Space Microdatacenter and Communication Co-design

One way to mitigate an ISL-bottleneck in context of SµDCs is to modify the network topology to increase the amount of data onboarded onto the SµDCs. Figure 12a shows how, by adding more receivers to a SµDC, the cluster topology can be changed from a ring, or ‘2-list’, to a ‘4-list’, or, more arbitrarily, a ‘\(k\)-list’ for even \(k\). While this may not help RF communication-based constellations due to limited available bandwidth, tremendous amounts of bandwidth is available in the optical frequencies, allowing linear growth in incoming data rate with the number of optical receivers [139]. Thus, for optical ISLs, \(k\)-lists for \(k > 2\) can be used to increase the SµDC’s incoming data rate at the cost of additional optical receivers on the SµDC and additional transmit power.

As \(k\) increases, the link distance between relay satellites grows. Optical ISL transmit power grows quadratically with distance [97], meaning a 4-list’s ISLs consume \(4\\times\) the power of a 2-list (while also transmitting \(2\\times\) the data). Also, this distance can eventually grow to such an extent that the ISL must aim through significant amounts of atmosphere. This results in atmospheric turbulence induced fading of the optical signal [161], degrading the channel capacity. If the distance is large enough, then the Earth’s landmass will directly block the signal. The specific value of \(k\) for which distance becomes a concern is dependent on the constellation’s formation: for evenly distributed — ‘orbit spaced’ — formations, maximum \(k\) is smaller than for tightly packed formations in which satellites are relatively close to one another.

Alternatively, SµDCs can be split Figure 12b — increasing the number of clusters in a ring-topology without increasing the compute power of the SµDCs in aggregate. By using smaller split SµDCs, costs associated with higher cluster counts (e.g., launch cost, booster fuel requirements, etc.) are mitigated. SµDC splitting is effectively a form of disaggregation, and thus can lead to increased total launch costs and constellation management costs. However, SµDC splitting is effective for all constellation formations, including orbit-spaced constellations which may see limited benefit from \(k\)-list topologies.

SµDC splitting and \(k\)-list topologies can be used in conjunction. Their benefits are orthogonal, and the increase in aggregate data rate into SµDCs scales multi-linearly with number of clusters (from splitting), and number of links into each SµDC (from \(k\)-lists). That is, the number of EO satellites supported by a \(k\)-list topology cluster is \(\\frac{k}{2}\) times those shown in Table 8, while SµDC splitting multiplies the number of clusters. Figure 13 shows that \(k\)-lists combined with SµDC splitting leads to significantly increased ISL communication capacity (the rate at which data can be transmitted from an EO satellite to an SµDC) in a frame-spaced constellation.

缓解 ISL 瓶颈的方法之一是修改网络拓扑结构，以增加 SμDC 接收数据的总量:

协同设计策略一： \(k\)-list 网络拓扑

• 方法描述： 通过在 S\(\mu\)DC 上增加更多的接收器，将集群拓扑结构从环形（或“2-list”）更改为“4-list”，或更通用的“ \(k\)-list”（ \(k\) 为偶数）
• 适用性： 这种方法主要适用于光学 ISL，因为光学频率具有巨大的带宽，允许传入数据速率随着光学接收器数量的增加而线性增长
• 权衡： 增加 \(k\) 可以提高 S\(\mu\)DC 的传入数据速率，但代价是 S\(\mu\)DC 上需要额外的光学接收器和额外的发射功率
• 限制因素： 随着 \(k\) 的增加，中继卫星之间的链路距离也随之增长，这可能导致光学 ISL 发射功率呈二次方增长（例如，“4-list”的 ISL 功耗是“2-list”的 4 倍），并且可能因距离过大导致信号穿过大气层而引发湍流衰减，甚至被地球陆地遮挡

协同设计策略二：S\(\mu\)DC 拆分 (S\(\mu\)DC Splitting)

• 方法描述： 拆分 S\(\mu\)DC，即在环形拓扑中增加集群数量，同时不增加 S\(\mu\)DC 的总聚合计算能力
• 优势： 使用更小的拆分 S\(\mu\)DC 可以减轻与更高集群数量相关的成本（例如，发射成本和助推器燃料需求）
• 适用性： S\(\mu\)DC 拆分对所有星座编队都有效，包括“轨道间隔”（orbit-spaced）的星座，这些星座可能从 \(k\)-list 拓扑中获益有限
• 权衡： S\(\mu\)DC 拆分是一种解聚合形式，可能会增加总发射成本和星座管理成本
• 组合使用： S\(\mu\)DC 拆分可以与 \(k\)-list 拓扑结合使用，它们的优势是相互独立的，可以实现总数据速率的多线性增长

拆分是否会失去“集中资源处理EO数据”的优势

一个很自然的疑惑是: 我好不容易把EO分散的资源/数据聚合在一起, 你又给分了出去, 啥意思?

其实上述疑惑的原因是混淆了 "算力分配" 与 "网络接入点"

这里的 "拆分" 是为了保持总的计算资源不变的情况下, 优化对这些资源的访问（即数据传输）

SμDC 拆分通过增加接收数据的接入点（即 SμDC 的数量）来缓解通信限制:

例如，与其部署一个 8 kW 的 SμDC，不如部署两个 4 kW 的 SμDC。总计算能力仍是 8 kW，但现在整个星座 拥有双倍的 ISL 接入端口， 从而使更多的 EO 观测卫星能够同时将数据卸载到 S\muDC 上进行处理

协同设计策略三：将 S\(\mu\)DC 转移到地球静止轨道 (GEO)

• 方法描述： 将 S\(\mu\)DC 放置在 GEO 轨道，以减轻 ISL 瓶颈
• GEO 缓解 ISL 瓶颈的机制： GEO 上的 S\(\mu\)DC 采用动态星形集群拓扑，利用 3 个 S\(\mu\)DC 彼此间隔 \(120^\circ\) 放置，确保每个 LEO EO 卫星在任何时间都能与至少一个 S\(\mu\)DC 保持视线 (LOS)
• 优势： 这种方法可以缓解 ISL 容量问题，并允许使用非常大的 S\(\mu\)DC
• 权衡/挑战： GEO 定位会带来更高的发射成本，需要更强的辐射加固，因为 GEO 轨道位于地球外范艾伦带，辐射能量更高

结果显示:

很显然, \(k\)-list 拓扑与 S\(\mu\)DC 拆分 这两个策略是正交的。因此理论上如果将二者相结合，应该可以更显著提高帧间隔星座（frame-spaced constellation）中 ISL 的通信容量（EO 卫星向 S\(\mu\)DC 传输数据的速率）!

实验见image13:

实验证明, 的确, 这些协同设计策略能够缓解通信瓶颈, 并实现 S\(\mu\)DC 的有效利用