Fall has arrived, and the SCinet team has begun its takeover of the Colorado Convention Center to start constructing the SCinet network for the SC23 conference. Volunteers of SCinet are among the earliest to arrive in Denver for the event called “Staging” week. At Staging, SCinet volunteers begin installing equipment borrowed through the SCinet contributors program under a tight deadline to enable basic network connectivity that will support the SC conference. However, for SC23, SCinet will not simply construct an ordinary network in Denver. This year’s version will provide an astonishing 6.71 Tbps (terabits per second) of network bandwidth to the convention center. This speed is sufficient to download the Library of Congress’ complete 21 petabyte (equivalent to 21000 terabytes) catalog in just over seven hours¹.

Achieving this goal necessitates forging robust partnerships with a multitude of external stakeholders, encompassing Research and Education (R&E) networks, commercial network providers, and researchers spanning the globe.

Wide Area Networks Need Wide Social Networks

Carlos Rojas-Torres

SCinet WAN Team Member

Leadership & Teamwork

The existence of the SCinet wide area network, or WAN, and collaborative efforts of SCinet contributors are absolutely essential for the realization of the conference network.

“The WAN team requires a strong and committed leadership team. The preparation and coordination of contributors, volunteers, and the SCinet committee is a multi-year endeavor,” explained Carlos Rojas-Torres, SCinet WAN team member from the University Corporation for Atmospheric Research (UCAR) and Front Range GigaPoP. “The team is composed of some of the most talented people in the industry, and it’s practically impossible to walk away from there without learning something new every single day.”

Relationships Matter

The WAN team’s ability to stay ahead of the numerous challenges they encounter each year can be attributed to the cultivation of these enduring multi-year partnerships and relationships. 

“The culture and relationships are what makes the WAN team successful,” said Chris Skaar, SCinet WAN team co-lead from University of Illinois Urbana-Champaign. “The leads for the WAN team have, over the years, fostered a culture of hard work, supporting one another, learning from each other, and developing long-lasting connections. We also try to help our contributors be as involved as possible and ensure that our team really learns about the products that they are donated.”

Chris Skaar

SCinet WAN Team Co-Chair

Big Tbps Requires Big Creativity

SCinet’s WAN design is intricately linked to the specific destinations researchers wish to connect to and the use cases that drive the presentation of their research at SC.

Cody Rotermund

SCinet WAN Team Deputy Chair

Putting the Puzzle Together

“Leading up to Staging week, it’s almost like we’re putting together a puzzle,” explained Cody Rotermund, SCinet WAN Team Deputy Chair from the Energy Sciences Network (ESnet). “We get some of the pieces through the end researchers or through the SCinet Network Research Exhibition (NRE) process. However, many of the puzzle pieces you need to form a complete picture aren’t really known until later in the year.  Essentially, to start putting an initial picture together, we do quite a bit of forecasting based in part on what we did in previous years. With that in mind, we’ll take a look at what fiber currently exists, and what we need to work with our contributors, to build out and go from there.”

A one-size-fits-all solution simply doesn’t exist. Each scenario demands a distinct technical approach and the judicious pairing of technology. The SCinet WAN leadership team invests a significant amount of time in assessing emerging technologies from WAN equipment contributors and aligning them with researcher use cases and connectivity requirements to meet the demanded network bandwidth.

“Once we have a handle on the initial fiber plant design, we then work with our hardware contributors and attempt to line up the best technical solution for the use case,” Rotermund added. “In some cases, a network carrier contributor might be able to provide end-to-end connectivity over their Dense Wavelength Division Multiplexing [DWDM] system. In other scenarios, we collaborate with our hardware contributors to connect into the carrier’s DWDM system using our own transponder modems.”

The World Is Our Stage: SCinet Global WAN Map

Finding the best technological solution often requires outside-the-box thinking. For example, Skaar noted that SCinet uses multiple contributors’ equipment to run “alien waves” across existing Line systems. An alien wave is when a third party uses their own transponder modems to transmit data over a carrier’s DWDM network. 

“This involves using Vendor A’s transponders across Vendor B’s line system and requires a lot of coordinated effort as many times these circuits must be manually configured on both vendors’ equipment,” Skaar described. “This year, we are utilizing ZR optics, which use tunable, optical frequencies in the same spectrum as DWDM to transmit data over shorter distances, under 10 km. This allows us to multiplex or ‘mux’ these wavelengths into our DWDM system from our local Point of Presence [POP] sites to the convention center, reducing the need for extra fiber.”

Testing, Testing, and More Testing

The performance tolerances for operating connections at speeds like 100 Gigabit Ethernet (GE), 400 GE, and beyond are exceptionally tight. The SCinet WAN team dedicates significant effort to collaborate with various network carriers and conducts thorough performance validation and testing throughout the Staging phase and in the days leading up to the SC23 conference. Time is of the essence during this critical period, and the WAN team’s innovative use of technology for remote testing has yielded significant benefits and results.

