Community Meeting | The MOSN community will release version 1.0 while promoting the evolution of the next-generation architecture

On February 24th, MOSN hosted its first community meeting of 2022.

The MOSN community put forward the new year's Roadmap at the meeting. Community members shared the experience of MOSN's implementation in different scenarios, and everyone opened up their minds and explored more possibilities that we can create.

MOSN Community Roadmap

MOSN's main goal in 2022 is to release MOSN 1.0 and open source a new out-of-the-box product. At the same time, it promotes the evolution of MOE (MOSN 2.0 architecture) and connects more ecological components.

With more and more users of MOSN landing, stability construction is also the key development direction this year, so everyone can use it with confidence.

Community meeting sharing guests

Users from Tantan Technology, Alibaba Cloud, and Qunar have actively shared their use cases in this conference for your reference.

Tantan Technology

Xie Zhengyao listed in detail the problems encountered in the process of MOSN landing and the pits stepped on, which provided a good idea for the follow-up optimization of MOSN.

There are many pits that we have already arranged, and the developers of MOSN are on the way to fill the pits.

Ali Cloud

Mu Yi listed the plan for the new fiscal year, and shared with you the tenant split of the cross-cluster service discovery scenario at the edge, as well as the new deployment form.

Qunar

Jialong pays more attention to the GC aspect in Roadmap, hoping to introduce some high-performance network frameworks and have more needs for performance optimization.

And I am looking forward to the holmes of the MOSN community, hoping to solve the problem when checking the problem.

QA review

1.Q: What are the landing scenarios and best practice blogs of MOE?

A: We will share the specific content at the first Service Mesh Summit 2022. There will be more landing scenarios this year. When trying to replace the access layer gateway, it will also be launched on a pilot basis. You can look forward to it!

2.Q: Can I understand that - if MOE is received in Envoy, I can write a filter through Go code to make it run on Envoy, and then I can use Envoy directly for my data link?

A: Yes, that's what it means. Now our demo can be played, but some interfaces have not been standardized. Now our internal sofagw is this architecture.

"demo document" ⬇️:
https://github.com/mosn/mosn/blob/master/pkg/networkextention/README-cn.md

3. Discussion on GC optimization methods

A: (1) It is indeed effective to reduce the GC frequency, which can reduce the repeated marks of long-lived objects. (2) However, this kind of pre-allocation is not very flexible. It is best to dynamically adjust the GC Percent to keep the GC goal at the expected water level.

In essence, it is a memory-for-CPU scheme. When the memory is sufficient, the water level of the GC goal is raised.

We also considered this situation when we engaged in holmes memory exception capture. https://uncledou.site/2022/go-pprof-heap/

Everyone spoke freely and opened their minds in this meeting. It is also the communication with users that makes MOSN's development plan and user needs complement each other.

Thank you for your active cooperation. With your help, the MOSN community will continue to promote performance optimization, technology implementation, and grow together with users.

We will also hold community meetings later, such as when MOSN releases a new version or has a big progress. Listen to user feedback, synchronize industry trends, and look forward to the next meeting~

If you want to book the next community meeting, learn more about MOSN community dynamics, DingTalk search: 21992058

Recommended reading of the week

Opens a new chapter of cloud native MOSN - Integrating Envoy and GoLang

Multiprotocol Extension Development Practice1621ecfaee022b

The exploration and practice of Service Mesh in Industrial and Commercial Bank of

The Next Five Years of Cloud Native