In the era of cloud computing, cloud vendors help us solve the maintenance and expansion of computing and storage resources, but users also face the problem of being bound by vendors.

In 2021, UC Berkeley proposed the concept of Sky Computing (“Sky Computing”), whose goal is to allow applications to run across multiple cloud vendors and achieve interoperability between multiple clouds. Computing power and data migration between clouds are two basic problems that Sky Computing needs to solve. With the development of virtualization technology, computing power migration is gradually being solved, and stateless computing-intensive tasks can be run indiscriminately on various cloud vendors. However, there are still major challenges in the safe and fast cross-cloud data migration and synchronization.

Most of the distributed systems in the background of cloud computing run on multiple servers in the same data center, with low network delay and high reliability. Cross-cloud means an increase in the physical distance between servers, resulting in increased network latency and reduced reliability. It is more difficult to maintain consistency between servers and establish consensus. Xline is a cross-cloud KV database developed using Rust, compatible with Etcd’s metadata storage interface. Xline is implemented based on the CURP protocol, and has better performance than Raft-based Etcd in the case of cross-cloud deployment. Experiments show that the delay is twice as low under ideal conditions. Using Xline can realize data sharing and access across data centers, and ensure data consistency, which is convenient for business systems to implement multi-site and multi-active deployment.

News Fast Delivery’s current Q&A with experts (February 1-February 7) we invitedShi JichengThe teacher discussed issues related to “Sky Computing” with everyone.

Issues that may be discussed include, but are not limited to:

• Advantages of Sky Computing over Cloud Computing
• Technical Challenges Facing Cross-Cloud Storage
• How Xline achieves low-latency cross-cloud data interaction
• Principle and selection of distributed consensus protocol
• How to create a low-coupling and reusable protocol framework
• Advantages and disadvantages of Rust development

If you have other related questions, you are welcome to actively ask!

Guest introduction

Shi Jicheng, co-founder and CTO of DatenLord, has worked in Internet companies such as Google, Microsoft, and Alibaba, and has many years of experience in research and development of operating systems and distributed systems.

In order to encourage active questions, we will draw 5 lucky members from the questioners after the Q&A to give a book of “Rust in Action”.

The consistent style of OSChina Q&A does not welcome any discussions and trolls that have nothing to do with the topic.

Below, everyone is welcome to ask Mr. Shi Jicheng questions about programming language design and development, and you can just reply to the questions.