Thanks to everyone who joined us for the July 18 DemoFriday with Plexxi and Iron Mountain! For those who couldn’t make the Plexxi and Iron Mountain Demo, you missed an informative presentation on how to use SDN to deliver a flexible, meshed network for managing transport and packet traffic between multiple data center fabrics.
After the demo, presenters from Plexxi’s product management division and Iron Mountain’s network services took questions from DemoFriday participants. Read the Plexxi and Iron Mountain Demo Q&A below, watch the full presentation and view archived assets, or check out the teaser video below.
What is the data center fabric throughput in terms of bandwidth?
Plexxi: Each switch contributes 24 or 48 10-Gb/s Ethernet links to the fabric, in a mesh form.
Is there any 40G?
Plexxi: Our switches have QSFP access for 4×10 or 1×40 (in various densities). The fabric is constructed as a mesh of 10-Gb/s Ethernet links.
What SDN protocol does your controller support?
What if you have more than two data center fabrics to connect? Do you have any examples for that?
Plexxi: You can connect multiple data centers together, with some topology restrictions. Contact firstname.lastname@example.org personally for specific details.
What is the max bandwidth Plexxi supports on a dark fiber side?
Plexxi: We use WDM for our 10-Gb/s Ethernet connections, so the maximum bandwidth supported over dark fiber depends on the network configuration and link properties. We can use passive or active optical systems to connect the system over dark fiber. With active systems, you can support up to 2 Tb/s of bi-directional capacity over a fiber pair.
Is the Plexxi SDN controller based on the OpenFlow standard?
Plexxi: It is not. Our controller calculates topologies rather than flows, and those topologies are passed to our switches.
Why not use OpenFlow for better interoperability?
Plexxi: Our solution is a combination of hardware and software. Many of the capabilities that we showed today are not possible using OpenFlow. Our network is run using topologies rather than explicit flow-forwarding rules. We are, however, looking at an OpenFlow interface onto the controller; timeframe is TBD.
How does the Plexxi controller handle redundancy?
Plexxi: The controller does not sit in the actual data path of user traffic, and its backend is a postgress database, which can be fully kept in sync with redundant versions.
Are you working with any open projects like OpenDaylight or OpenStack?
Plexxi: Yes and yes. We are part of OpenDaylight (our Affinity definitions have been donated to ODL) and we are working in ODL to create ODL-to-Plexxi-Controller interactions. We have started OpenStack integration just recently, to extract information from OpenStack and use it to drive network configuration and provisioning.
Any automation integration with Puppet?
Plexxi: We have done a sample integration with Puppet. We have full REST APIs with a Python library, and we did a sample Puppet integration that would create affinity groups based on their Puppet role.
Can you characterize the cooling and TCO challenges of your data centers compared to a large-scale cloud data center?
Iron Mountain: Iron Mountain’s National Underground data center is one of the most energy-efficient facilities in the world. As a former limestone mine located 220 feet underground, the facility has built-in natural efficiencies — the limestone walls and ribs that frame the white space serve as a natural heat sink and absorb 1.5 BTU/hr/ft2. Additionally, the facility has a 4-acre underground aquifer which we have configured as a geothermal cooling system to maintain the ambient temperature at a steady 52 degrees Fahrenheit year-round. All of these attributes combine to yield PUE ratio and TCO that is among the lowest commercially available today. Additional details on the energy efficiency of this facility can be found here.
Do most of your customers use an IaaS model, and do you consider interoperability with a hybrid cloud approach?
Iron Mountain: Iron Mountain is both carrier-neutral and cloud-neutral in our data-center colocation facilities. One of the differentiating features of our facilities is our data center marketplace, which provides customers with access to a wide array of cloud service providers that offer “as-a-service” solutions for cloud storage, archiving, disaster recovery, etc. Based on our secure, resilient platform and ecosystem of providers, customers can and do choose to go beyond IaaS/private cloud to establish hybrid cloud deployment models.
If SDN takes into consideration the IPv6 transition, what are the supported transition mechanisms?
Plexxi: The most important part of a software-defined network is that it operates at a higher layer than IP (Layer 3). A software-defined network is abstracted from the networking and routing layer. Our philosophy is that we abstracted the network to groupings of applications/customers/tenants and define policies on how the groups need to communicate. It does not matter whether the network uses IPv4, IPv6, MAC address, or some other mechanism to identify connected entities.
Any QoS-related considerations between switches?
Plexxi: For a Plexxi view on QoS, go the following link: https://www.youtube.com/watch?
Can I connect a Plexxi DTF over alien wavelengths?
Plexxi: Yes, Plexxi DTF can be connected over an alien wavelength by using transceivers with the appropriate WDM channel. The available channel information can obtained from the customer or service provider in order to select the appropriate transceivers.
Is a DTF only between data centers?
Plexxi: No. Plexxi’s data-center transport fabric is about connecting pools of resources. Those groups can be server farms, storage clusters, data-center pods, or actual data centers. Anywhere there is a physical separation of resources (small or large), Plexxi’s data-center transport fabric can be used to provide high-capacity, low end-to-end latency connectivity with superior control.
Can a Plexxi data-center transport fabric connect non-Plexxi networks?
Plexxi: Yes. Plexxi can be deployed as an intermediary between networks in place of WAN gateways or optical transport gear. In this scenario, Plexxi will provide load balancing across interconnect links. Traffic control extends as far as the Plexxi gear reaches, but that does not mean that data centers have to be end-to-end Plexxi.