Rebirth: Starting from Lighting Up the Tech Tree

Chapter 065 Technical Breakthroughs

At 8 p.m., the lights in conference room 402 were on full display.

Seven people sat around the long table: Zuo Cheng, Yu Ying, Zhang Lei, Chen Hao, Liu Wei, Fang Ze, and the new intern, Xiao Lin. Each person had a notebook and draft paper in front of them, and the whiteboard was covered with formulas and flowcharts.

"We have two core issues to solve," Zuo Cheng said, standing in front of the whiteboard with a marker in hand. "First, the latency compensation algorithm for the satellite-to-ground link; second, the concurrent access mechanism for massive numbers of terminals."

Yu Ying added, "Professor Li meant that the solution must be supported by theoretical principles, and cannot be based solely on engineering experience."

"We have the theory," Zuo Cheng said, "but the challenge lies in combining theory with engineering."

Chen Hao adjusted his glasses: "The latency variations in satellite-to-ground links follow a pattern and can be described using a mathematical model. The problem lies in how to update the parameters of this model in real time. There are too many variables: the satellite's position, the ground station's position, atmospheric interference, etc."

"Use Kalman filtering?" Yu Ying asked.

"Kalman filtering is too computationally intensive," Chen Hao shook his head. "The processors on the satellite have limited performance and can't run complex filtering algorithms."

Zuo Cheng stared at the formulas on the whiteboard, his mind racing.

In his previous life, the Sky Dome project used a simplified version of the Kalman filter, but the results were not ideal. In this life, he wants to do better and needs to find a more lightweight solution.

"I've got an idea," Zuo Cheng said. "Instead of real-time filtering, we can use a predictive model. Based on the satellite's orbital parameters and historical latency data, we can train a lightweight neural network to predict the latency value at the next moment."

"A neural network?" Fang Ze frowned. "That requires training data. Where do we get that?"

"Use a lightweight model," Zuo Cheng said. "The training data comes from the first two phases of the Sky Dome project, and Blue Bay Communications should have complete link data. The model's inference will be performed at the ground station, while the satellite will only run the forward propagation, keeping the computational load under control."

Zhang Lei wrote in his notebook: "I will contact Zhou Henian tomorrow."

"The second problem is the massive number of concurrent terminals." Zuo Cheng drew a diagram on the whiteboard. "The goal of the third phase is to support more than 100,000 terminals online simultaneously, but the current architecture can't handle that."

Yu Ying said, "Professor Li's team has done relevant research and developed a hierarchical access architecture that divides terminals into different priorities, with high-priority terminals accessing first and low-priority terminals waiting in a queue."

How exactly are priorities assigned?

"By business type." Yu Ying pulled a paper from her bag. "This is the research result of Professor Li's team, which divides businesses into three categories: emergency communication, general IoT, and data collection. Emergency communication has the highest priority and must guarantee real-time performance; general IoT is next, allowing for a certain degree of delay; data collection has the lowest priority, tolerating greater delays, but the data volume is usually very large."

Zuo Cheng nodded: "That's a good idea. But we also need a dynamic load balancing mechanism to prevent the system from crashing due to too many users in a particular area."

A brief silence fell over the meeting room. Everyone was deep in thought, their fingers tapping unconsciously on the table.

Zuo Cheng closed his eyes and brought up the system interface in his mind.

The current score is 123 points. There is one more technical analysis session remaining, with a validity period of five days.

"Using technical analysis."

[Technical Analysis Launched]

[Objective: Architecture for handling massive concurrent client access]

[Consumption: 1 technical analysis session]

[0 times remaining]

A cool sensation flooded my mind, and my thoughts became exceptionally clear.

Zuo Cheng constructed a model in his mind. Layered access was correct, but not enough. A dynamic scheduling mechanism needed to be added on top of the layering.

Each satellite in each region acts as an independent access node, maintaining a local queue. When the queue length exceeds a threshold, load balancing is automatically triggered, transferring some terminals to satellites in adjacent regions.

This allows for global load balancing and avoids single-point overload.

Zuo Cheng opened his eyes, walked to the whiteboard, and began to draw a new architecture diagram. His pen moved quickly across the whiteboard, drawing boxes and arrows.

"I've come up with an improvement plan," he said. "Based on layered access, we can add dynamic load balancing. Each satellite node maintains a local queue, and when the queue exceeds a threshold, a load balancing is automatically triggered. The load balancing strategy can use consistent hashing to ensure that the same terminal is always assigned to the same satellite, avoiding the overhead caused by frequent switching."

Chen Hao's eyes gradually lit up as he looked at the diagram on the whiteboard.

"This would greatly improve the system's fault tolerance. Even if one satellite malfunctions, the workload can be automatically transferred to other satellites."

"That's right," Zuo Cheng said. "This is the architecture we need."

Chen Hao stared at the whiteboard for a while, then suddenly said, "This solution is a generation more advanced than Huaxin's centralized architecture. They aggregate all terminals into a central node for processing, while ours is distributed, with each satellite node independently handling local traffic, only switching over when the load is unbalanced."

"Yes, that's our advantage," Zuo Cheng said. "Huaxin's solution works fine when the number of terminals is small, but once it exceeds 100,000, the central node becomes a bottleneck. Our distributed architecture naturally supports horizontal scaling."

Yu Ying took notes, occasionally interjecting, "What is the theoretical basis of this architecture? I need to explain it to Professor Li."

"Queuing theory plus distributed consistency," Zuo Cheng said. "Each layer is a queuing system, and layers maintain consistency using a distributed protocol."

"That's enough." Yu Ying nodded. "I can organize it into a document tomorrow."

The meeting finally ended at 2 a.m.

Zuo Cheng took photos of the contents of the whiteboard for archiving, Yu Ying compiled the meeting minutes, and the others left one after another.

"Thanks to you today," Zuo Cheng said to Ying. "Without your academic background, I simply couldn't explain many theoretical issues clearly."

Yu Ying smiled and said, "I've learned a lot too. You have a strong engineering mindset; you can turn complex theories into feasible solutions."

"Let's learn from each other," Zuo Cheng said. "Let's go, I'll take you home."

"No need, I'll take a taxi myself. You have to get up early tomorrow."

"I send you."

Yu Ying looked at him and did not refuse again.

The city of Hangzhou was quiet in the early morning, the streetlights casting a dim yellow glow on the road. Zuo Cheng drove slowly, and neither of them spoke, but the atmosphere was not awkward.

"Brother," Yu Ying suddenly spoke.

"Um?"

"Thank you for letting me join 402."

"I told you, this isn't charity." Zuo Cheng looked ahead at the road. "402 needs you, and I need you too."

Yu Ying's lips curled up slightly as she looked at the night view outside the window.

"I know."

The car stopped downstairs at Yu Ying's dormitory building. Zuo Cheng watched her go upstairs before turning the car around and leaving.

He didn't go straight back to his dorm; instead, he went to the company. The framework of the technical solution was finalized, but many details still needed refinement. There were only two days left before he had to submit the solution to Professor Li, and time was tight.

The office was empty. Zuo Cheng turned on the desk lamp and began writing a document.

Outside the window, the night sky over Hangzhou began to lighten, and a new day was about to begin.