Network Switch Fast OS Upgrade Without Reboot
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network switch upgrades and reloads result in extended downtime, causing network disruptions due to the inability to route packets for 5 to 10 minutes, which affects connectivity and can lead to virtual machine crashes in large networks.
Innovation Solution
Implementing a fast upgrade/reload technique that allows a network device to download and install a new operating system without rebooting, by disabling DMA transfers and NPU access, delaying NPU reset until the new OS starts, and programming the NPU with new forwarding information to maintain packet forwarding during the process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a network switch performs a traditional upgrade or reload, then the operating system is updated or the device is rebooted, but the network switch becomes unable to route packets for 5 to 10 minutes causing network disruption
Solution Approach 1:
The patent applies preliminary action by downloading the new operating system image while the network switch is still running the existing OS. This allows the upgrade process to begin in advance, with the new image being stored in non-volatile memory before the actual switchover occurs, thereby reducing the time the switch is non-operational.
Solution Approach 2:
The patent maintains continuity of useful action by keeping the network switch operational throughout the upgrade process. The existing OS continues to route packets while the new OS is prepared, and the transition is orchestrated to minimize interruption. The switch only becomes non-operational for a brief moment during the actual OS switchover, not for the entire upgrade duration.
2Adaptability or versatility
If a network switch performs a traditional reload, then the device is rebooted to apply configuration changes, but connectivity is lost and virtual machines crash for 5 to 10 minutes
Solution Approach 1:
Configuration changes and the new OS image are prepared and validated in advance while the network switch remains operational. This preliminary preparation ensures that when the reload occurs, the configuration is already optimized and ready to be applied, minimizing the disruption to network throughput.
Solution Approach 2:
The patent introduces an intermediary mechanism where the new OS image and configuration are staged in non-volatile memory and validated before the actual reload. This intermediary preparation phase allows the switch to maintain normal operations while the upgrade components are readied, reducing the impact on network productivity.
3Loss of time
If a network switch loses connectivity during upgrade, then existing communication links are torn-down and virtual machines crash, but this causes large-scale network disruption
Solution Approach 1:
The new operating system image is downloaded and validated in advance while the network switch continues to operate normally. This preliminary action ensures that when the switchover occurs, the new OS is ready and verified, minimizing the duration of connectivity loss and preventing cascading network disruptions.
Solution Approach 2:
The patent implements beforehand cushioning by validating the new OS image and preparing the upgrade environment before the actual switchover. This preparation phase acts as a cushion, ensuring that the transition is smooth and minimizing the risk of extended outages or cascading failures in the network.
Data Source
AI summary
A network device includes a control processor and a network processor unit (NPU) to forward network packets between network ports of the device. The processor downloads a new operating system (OS). The processor performs a fast upgrade to the new OS using a fast upgrade/reload technique to isolate the processor from interruptions and, without rebooting the processor, start the new OS in place of an existing OS. The fast upgrade/reload also includes delaying a reset of the NPU so that the NPU continues to route the network packets based on existing forwarding information. The fast upgrade/reload includes resetting the NPU only when the new OS first accesses the NPU, and then programming the reset NPU.