Software Reliability Analysis of NASA Space Flight Software: A Practical Experience
Software Reliability Analysis of NASA Space Flight Software: A Practical Experience
复制标题
NASA太空飞行软件的软件可靠性分析：实践经验
DOI：
--
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Issac Mcginnis
中科院分区：
文献类型：
--
作者：
H. Sukhwani;Javier Alonso;Kishor S. Trivedi;Issac Mcginnis
本文对某航天使命飞行软件进行了可靠性分析。对于我们的分析，我们使用在飞行软件开发过程中收集的缺陷报告。我们发现该软件是在多个版本中开发的，每个版本跨越所有软件生命周期阶段。我们还发现，软件版本是针对四种不同的硬件平台开发和测试的，从现成的或仿真硬件到实际的飞行硬件。对于表现出可靠性增长或衰减的版本，我们拟合软件可靠性增长模型（SRGM），否则我们拟合分布函数。我们发现，大多数版本表现出可靠性增长，对数逻辑（NHPP）和S形（NHPP）作为最适合的SRGMs。对于经历可靠性衰减的版本，我们调查了相同的原因。我们发现，这样的版本是第一个软件版本进行测试的一个新的硬件平台，因此，他们遇到了重大的硬件集成问题。此外，这些版本似乎是在时间压力下开发的，以便更快地在新硬件平台上开始测试。此类版本的可靠性增长较差，因此预测故障率较高。其他问题包括硬件规格的变化和供应商的交货延迟。因此，我们的分析为管理层提供了关键的见解和输入，以改进软件开发过程。由于美国航天局已转向产品线工程的飞行软件开发，未来的空间任务的软件将以类似的方式开发，因此，对这个使命的分析结果可以被认为是未来的飞行软件任务的基线。
In this paper, we present the software reliability analysis of the flight software of a recently launched space mission. For our analysis, we use the defect reports collected during the flight software development. We find that this software was developed in multiple releases, each release spanning across all software life-cycle phases. We also find that the software releases were developed and tested for four different hardware platforms, spanning from off-the-shelf or emulation hardware to actual flight hardware. For releases that exhibit reliability growth or decay, we fit Software Reliability Growth Models (SRGM), otherwise we fit a distribution function. We find that most releases exhibit reliability growth, with Log-Logistic (NHPP) and S-Shaped (NHPP) as the best-fit SRGMs. For the releases that experience reliability decay, we investigate the causes for the same. We find that such releases were the first software releases to be tested on a new hardware platform, and hence they encountered major hardware integration issues. Also such releases seem to have been developed under time pressure in order to start testing on the new hardware platform sooner. Such releases exhibit poor reliability growth, and hence exhibit high predicted failure rate. Other problems include hardware specification changes and delivery delays from vendors. Thus, our analysis provides critical insights and inputs to the management to improve the software development process. As NASA has moved towards a product line engineering for its flight software development, software for future space missions will be developed in a similar manner and hence the analysis results for this mission can be considered as a baseline for future flight software missions.