In-Orbit Processing or Not? Sunlight-Aware Task Scheduling for Energy-Efficient Space Edge Computing Networks

With the rapid evolution of space-borne capabilities, space edge computing (SEC) is becoming a new computation paradigm for future integrated space and terrestrial networks. Satellite edges adopt advanced on-board hardware, which not only enables new opportunities to perform complex intelligent tasks in orbit, but also involves new challenges due to the additional energy consumption in power-constrained space environment.In this paper, we present PHOENIX, an energy-efficient task scheduling framework for emerging SEC networks. PHOENIX exploits a key insight that in the SEC network, there always exist a number of sunlit edges which are illuminated during the entire orbital period and have sufficient energy supplement from the sun. PHOENIX accomplishes energy-efficient in-orbit computing by judiciously offloading space tasks to “sunlight-sufficient” edges or to the ground. Specifically, PHOENIX first formulates the SEC battery energy optimizing (SBEO) problem which aims at minimizing the average battery energy consumption while satisfying various task completion constraints. Then PHOENIX incorporates a sunlight-aware scheduling mechanism to solve the SBEO problem and schedule SEC tasks efficiently. Finally, we implement a PHOENIX prototype and build an SEC testbed. Extensive data-driven evaluations demonstrate that as compared to other state-of-the-art solutions, PHOENIX can effectively reduce up to 54.8% SEC battery energy consumption and prolong battery lifetime to 2.9× while still completing tasks on time.

随着天基能力的快速发展，空间边缘计算（Space Edge Computing, SEC）正成为未来天地一体化网络的一种新兴计算范式。卫星边缘节点采用先进的星上硬件，这不仅为在轨执行复杂智能任务带来了新机遇，也因在功率受限的空间环境中增加了额外能耗而引入了新挑战。

本文提出了一种面向新兴SEC网络的高能效任务调度框架 PHOENIX。PHOENIX 利用了一个关键洞察：在SEC网络中，总存在一些在整个轨道周期内都受到光照、并能从太阳获得充足能量补充的“光照边缘”节点。PHOENIX 通过将空间任务审慎地卸载至“光照充足”的边缘节点或地面站，来实现高能效的在轨计算。

具体而言，PHOENIX 首先构建了空间边缘计算电池能量优化（SBEO）问题，其目标是在满足各种任务完成约束的同时，最小化网络的平均电池能耗。接着，PHOENIX 融合了一种光照感知调度机制来求解SBEO问题，并高效地调度SEC任务。最后，我们实现了一个 PHOENIX 原型系统并搭建了一个SEC测试平台。广泛的数据驱动评估表明，与其他先进解决方案相比，PHOENIX 能有效降低高达 54.8% 的SEC电池能耗，并将电池寿命延长至 2.9 倍，同时仍能确保任务的按时完成。

Introduction

With the rapid evolution in the aerospace industry, emerging low earth orbit (LEO) satellite mega-constellations not only extend the network boundary of today’s terrestrial Internet, but also spawn an innovative computation paradigm: “space edge computing (SEC)”. Based on advanced on-board hardware, emerging SEC technologies combine the capabilities of satellite communication and edge computing to provide edge-like services right at the satellite [1]–[3], and further enable a series of intelligent space applications such as smart remote sensing [4], autonomous debris detection and avoidance [5], and Internet of space things [6], etc.

While SEC has broad application prospects, it also involves new technical challenges in the energy-constrained outer space environment. On the one hand, fully realizing the promising capability of SEC requires extra advanced on-board hardware to support complex space missions, e.g., exploiting satellite GPUs [7], [8] for data inference and deploying high-speed intersatellite communication links (ISL) for cross-edge collaborative processing [9]. On the other hand, these additional payloads inevitably involve more energy consumption on satellite edges. Since the energy usage can significantly affect the execution of SEC tasks as well as the battery lifetime (as we will introduce later in §II), accomplishing energy-efficient space task execution is undoubtedly a crucial problem for futuristic SEC networks.

