CALIENT Technologies, Inc. has announced its LightConnect™ Fabric Manager interconnect orchestration software for the S-Series family of Optical Circuit Switches (OCS).
LightConnect was developed to enable new virtual Pod (V-Pod)-based data centers by facilitating the orchestration of an OCS-based interconnect fabric for intra-data center connectivity allowing compute and storage resources to be shared between physical Pods in data centers.
Pod-based data centers, as popularized by Google, Facebook, Amazon, and others, are built using an easily replicated, packaged pod architecture including compute, storage and networking functionality. Pods are largely popular due to the practicalities of scaling in very large data centers where growth requires that compute resources can be brought online in a modular fashion, while the Pod also simplifies maintenance activities.
While Pods are a flexible way to grow data center functionality, their self-contained nature creates a natural resource boundary that limits intra-Pod resource sharing. This is important because workload compute demands are elastic, varying as often as several times a minute, and there must always be additional available resources to handle peak demands.
In order to allow plenty of capacity for peak demands, Pods typically run at low average levels of utilization, with less than 40% being typical according to industry sources.
Factoring in that 85% of data center hardware capital expenditures are associated with servers and storage, this means that on average, a massive amount of data center investment is wasted. This was noted by Google’s Urs Hoelzle at the 2013 “How Green Is The Internet” symposium, where he argued that, “server and data center under-utilization is one of the primary sources of waste and inefficiency in computing.”
The LightConnect Fabric-driven V-Pod solution overcomes this challenge by creating an optical fabric that allows racks or rows of compute resources to be shared between physical Pods across the data center to accommodate varying workload demands.
Now data center operators can run Pods at high levels of utilization, with the knowledge that additional compute resources can be allocated dynamically from other Pods if needed. As a result, the entire data center can be dimensioned to support an overall average peak demand, rather than having to do so for each individual Pod.
Core functions of the Fabric Manager include a topology manager and a cross-connect manager. The topology manager maintains a database of the data center network topology including which layer 2/3 switches and ports are connected to specific optical ports on the LightConnect Fabric.
Similarly, the cross-connect manager maintains a complete database of all optical cross connects in the LightConnect Fabric. Together, these two core functions and their databases allow the Fabric Manager to maintain a complete view of overall topology and optical connectivity in the data center.
The Fabric Manager also maintains a consolidated dashboard summary with operational status of all optical circuit switches including alarms, events, and port utilization. The Fabric Manager is a CentOS-based Linux Application, capable of running on off-the-shelf hardware or any virtual machine
“Operational cost efficiency is a compelling competitive advantage in cloud data centers, which makes any innovation that improves compute economics — the biggest contributor to cost — an essential consideration for data center architects and managers,” said Atiq Raza, chairman and CEO of CALIENT. “Together, the new LightConnect Fabric and Fabric Manager enable a completely new paradigm in data center design and optimization that will support tremendous improvements in server and storage utilization. It is simply not possible to achieve these results with existing layer 2/3 switching solutions, because the round-trip latency of packet-switching is several microseconds every time the hop is made and that is far too great a penalty.”