Network Computing is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Liquid Computing Wants To Virtualize Everything

I've been digging into I/O virtualization lately and ended up speaking with some folks who want to virtualize everything. Everything. Data I/O. Network. Entire physical servers. Liquid Computing is hoping to be the next major disruption in the force.
Greg McElheran (CEO), Mike Kemp (CTO), and many of the staff of Ottawa-based Liquid Computing have a strong telco background. Frankly, this shows through in the concept and design of the company's fabric computing solution. While I think they were a bit uncomfortable with mainframe analogies, the modular design and highly scalable architecture would make any Big Iron fan comfortable.

From a virtualization standpoint, Liquid takes abstraction to a second level. VMware or Xen abstract guest OSes from underlying hardware. The LiquidIQ platform abstracts server resources from a flexible hardware pool. If this company takes off, virtualization terminology is going to get more confusing... Administrators can provision a virtual physical server from available CPUs, memory, etc. housed in the LiquidIQ chassis. Once everything (including VPN, firewall, storage paths) has been selected using the management console, the "new" server can be built as a traditional 1x1 platform or as an ESX or Xen host running guests as needed.

Hardware specs can be formidable: each chassis can run up to 20 compute modules (blades) supporting Opteron quad-cores. Up to 12 chassis can be linked. If I'm doing my math correctly, that yields 960 sockets. That's a bunch of available CPUs. Each compute module maxxes out at 64 GB of RAM. Each chassis runs a 200 Gbps backplane. Up to 10 10-Gb Ethernet and 10 dual-port 4 Gb HBAs can be configured per chassis. Everything is hot swappable; you think Vmotioning is neat? How about live-migrating a "physical" server or adding additional physical CPUs or memory on the fly?

Personally, I'd like to see the specs on a VM migration from one virtualized physical server to a second on the same chassis at 200 Gb. It makes our virtualization test lab GigE switch seem a wee bit poky.

Liquid Computing is just getting started; it has a handful of customer sites in production. The principals talk about "fabric computing" with an almost religious fervor. Check 'em out.