Rocel grew up on the island of Maui and graduated from H.P. Baldwin HS in 2008. He moved to O‘ahu and is currently attending Honolulu CC, majoring in Computing, Electronics & Networking Technologies (CENT). He will be transferring to UH West O‘ahu to complete his bachelor’s degree. Rocel is studying for Cisco CCNA, CompTIA A+, and Security+ certifications, and he plans to pursue a career in networking and systems administration. In his free time, he enjoys hanging out with friends and playing the drums.
Home Island: Maui
High School: H.P. Baldwin High School
Institute when accepted: Honolulu Community College
Automatic Backup ISP Failover Solutions for
the Subaru Telescope Network
Project Site: Subaru Telescope
Mentor: Kiaina Schubert
Internet connectivity failures cause major disruptions at Subaru Telescope — they lead to loss of productivity and large inconveniences for staff workers and researchers within the Hilo base facility. Currently, a manual cut-over method using a secondary Internet Service Provider (ISP) is applied in the event of service loss. Subaru’s primary Internet service, a 1-Gb/s line through Institute for Astronomy at UH, can only be monitored from the Subaru base to the next hop; link states further upstream are unknown. This project involved researching and designing methods that would detect anomalies in upstream links and automatically re-route Internet traffic through a secondary ISP, a 10-Mb/s line through Hawaiian Telcom. The key factor that came into play through this type of networking design change is how well the system will work within a set of known and unknown rules. These knowns and unknowns included the current Subaru Telescope network (STN), and the third parties UH and Hawaiian Telcom. Ultimately, these factors decide a given plan’s feasibility. Our first solution was to implement a dynamic routing protocol that would build and maintain routes automatically as upstream availability changes. Routers along the UH upstream path participate in dynamic routing and relay information of remote paths through one another. This is the preferred, industry-standard solution for this type of wide-area-network “failover” and was proven to work through our testing and simulation environments. The second solution involved use of a feature on the STN firewall called Path Monitoring through Policy-Based Forwarding. The firewall would be able to monitor a given IP address on the Internet through ICMP or “ping” messages, and take down or bring up a link as the monitor sends and receives replies of these messages. This setup required extensive testing as to how well it integrated into the current network. We were fortunate enough to be able to implement and test this solution into the live STN network during a scheduled UH Internet outage. Automatic cut-over was successfully established through this method, although full integration into STN needed to be fine-tuned.