Fabric Computing

Arthur Cole

Arthur Cole spoke with Dr. Sharad Mehrotra, president, CEO and founder, Fabric7.

 

Cole: x86 technology has been the norm for more than 10 years. How would you rate the latest developments in that space - paradigm shifts or examples of the law of diminishing returns?
Mehrotra: I think we are seeing a dramatic change in what x86 can accomplish because of the advent of 64-bit and dual-core processing. They're going to take x86 architectures to the next level, comparable to the highest-performing PC platform in the world. Essentially, we will have the ability to connect many processors to create scalable multicore processing capabilities, and that's going to lead to a complete resurgence in x86 computing.


If you look at what we've been able to do with the Opteron, we've created a very powerful enterprise server architecture that matches or exceeds a lot of the features that you find in many other servers: partitioning, advanced I/O, virtualization. These are cutting-edge features that we've been able to add on top of Opteron.

 

Cole: What is fabric computing and how does it improve upon current server technology?
Mehrotra: The simplest way to think about it is the next-generation architecture for enterprise servers. Fabric computing combines powerful server capabilities and advanced networking features into a single server structure. The Q160, our flagship machine, starts with a scalable multiprocessor complex built around the Opteron. We've decomposed the processor complex into separate servers using hardware partitioning, and then added a powerful crossbar switch to provide virtualized I/O for networking and storage, plus built-in switching at Layer 2 and Layer 3. We end up with a new kind of server that can be partitioned and configured on the fly into different-sized servers using one chassis. When more capacity is needed, you can network a group together without third-party switching.


In the fabric computing view, resources are no longer tied to a single machine. A customer buying a typical server does not know exactly how to configure it or what applications to run. In our systems, you're not locked into a predetermined set of assets. You can reconfigure on the fly without adding software layers that slow everything down. Everything is done on hardware at full speed. Remember, we're not talking about just changing CPU memory. We're talking about changing the network I/O. It reduces a lot of the complexity that customers struggle with. You no longer reconfigure machine by machine. You have complete control of the entire fabric.

 

Cole: What sort of system or network reconfiguration is needed to implement fabric computing?
Mehrotra: None, actually. We were very careful about that. We knew we had to fit into a world where IT managers have a tremendous amount of equipment already in place. As a start-up, we can't depend on forklift upgrades. Our machines do not disrupt the IT infrastructure in the least. We connect to other infrastructures through Ethernet and Fibre Channel, and we run unmodified versions of Linux and Windows. Everything works below the OS and within the fabric with no changes to applications.



In the longer term, if massive adoption of fabric systems takes place, there will be some changes, but they will be beneficial to customers. As new Fabric7 systems get introduced into a fabric, you begin to remove more and more boxes from a building. Over time, you'll see fabric computing move to gradually replace legacy systems.



Add Comment      Leave a comment on this blog post

Post a comment

 

 

 

 


(Maximum characters: 1200). You have 1200 characters left.

 

 

Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.


 

Resource centers

Business Intelligence

Business performance information for strategic and operational decision-making

SOA

SOA uses interoperable services grouped around business processes to ease data integration

Data Warehousing

Data warehousing helps companies make sense of their operational data