随着航空航天工业的飞速发展，新兴的低地球轨道（LEO）巨型卫星星座不仅扩展了当今地面互联网的网络边界，也催生了一种创新的计算范式：“空间边缘计算” (Space Edge Computing, SEC)。基于先进的星上硬件，新兴的SEC技术将卫星通信与边缘计算的能力相结合，直接在卫星上提供类边缘服务 [1]–[3]，并进一步赋能了一系列智能空间应用，例如智能遥感 [4]、自主碎片探测与规避 [5] 以及空间物联网 [6] 等。

尽管SEC具有广阔的应用前景，但在能源受限的外太空环境中，它也带来了新的技术挑战。一方面，要完全实现SEC的巨大潜力，需要额外的先进星上硬件来支持复杂的空间任务，例如，利用卫星GPU [7], [8] 进行数据推理，以及部署高速星间通信链路（ISL）以支持跨边缘协同处理 [9]。另一方面，这些额外的有效载荷不可避免地会给卫星边缘节点带来更多的能源消耗。鉴于能源使用会显著影响SEC任务的执行以及电池的寿命（我们将在第二节中详细介绍），实现高能效的空间任务执行无疑是未来SEC网络的一个关键问题。

Task offloading, which has been well-studied by the conventional mobile computing community over the past decade, is an effective approach for optimizing energy consumption on power-constrained devices [10], [11]. The core idea behind traditional task offloading is to transfer the energy-intensive tasks from the power-constrained devices to a high-performance server with sufficient power supplement (e.g., a cloud), and receive the results after remote execution. Energy can be saved if the network transmission consumes less energy than local task execution. In an SEC scenario, a straightforward method to apply task offloading for energy-efficiency is to transfer space tasks to a nearby ground station which typically has sufficient computation capability and power supplement. However, due to huge amount of SEC data [12], naively offloading all tasks to the ground can easily overwhelm the satellite downlink [8] and involve high latency which is unacceptable for time-sensitive SEC tasks like satellite-based wildfire monitoring and rescue.

在过去的十年里，任务卸载作为一种在功率受限设备上优化能耗的有效方法，已在传统的移动计算领域得到了充分研究 [10], [11]。

传统任务卸载的核心思想，是将能源密集型任务从功率受限的设备转移到拥有充足电力供应的高性能服务器（例如云端），在远程执行后接收结果。如果网络传输消耗的能量少于本地执行任务所消耗的能量，便可实现节能。

在SEC场景中， 一个直接应用任务卸载以实现节能的方法，是将空间任务传输到通常具备充足计算能力和电力供应的地面站 。然而，由于SEC数据量巨大 [12]，将所有任务简单地卸载到地面很容易使卫星下行链路不堪重负 [8]，并会引入高延迟，这对于像星基野火监测和救援这类时间敏感型SEC任务是不可接受的。

To address the limitation of existing offloading approaches, this paper presents PHOENIX, an energy-efficient and deadline-aware task scheduling framework for emerging SEC networks. The design of PHOENIX stands upon a series of important insights obtained from today’s LEO satellite constellations. First, sub-systems in a satellite edge are powered directly by solar panels with sufficient power supplement when the satellite is illuminated by sunlight, and are powered by a rechargeable battery when the satellite enters the earth’s shadow. Thus, the key to energy optimization is to reduce the energy consumption of the satellite battery. Second, as satellites move, we observe that there dynamically exist a number of orbit planes where satellites in these orbits are exposed to sunlight with near-100% sunlit ratio. Carrying out computation tasks on these “sunlit edges” does not consume battery power. Third, an SEC task is typically associated with a time-to-completion requirement, especially for time-sensitive applications. Taken them together, PHOENIX accomplishes energy-efficiency by dynamically and judiciously offloading space tasks to computational nodes with sufficient power supplement subjecting to various task deadlines. To this end, PHOENIX makes scheduling decisions based on the following options: (i) processing the data locally on a satellite edge; (ii) offloading SEC tasks to other proper “sunlit edges”; or (iii) offloading SEC tasks to ground stations.

