SpaceWire Network Support Algorithm as Part of Decentralized Plug-and-Play Algorithm St. Petersburg State University of Aerospace Instrumentation Institute for High-Performance Computer and Network Technologies Elizar Fortyshev, Ksenia Khramenkova
Introduction Plug-and-Play algorithms allow automate the process of configuring and administering onboard systems according to standard SpaceWire. One of the tasks in the development of decentralized Plug-and-Play algorithm is the creation of algorithm for network managers support, in which failure one or more network managers will not cause to fall of onboard network. 2/13
The SpaceWire standard SpaceWire standard is an ECSS standard, which aims to use in the management, development and product quality in aerospace projects. SpaceWire supports the integration and testing of the onboard systems through the introduction of appropriate hardware directly in the data-processing equipment. 3/13
Basic methods configuration and administration of SpaceWire routers Manual Automatic 4/13
Automated method configuration and administration (Plug-and-Play) Centralized Plug-and-Play Algorithms Decentralized 5/13
Problem in decentralized algorithm Plug-and-Play Every network manager administrates only its subnet, and in case of manager's failure, its part of network will become uncontrollable, which can lead to failure of the entire network and other negative consequences. 6/13
Variants for solving this problem Enlargement of regions Duplication of managers SpaceWire Network Support Algorithm Combination method 7/13
Enlargement of regions Idea of the method: neighbouring manager must take subnet of the failed manager. In voting for the role of region s coordinator were used basic principles of circular synchronization algorithm in a distributed system, which is based on the use of the ring (physical or logical) without a marker. 8/13
Duplication of managers Idea of the method: each network manager has a duplicate manager that may enters work on subnet instead of main manager. Work reliability two parallel-connected network managers, namely probability that at least one of them will be work, equals a sum of three probabilities of possible favourable outcomes: neither the main manager nor the duplicate manager will break down; main manager fails, but duplicate manager will be working; duplicate manager fails, but the main manager will be working. 9/13
Combination method Idea of the method: combination two methods: duplication of managers and enlargement of regions. Duplicate has a same logical and regionally logical address like at main manager. Main manager and duplicate connected to router in parallel. 10/13
Comparative analysis of these methods Advantages: Limitations: Enlargement of regions No additional material costs in construction of the network Limitations on maximum number of devices that can be serviced by one manager because of its limited memory Duplication of managers Allows increase reliability of the network and avoid conflicts between neighboring managers To construct the network requires twice managers Combination method Combines reliability of 1 st and 2 nd methods and can save network control as long as possible Can not always be applied because imposes strong demands for memory managers and material costs 11/13
Conclusions: Using a decentralized algorithm allow to improve network reliability, its setting and monitor status of devices. On the other hand, it imposes additional requirements on the network and managers. SpaceWire Network Support Algorithm is regarded for solving the problem in case failure of one or more managers. Most reliable method of the algorithm presented in this work is a combination method. However, he imposes serious requirements to memory, managers and other costs. Duplication of managers can increase network reliability and avoid conflicts between neighbouring managers. However, it is needed twice as many managers to build network, that not always possible to implement at onboard network. Enlargement of regions is an alternative to duplication, however, there are restrictions (due to the limited memory of managers) on maximum number of devices that can be serviced by one manager. 12/13
Thank you for attention! 13/13