Hi, community: As the previous thread mentioned[1], Sheng and I had a long time discussing a dedicated database for storing skywalking hyper-modal data. The first idea could be traced back to the time I joined SkyWalking, which still was Sheng's personal project. The main architecture is discussed in my graduate thesis in 2017, with the title of "The design and implement of tracing database system". In that paper, I discussed a potential database implementation to handling tracing data and design an abstract model to save the data.
After that, several issues blocked me from pushing this work till I figured them out about one month ago. Simultaneously, the debate of the ES's license transferring caused Sheng and me to decide to kick the project off. In recent weeks, I composed a list[2] to trace the main ideas and tech stacks that could affect the final architecture. If you have any ideas, feel free to comment on them. >From my perspective, we would verify some potential data modals before starting the design. Based on that, I begin a study[2] to choose an idea data modal from several candidates. In this study, I should set up some stimulation workloads to test data models. These workload designs need your help to make them as close to the real scenarios. If anyone has any ideas, pls let me know. In the next few weeks, I will be on the study. Once some results merge, the community will get the updates. * 1. https://github.com/apache/skywalking/issues/6219#issue-787602819 * 2. https://docs.google.com/document/d/1qzzJ3caBtAFmFBBoB0VoETqWAUMP2jSR4UR3C3Jykos/edit?usp=sharing * 3. https://docs.google.com/document/d/1bvvycAe1MlXKIJPdGUFBS8XDI7nPpIpI44RKwH_UkJc/edit?usp=sharing -- Hongtao Gao Apache SkyWalking && Apache ShardingSphere Twitter, @hanahmily
