Cumulus Networks enables us to give users maximum choice in hardware, software and tools so they can build a solution based on the problem at hand instead of deploying a solution because it’s the only thing a vendor provides. With Cumulus Linux as the networking OS, networking gear is finally at par with servers because users can manage their networking switches using the same tools as a Linux server. Because getting a switch up and running is now as familiar as deploying a server, our users have significantly reduced their OPEX investment.
Higher Education and Research
Choice and Simplicity
Open Compute Project (OCP), Quanta, Edge-Core, Ansible
The University of Texas at San Antonio (UTSA) is a state research university in San Antonio, Texas, United States. With over 30,000 students, it is the largest university in San Antonio. UTSA is also home to the first Open Compute Project (OCP) Certification and Solution Laboratory in North America. The laboratory serves the goal to certify Open Compute technologies and key workloads for large enterprises while educating its own students on cutting edge open source technologies and providing research opportunities.
Having seen at first hand the benefits of open technologies and ecosystems, it was only natural for UTSA to adopt Cumulus Linux, the Linux network OS that enables a rich open ecosystem and that works on top of the Open Network Install Environment (ONIE), the essential building block for Open Networking.
As a rapidly evolving environment, the data center of UTSA must grow at a rapid pace while enabling research projects based on the latest technologies, within the constraints of a tight budget. Challenges faced by UTSA team included:
Research projects rely on the latest technologies. While it was fairly easy to leverage and repurpose assets for the compute environment, the same could not be said about networking. With a vertically integrated stack comprising networking hardware, an operating system, and features, it was difficult to reuse existing assets while optimizing budgets for specific research projects and planning for potential vendor-specific exceptions.
To get away from vendor lock-in, UTSA was evaluating number of options, including adopting Open Networking model, the disaggregation of hardware from the operating system. Open Networking would enable them to get to the best technology at the right cost, where, once they operationalized the software layer, they could swap out the hardware as needed.
Beyond budget constraints, having various vendor networking solutions also means that the staff needs to understand various user interfaces (different CLIs) and tools. Each vendor has their own hardware, features and tools that are proprietary and are not easily repurposed. Networking operations have been a fairly manual process, adding to time constraints for limited IT staffs. The university had spent many cycles in the past migrating from one vendor to another; the process had been lengthy and took time to adjust. Proprietary technology and not interoperable added to the overall complexity.
UTSA leveraged OCP networking technologies with Cumulus Linux in the UTSA Cloud and Big Data Laboratory and in various cloud projects for research and education. Over 90% of server ports are based on OCP technology. With Cumulus Linux, now two thirds of the network infrastructure are also based on OCP networking with Quanta and Edge-Core switches running Cumulus Linux. Pods are dedicated for research projects, specific solutions (like Big Data and OpenStack) or for staff services, so there is not a single topology in place. Instead, the model is to build a flexible solution that can be adapted to many use cases.
Open Network Install Environment (ONIE) enabled choice. Users get the option to run the operating system they want on the hardware device they desired. If anything happens to the hardware, it could easily be swapped. Once swapped, ONIE discovered the operating system of choice to install on the hardware. OPen Networking offered flexibility and choice to UTSA to select their choice of hardware and software.
The disaggregation goes one step further with Cumulus Linux: As a native linux network OS, it supports both networking and a range of widely available server packages, mostly open source. Switching from the world of various network operating systems, the team rapidly felt comfortable with Linux, a familiar tool. UTSA was accustomed to teaching every administrator JunOS or IOS and how to do vendor-specific network operation tasks. As a result, the training was esoteric and specific to these vendor implementations. With Cumulus Linux, the switch is just another server with many accelerated ports; the administrator only needs to understand Linux. Since UTSA computer science graduates are skilled in UNIX/Linux, adoption was easy once they complete the program.
Before Cumulus Linux, operations within UTSA were manual. It could take an hour to half a day to get a new switch back up and running. The process was tedious and completely manual with opportunities for errors. With legacy switches, a failure would be associated with an SNMP trap. In the new model, with Cumulus Linux, time to replace hardware is constant and the rest is an automated, streamlined process that takes few minutes to fully operationalize the switch. The correct image is automatically installed with ONIE; then Cumulus Linux’s zero touch provisioning and configuration management tools take over. Since UTSA leveraged Cumulus Linux as their network operating system, they were also able to leverage their own preferred automation application, Ansible, for configuration management throughout the entire network stack.
By leveraging Open Networking technologies with Cumulus Linux, UTSA achieved simplified their stack and got flexibility that resulted in reducing both CapEx and OpEx. Benefits of deploying Open Networking and Cumulus Linux included:
Freedom of choice at every layer – choice of hardware, choice of operating system, choice of applications (CAPEX/OPEX reduction)
High capacity fabric at lower cost (CAPEX reduction)
Budget flexibility with a software subscription and efficient reuse of all assets: hardware, software and tools (CAPEX/OPEX reduction)
Leverage Linux know-how with Linux-savvy staff and students (OPEX reduction)
Consistent data center operations with standard Linux tools (OPEX reduction)
Fast and reliable automated environments with reduced risks, and 4-8x operational improvement (OPEX reduction)
Networking education acceleration