Support limit of 16 cores/logical processors (e.g., 2 processor, quad-core systems is 8 cores; or 4 processor, quad-core system is 16 cores)”

I guess all the talk about innovation is doable by a proprietary vendor is now put to rest. FOSS innovation wins, again. Even after acquiring a proprietary product from Viridian, MS is unable to deliver and continues to scale back on it’s vapourware offerings – but, hey, it is vapourware, so no surprises here.

Leave a Reply