I had a chance to speak with an interesting newcomer to the world of virtualization, Xsigo. Jon Toor, Xsigo's VP of Marketing, and I chatted about Xsigno's VP780 Network Director. I came away thinking that it's amazing how concepts come back again and again if one waits long enough.
One of the things that made mainframes perform and scale so well was the developers put intelligence wherever necessary to off-load the main processor so it could stay busy executing application code. These secondary computers were often called "channels" or "I/O directors." Once a message to a network or storage device was set up, the secondary computer would manage the whole process of sending or receiving data. The main processor was only interrupted when either an unrecoverable error occurred or the communications task was completed and the data was safely stored in memory.
As industry standard systems are being pressed into service to ever more critical business functions (and take work away from mainframes and midrange systems), the developers of these industry standard systems have run into problems that the mainframe folks solved 30 years ago. Often, they turn to the same approach that was previously successful.
Network and storage I/O have been bottlenecks for industry standard systems for quite some time. A focus on simplifying and amplifying the abilities of industry standard systems can be seen in what Xsigo is offering.
They've developed a network and I/O "director" that is essentially a specially designed processor that has been optimized to handle network traffic. Each industry standard system needs only have a single connection to this director and it is provided completely virtualized access to a broad array of network and storage devices. This would most certainly reduce the cost of these industry standard systems because it would be possible to reduce the number of I/O adapters, cables and headaches installed.
If you pop over to their site and read about what they're doing, I'm sure that you'll 1) see concepts you've seen before being implemented in a new way and 2) might see a potential solution to your organization's network and I/O bottleneck.