为了解决现有卸载方法的局限性，本文提出了 PHOENIX，一个面向新兴SEC网络的高能效且具备截止时间感知的任务调度框架。PHOENIX的设计基于一系列从当今LEO卫星星座中获得的重要洞察。

首先，当卫星受太阳光照时，其子系统由太阳能电池板直接供电，电力充足；而当卫星进入地影区时，则由可充电电池供电。因此，能源优化的关键在于减少卫星电池的能量消耗。

其次，我们观察到，随着卫星的移动，动态地存在着一些轨道平面，其中的卫星几乎以100%的光照比率暴露在阳光下。在这些 “光照边缘” 上执行计算任务不消耗电池电量。

第三，一个SEC任务通常关联一个完成时间要求，对于时间敏感型应用尤其如此。综合这些因素，PHOENIX 通过在满足各种任务截止时间的前提下，动态且审慎地将空间任务卸载到电力供应充足的计算节点，从而实现高能效。

为此，PHOENIX 的调度决策基于以下选项：

(i) 在卫星边缘本地处理数据

(ii) 将SEC任务卸载到其他合适的“光照边缘”

(iii) 将SEC任务卸载到地面站

Specifically, PHOENIX calculates the decisions in two steps. First, given the SEC network information and the completion time requirements of various tasks, PHOENIX formulates the SEC Battery Energy Optimization (SBEO) problem which targets at minimizing the maximum depth-of-discharge (DoD) of all satellite edge batteries, while satisfying various network, computation and application-level constraints. DoD is an important metric that characterizes the energy usage of a battery and can affect the lifetime of the rechargeable battery as well as the satellite itself. However, efficiently solving the SBEO problem is non-trivial since we prove its NP-hardness and multiple concurrent space tasks can compete for the dynamic computation and network resources in the SEC network.

Second, PHOENIX incorporates a sunlight-aware dynamic SEC task scheduling mechanism which decomposes the original SBEO problem and adopts a series of heuristic algorithms to calculate appropriate scheduling decisions efficiently. Specifically, to reduce the problem complexity, PHOENIX jointly combines coarse-grained orbit-level and find-grained per-satellite task scheduling to calculate near-optimal task assignments.

具体来说，PHOENIX 分两步计算决策。

首先，给定SEC网络信息和各种任务的完成时间要求，PHOENIX 构建了空间边缘计算电池能量优化（SBEO）问题。 该问题旨在最小化所有卫星边缘电池的最大放电深度（Depth-of-Discharge, DoD），同时满足网络、计算和应用层面的各种约束。 DoD 是一个表征电池能量使用情况的重要指标，它会影响可充电电池乃至卫星本身的寿命。然而，高效求解SBEO问题并非易事，因为我们证明了其NP困难性，并且多个并发的空间任务会在SEC网络中竞争动态的计算和网络资源。

其次，PHOENIX 融合了一种光照感知的动态SEC任务调度机制 ，该机制分解了原始的SBEO问题，并采用一系列启发式算法来高效地计算出合适的调度决策。具体而言，为降低问题复杂度，PHOENIX 联合运用了粗粒度的轨道级调度和细粒度的单星级任务调度，以计算出近优的任务分配方案。

We build a data-driven hardware-in-the-loop SEC testbed and implement a PHOENIX prototype upon it. Our testbed integrates large-scale SEC network simulation, and low-power computational hardware that has been verified in real space environments. Extensive evaluations demonstrate that PHOENIX can outperform other state-of-the-art SEC approaches in terms of energy consumption, battery lifetime, task deadline satisfaction etc., under various experiment configurations.

Contributions of this paper can be summarized as follows: (i) we formulate the SEC battery energy optimization (SBEO) problem and expose the technical challenges of solving it efficiently and effectively; (ii) we propose PHOENIX, a novel sunlight-aware energy-efficient task scheduling framework for optimizing satellite battery usage and extending the lifetime of SEC networks; (iii) we implement a PHOENIX prototype and conduct extensive data-driven, hardware-in-the-loop experiments to demonstrate the effectiveness of PHOENIX.

我们构建了一个数据驱动的硬件在环SEC测试平台，并在其上实现了一个 PHOENIX 原型。我们的测试平台集成了大规模SEC网络仿真和已在真实空间环境中得到验证的低功耗计算硬件。广泛的评估表明，在各种实验配置下，PHOENIX 在能耗、电池寿命、任务截止时间满足率等方面均优于其他先进的SEC方法。

本文的贡献可总结如下：(i) 我们构建了SEC电池能量优化（SBEO）问题，并揭示了高效、有效求解该问题的技术挑战；(ii) 我们提出了 PHOENIX，一个新颖的、光照感知的、旨在优化卫星电池使用和延长SEC网络寿命的高能效任务调度框架；(iii) 我们实现了一个 PHOENIX 原型，并进行了广泛的数据驱动、硬件在环实验，以验证 PHOENIX 的有效性。