By employing a micro-electromechanical systems (MEMS) optical switch, the WAN team has the ability to implement dynamic fiber topology adjustments within the convention center, eliminating the need for manual fiber swapping in the field. This innovation enables them to programmatically establish connections between WAN circuits and test equipment. Once the connections have been validated, they can be seamlessly redirected to the appropriate router or switch, streamlining network operations and enhancing efficiency.

“This saves us valuable time of manually repatching fiber jumpers, as well as providing extra optical measurements, which is crucial in troubleshooting,” Skaar added.

Share Your World with Us

Are you seeking hands-on, experiential training in the construction and operation of a wide area network? Join SCinet and learn from some of the most brilliant minds in the field! The SCinet WAN Team includes optical engineers from diverse backgrounds, including R&E, government, industry, and HPC. As Rojas-Torres strongly encouraged, SCinet offers a valuable opportunity for knowledge and skill acquisition in the world of wide area networking. 

“Don’t hesitate: go for it!” he implored. “The friendships, professional networking, and exposure to new technologies you’ll experience while being part of the WAN team will stay with you forever.”

Learn more about how to participate in SCinet for SC24:

Participate in SCinet

More About WANs

What Is a WAN and What Makes It Work?

A Wide Area Network (WAN) encompasses a set of technologies and service provider networks designed to interconnect various Local Area Networks (LANs) over extensive distances. A LAN enables networked devices, e.g., computers, to communicate and share resources within a confined geographical area or under a single administrative entity, such as a school, data center, or business. If you think of the LAN as an isolated island, the WAN acts as the bridge that facilitates the exchange of data and traffic between these distinct islands.

To illustrate with a practical example, consider a scenario where a university located in New York operates its own LAN and seeks to establish a connection with a remote resource hosted in the Cloud, which resides within another LAN situated in Los Angeles. In this instance, the WAN plays a crucial role by facilitating long-distance communication between these two organizations. This communication is achieved using fiber-optic cables as the transmission medium and specialized Dense Wavelength Division Multiplexing (DWDM) equipment, which enables efficient multiplexing and transmission of data over various wavelengths, ensuring reliable connectivity between the two LANs separated by a significant geographical distance.

DWDM and Optical Networking

A device called a transponder modem takes one or multiple network Ethernet connections from a LAN and encodes each connection into a specific frequency range within the visible light spectrum, also referred to as a channel or wavelength. The DWDM equipment functions like a prism, taking these individual wavelengths generated by the transponder and amalgamating them into a composite signal for transmission over a single fiber. The bandwidth capacity of each of these wavelengths varies based on the optical properties of the fiber, such as distance, loss, and reflection. It is common to encounter DWDM equipment that can support up to 80 distinct wavelengths over a single pair of fibers. This capability allows the WAN provider network to pack a substantial amount of network bandwidth over a single fiber efficiently.

After the DWDM equipment transmits the composite signal onto the fiber, signal degradation becomes more pronounced the farther it has to travel. To counter this, amplification equipment is strategically placed along the fiber route to rejuvenate the signal. Eventually, it reaches the DWDM equipment at the opposite end, which then separates the composite signal into its original wavelengths. These wavelengths are received by the transponder modem, which converts them back into individual Ethernet connections for the LAN.

References

¹ https://www.loc.gov/programs/digital-collections-management/about-this-program/frequently-asked-questions/

SCinet Diamond Contributor

SCinet Routing & Cisco NSO

Over the years, the Routing Team at SCinet has piloted different automation solutions to enhance their collective ability to scale and accomplish more in less time. For SC23, the Routing Team has partnered with Cisco to use the NSO orchestrator application. This app helps standardize configurations, simplify the deployment of new devices, and minimize human error.

Corey Eichelberger

SCinet Routing Team Co-Chair

“By leveraging the programmatic interfaces of modern network hardware and automation technologies, the Routing Team can realize efficiencies in how we build and manage the SCinet network,” explained Corey Eichelberger, SCinet Routing Team Co-Chair from the Texas Advanced Computing Center (TACC). “This allows the Routing Team to free up valuable time and resources for our volunteers to work on the more experimental projects that SCinet is known for.”

Kalina Dunn

SCinet Routing Deputy Chair

There is also a skills gap across the industry for network engineers who understand how to write software. Routing Team Deputy Chair Kalina Dunn, from GlobalNOC at Indiana University, explained: “Using technologies like Cisco NSO, we are also looking to provide our volunteers with NetOps experience and training, which are typically hard to come by in the networking industry. By letting them learn and experience how these technologies work, they can bring their experience back to help solve similar challenges their home institutions are facing.”

Cisco NSO Architecture

Cisco NSO is a multi-vendor orchestration platform that provides a single programmatic API (application programming interface) to the network [1].

