Cloud & Data Center

Fiber Mountain Helps Data Centers Scale Affordably

By Erik Linask, Group Editorial Director  |  July 08, 2014

INTERNET TELEPHONY magazine recently interviewed M. H. Raza, founder and CEO of

Fiber Mountain Inc., about how the company is leveraging data center infrastructure management, software-defined networking, and virtualization to take data centers to the next level.

What is data center infrastructure management, and what is its role in today’s highly complex data centers?

DCIM software collects, analyzes, organizes and represents critical data center information and metrics, and associates them with the physical location of infrastructure assets in the data center. In addition to tracking power, heat and space, DCIM is used for tracking of virtual resources, software licenses, and various other metrics. Many systems have a physical topology diagram of network layout and equipment, again associated with each physical location, however the documentation process to achieve this is manual, costly to maintain, and usually not 100 percent accurate as it requires frequent human audits. We see DCIM software evolving to include automatic discovery of network devices such as servers, switches, storage, patch panels and cable, and also to be able to automatically and accurately document these devices and cables into cabinets, racks and connectivity diagrams, without human intervention or audit, which will be significant as data centers become more complex.

What trends are driving evolution in data centers and creating demand for DCIM and SDN capabilities?

DCIM and SDN together, this is a very interesting question! The underlying trend that is driving evolution in data centers is growth in size of network infrastructure; the network is growing both in capacity and complexity. DCIM is required to document capacity and help plan and manage the growth. SDN addresses the complexity side of the equation by promising to simplify the network through virtualization. The challenge is that today’s DCIM solutions do not have the intelligence to dynamically associate network devices with their physical location, and consequently lack the ability to associate packet flow metrics with the location of physical assets.  Hence, it is difficult to make the leap that DCIM solutions can document network metrics from virtually abstracted layers of a software-defined network when the DCIM solution is not even able to dynamically document the very hardware that is used by the SDN. We are now looking at next generation network-aware DCIM solutions, which will be able to bridge the gap between the physical and the logical, and understand network flows and congestion as it relates to each piece of physical hardware and corresponding physical paths and locations, and be able to dynamically document the changing abstraction layers. Applications on these layers include cloud apps, content delivery, business continuity, disaster recovery and security.

SDN is driving network transformation through the virtualization and programmability of switching and routing infrastructure, creating much more agile and flexible network environments. Are current SDN standards and capabilities enough to accommodate these trends?

The level of agility and flexibility that SDNs provide will continue to evolve for a long time. We started by creating programmable abstraction layers over existing switch and router infrastructures; and we knew that progress in SDN would change the way we build and architect networks. Industry collaborative efforts under way today aim to increase multivendor interoperability, including white box devices, and [are] a precursor to the evolution that is coming in SDN capabilities. I also believe that advancements in photonics and optical switching along with SDN’s inherent ability to control packet flows will drive a network infrastructure transformation which will require more discussions in standards bodies and start a race toward new features and capabilities.

How is Fiber Mountain filling the gap in current SDN models?

Traditional networks are based on switching and routing principals that were formed decades ago when the problem sets were different than today. With SDN we have an opportunity to transform the network, but to get there we need software control of every hardware element that we install in our data centers, from the servers all the way down to the physical layer cabling that connects the servers to the network. Fiber Mountain’s definition of network virtualization or SDN includes the software controlled configuration and reconfiguration of switches, servers, storage, and also the physical layer which is the lowest building block for a connected group of devices, or network. Fiber Mountain is presenting clients with a standards-based solution through a revolutionary fiber core allowing direct connection of critical data flows between resources.   

Explain Fiber Mountain’s Glass Core and what it means to traditional packet aggregation points in the network core and end-of-row and top-of-rack switches.

Over the past years, as bandwidth needs have grown, network infrastructure in the data center has gone through several transitions, one of these being the proliferation of packet aggregation points. First there was the core switch, then we created a second packet aggregation point at the edge of the row (end-of-row switch), and then a third packet aggregation point at the edge of the rack (top-of-rack switch), and now a fourth packet aggregation point at the edge of the server. Fiber Mountain has developed a glass core, which replaces several packet-processing hops in the network, allowing direct connections between edge devices. This results in less packet processing in the network, which directly translates to fewer switches, less power consumption, less heat dissipation, less space requirement, and a better TCO model for the data center.

What is Fiber Mountain’s network fabric, and how does it enable the kind of infrastructure provisioning and management Fiber Mountain envisions?

Fiber Mountain provides an optical core, which is made up of a high density of fiber-optic cable strands. These strands of fiber are a consumable resource that can be programmed via software to deliver 10, 40, or 100gbps connectivity between any two points. Network virtualization is not complete without the ability to virtualize switches and the cable that connects the switches.

What does the increased fiber density mean for network latency?

Network designs proposed by Fiber Mountain definitely have a higher level of fiber cable density, particularly in the core of the network. The increased fiber connectivity results in a reduction of core switches and latency associated with these switches. A Fiber Mountain network has far less latency than traditional three-tier or two-tier networks.   

Network discovery and diagramming, especially down to the individual cable level, is a cumbersome task. How does Fiber Mountain ease this burden?

Fiber Mountain’s Alpine Orchestration System separates the control and management plane from switches and other active devices, and also creates an abstracted control and management plane for fiber optic cabling. The ability to have a unique identifier for each fiber optic cable strand, and the ability to communicate in real time with each strand, is the key ingredient to dynamically discovering cable connectivity and reconfiguring it within the glass core.

How will data centers deploy Fiber Mountain technology within their existing infrastructures?

Fiber Mountain technology is designed and developed to co-reside with a customer’s legacy network. A customer may choose to integrate the technology at a pace that they are comfortable with, a single rack or an entire row at a time. The technology will work seamlessly with an existing infrastructure as it has industry standard interfaces and is tested with best-of-breed network devices.

How will Fiber Mountain’s approach to SDN change how and where data centers are built?

Fiber Mountain has the unique advantage of being able to virtualize servers, switches, and cabling. This truly enables remote management and control of the data centers that are geographically dispersed. The ability to manage multiple data centers from a single pane of glass also helps reduce operations cost. Fiber Mountain creates a foundation where customers will be able to manage remote data centers with the same ease as managing a local data center and also be able to scale the data center as their needs grow.

How will the idea of connectivity virtualization change data center economics?

Fiber Mountain’s technology provides new simplicity, higher performance, lower latency, and real-time managed networks that can also boast a significant reduction in operating costs with less power, heat, cooling, and space. Data center economics can be summarized as: Two times the capacity for half the cost.




Edited by Maurice Nagle
blog comments powered by Disqus