Openmldb meetup No.4 meeting minutes

The content of the meeting

OpenMLDB Community in 2022 year 6 month 25 The fourth session was held on August meetup, The videos and materials related to the meeting are as follows ：

OpenMLDB PMC core member Lu Mian , From low cost 、 High performance online and offline consistency feature platform open source solution , Introduce the consistent production level feature platform online and offline OpenMLDB And its v0.5.0 New performance improvements in the new version 、 Cost reduction 、 Flexibility adds three features . Video link ：https://www.zhihu.com/zvideo/1525167543303700480 The data link ：https://pan.baidu.com/s/12Vk0syf-WgqQ1uBAXtTNBw Extraction code ：open

Apache RocketMQ PMC member ding , A message processing platform tempered by trillions of flood peaks ——Apache RocketMQ For topic , Help you understand Apache RocketMQ High availability architecture and cloud native era RocketMQ New challenges . Video link ：https://www.zhihu.com/zvideo/1525170014009634816 The data link ：https://pan.baidu.com/s/1vU8K1CA7D1aUO3-vggEXRw Extraction code ：open

Apache RocketMQ connect contributor Sunxiaojian , The depth resolution Apache RocketMQ Technical principles and application scenarios of , borrow RocketMQ OpenMLDB Connector Lead the audience to explore how to base on RocketMQ Quickly build a lightweight data flow platform . Video link ：https://www.zhihu.com/zvideo/1525170045735370752 The data link ：https://pan.baidu.com/s/1fJHmiC_-awjKPtkQFvaTYg Extraction code ：open

Discussion and exchange

Q1： Excuse me, Scala Interface in OpenMLDB Is it in the follow-up development plan of ？ A1： Scala In theory, we can also call our Java SDK To connect .Scala Native SDK At present, in our Roadmap On , But it has not been specifically promoted . If you are interested in this requirement , Also welcome to join the community to develop with us .

Q2： OpenMLDB Support vectorization ？ A2： at present OpenMLDB Mainly structured tabular data , The vector search function is not supported （ There are many excellent open source tools to support this part of the function ）. If you want to store vectorized data , In fact, it can also be converted into tabular form , If you need vectorized data for feature calculation, you can use OpenMLDB .

Q3： Excuse me, OpenMLDB How to set lable？ If it is only characteristic calculation , Which model is better for training ？ A3： OpenMLDB Provide feature calculation function , But we can't do model training and reasoning , therefore label This column is in OpenMLDB There is no need for special treatment here , Just keep it and give it to the later model training . There are many options for model training . You can refer to our case , Pick up LightGBM, Or other models , such as XGBoost, Sklearn etc. .

Q4： I've just been in touch recently featurestore product , Find out OpenMLDB and Feast Comparison of such open source products , Looks missing feature view Metadata management . as well as feast This kind of product is to give computing power to other cloud computing ,OpenMLDB More importantly sql engine Computing power . I don't know if the understanding is correct ？ A4： feature view The definition of seems to be the same as we use SQL Definition feature The concept is similar , We use forms schema, as well as SQL To describe the data source format and feature Decimation logic . Of course, this one Feast The definition of is more concrete , We will learn this description . Comparison Feast We emphasize real-time feature computation , Based on the freshest data , Instead of offline feature calculation and online retrieval .

OpenMLDB Community

I would like to thank you for your meetup Strong support , If you want to know more OpenMLDB Or participate in Community Technical Exchanges , Relevant information and interaction can be obtained through the following channels . Github: https://github.com/4paradigm/OpenMLDB Official website ：https://openmldb.ai/ Email: [email protected] OpenMLDB WeChat ac group ：