“Cisco NSO can provide turnkey solutions to automate almost any hardware with an IP address and login, providing the HPC community a multi-vendor tool that will manage different vendors’ nuances of managing devices via the Cisco NED (Network Element Driver),” said Patrick Finnegan, a Network & Security Architect at Cisco and SCinet Routing Team member.

Combining various technologies to address the permutations of vendor equipment is a longstanding challenge for SCinet. Routing Team member Bartosz Drogosiewicz, from ICM University of Warsaw, offered his take on the challenges posed by the multi-vendor issue.

“The industry is at this weird but perfectly understandable point where every vendor tries to have its own automation solution that is only compatible with their own products,” Drogosiewicz said. “If you’re working in a multi-vendor environment, it becomes a challenge to orchestrate all the different devices and services. There’s a lot of little things and quirks that pile up while discovering nooks and crannies of old and new devices. That becomes the hard part: the little things that need to be documented and passed on.”

Bartosz Drogosiewicz

SCinet Routing Team Member

Patrick Finnegan

Cisco Network & Security Architect

An Automation Journey

The SCinet Routing Team has long been a proponent of technologies to simplify management and provide more intelligent networks. While scripting against network elements is a concept that has been applied for decades, the Routing Team began adopting more robust techniques, such as Software Defined Networking (SDN) and OpenFlow, as far back as SC15.

OpenFlow, OpenDaylight, and Ryu (SC15–SC17)

OpenFlow was a transformative step toward adding more intelligence and programmability to the network’s dataplane. This provides more granular control over traffic flows in the network. OpenFlow was first introduced by the Open Networking Foundation in February 2011 [2].

The SCinet Routing Team leveraged the multi-vendor aspect of the SCinet network to test the robustness of equipment support for the OpenFlow protocol. At SC15, the Routing Team deployed the OpenDaylight OpenFlow controller. Network service that once required individual engineers logging into many devices, OpenDaylight could deploy between the core network and the booth at the push of a button [3]. This experiment was continued as part of SC17 using the Ryu SDN controller [4].

Ansible/NAPALM and Faucet (SC18)

Despite OpenFlow being adopted in Research and Education networks, industry support for OpenFlow among hardware vendors declined [5]. The Routing Team began to investigate alternative open-source projects being leveraged for server automation and orchestration. 

During this period, the SCinet network was split between two different automation/orchestration frameworks. The first half of the network services were deployed using a combination of Ansible and NAPALM. The NAPALM library provided a multi-vendor unified API for programmatic network equipment access. The Ansible application provided an orchestration engine and the ability to automate workflows [6].

A model of configuration intent was exported from the SCinet Database (called Intranet). The final piece of this solution was the Flask web container application developed by the Routing Team, which acted as a web user interface front end for Ansible/NAPALM and the SCinet Intranet export script called GenConfig.

The other half of network services were deployed using an OpenFlow controller called Faucet. Despite OpenFlow adoption being on the decline, Faucet provided a unique solution as vendor OpenFlow support must be certified as compatible by the Faucet Foundation. This approach helped ensure consistency of OpenFlow support between hardware vendors. 

Either the hardware is compliant with the required OpenFlow features, or it is not. Faucet also separates the SDN controller function from the operation of the network devices. An operator could update the software on the Faucet controller, and the network will continue operating using its current known good configuration [7].

Cisco NSO (SC19–SC23)

Despite previous success, the multi-vendor aspect of SCinet continued to present challenges for managing configurations and introducing new contributors in the network. Groundwork began in SC19 to build a virtual SCinet R&D network using the GNS3 hypervisor, allowing multi-user topology management. This allowed the Routing Team to expand their R&D cycle for SCinet from a few weeks to year-round. Using network design data from previous SCinet networks, the Routing Team leveraged this environment to pilot configuration management and orchestration with NSO.

Find Your Own “Route” with SCinet

In addition to research and education networks, getting hands-on experience in a multi-vendor network and learning from the brightest minds across educational institutions, government agencies, and HPC sites are among the many reasons volunteers are drawn to SCinet. Drogosiewicz offered his own reasoning about why others may want to join.

“SCinet builds something totally unique,” he added. “I call it ‘The Woodstock of IT.’ It’s a three-day festival with the world’s fastest temporary network built and disassembled within a week. Being a part of the Routing Team means that you’re doing all the different aspects of architecture, design, R&D, configuring, and troubleshooting. Everyone participates in every aspect of it, and it’s a unique chance to broaden your knowledge.”

Learn more about how to participate in SCinet for SC24:

Participate in SCinet

References:

[1] https://developer.cisco.com/docs/nso/guides/#!introduction-to-nso/architecture-overview

[2] https://web.archive.org/web/20110326024026/http://www.opennetworkingfoundation.org/?p=7 

[3] https://www.opendaylight.org/

[4] https://ryu-sdn.org/

[5] https://blog.ipspace.net/2022/05/openflow-still-kicking.html

[6] https://github.com/napalm-automation/napalm-ansible

[7] https://faucet.nz/