search

matrixorigin / matrixone

Hyperconverged cloud-edge native database
https://docs.matrixorigin.cn/en
Apache License 2.0
1.79k stars 276 forks source link

[Question]: Transaction layer architecture #4900

Open intech opened 2 years ago

intech commented 2 years ago

Describe your problem

Could you describe the solution to the concurrency problems in MVCC and the layers above?

Ariznawlll commented 1 year ago

useless issue

fengttt commented 1 year ago

useless issue

en, this is obviously useful. It says our architecture doc is lacking.

intech commented 1 year ago

When I first got to your product and read the stated features and guarantees, I had many architectural questions. I asked just one question. I know this question is challenging to implement, and answering it will show how much you know about the claimed guarantees. It's been seven months, and the answer is still not in this issue or the system design documentation.

I took another look at the architecture page from the documentation. After flipping through several pages of issues, I still conclude that the system is under-documented and raises many questions about its application.

You could improve this by providing more architectural features, not just in the form of what you guarantee but how you do it. What's missing is the algorithm used and links to publications. And about the limitations currently known related to implementing these algorithms.

It would also be great to see a report on Jepsen testing.

dengn commented 1 year ago

When I first got to your product and read the stated features and guarantees, I had many architectural questions. I asked just one question. I know this question is challenging to implement, and answering it will show how much you know about the claimed guarantees. It's been seven months, and the answer is still not in this issue or the system design documentation.

I took another look at the architecture page from the documentation. After flipping through several pages of issues, I still conclude that the system is under-documented and raises many questions about its application.

You could improve this by providing more architectural features, not just in the form of what you guarantee but how you do it. What's missing is the algorithm used and links to publications. And about the limitations currently known related to implementing these algorithms.

It would also be great to see a report on Jepsen testing.

Sorry. this project is under intense development, documentation is more on the usage side. @XuPeng-SH can you explain about this question: the solution to the concurrency problems in MVCC and the layers above?