# GSoC 2020 Apache Proposal # Apache RocketMQ Scaler for KEDA ## Application Name : Hien Nguyen University : Haaga-Helia University of Applied Sciences - Bachelor of Information Technology - (Location: Distributed system, Cloud(AWS, Azure), Golang, Maven, Docker, Kubernetes ## GSoC - Apache RocketMQ Scaler for KEDA proposal ## Context KEDA allows for fine-grained autoscaling (including to/from zero) KEDA project for new feature request for scaler for Apache RocketMQ. KEDA is written in Go (gPRC). KEDA has ScaledObject for the service, Metrics adapter, Scaler, Controller. RocketMQ operator deployment

扩展能力之间的冲突关系，如何有效管理？如何有效的对运维透出？ apiVersion: "app.alibaba.com/v1" kind: CronHPA metadata: name: cron-scaler spec: timezone: Asia/Shanghai schedule: - cron: '0 0 6 * * ?' minReplicas: 20 "the docker image of the function" } } 1 kubectl get traits 2 NAME AGE 3 cron-scaler 19m 4 auto-scaler 19m ## 可发现、可管理的运维能力：OAM Traits System ## 发现运维能力 ## $ oamctl trait-list |NAME|VERSION|PRIMITIVES| metadata: name: cron-scaler annotations: version: v1.0.0 description: "Allow cron scale a workloads that allow multiple r spec: appliesTo: kubectl get traits crob-scaler -o yaml - core

This report documents the results of a security assessment targeting the Vitess software database scaler. Funded by the CNCF / The Linux Foundation, this project was carried out by Cure53 in February 2019 of this Cure53 assessment funded by CNCF / The Linux Foundation certify that the Vitess database scaler is secure and robust. This very good outcome is achieved by limiting the attack surface, taking appropriate final test report. In light of this February 2019 project, Cure53 concludes that the Vitess database scaler is mature and secure. Therefore, it is deemed fit-for-purpose as far as deployment in modern scalable

Scikit-learn主要用法 ## 数据预处理 使用Scikit-learn进行数据标准化 from sklearn.preprocessing import StandardScaler 构建转换器实例 scaler = StandardScaler() Z-Score标准化 拟合及转换 $$ x^{*}=\frac{x-\mu}{\sigma} $$ $$ \begin{aligned}&a m}(x^{(i)}-\mu)^{2}\\ &\mu=\frac{1}{m}\sum_{i=1}^{m}x^{(i)}\\ \end{aligned} $$ $$ \mathtt{scaler.fit\_{t}ransform(X\_{t}rain)} $$ 处理后的数据均值为0，方差为1 ### 2. Scikit-learn主要用法 ## 数据预处理 使用Scikit-learn进行数据变换

– 运维能力 ## ☀️ ☀️ ☀️ apiVersion: core.oam.dev/v1alpha2 kind: TraitDefinition metadata: name: scaler appliesToWorkloads: - containerizedworkloads.core.oam.dev # arbitrary workload definition # support cmd: - sleep - "1000" traits: - name: scaler properties: replicas: 10 - name: metrics properties:

- componentName: oamfrontend instanceName: oam-fe traits: - name: manual-scaler properties: replicaCount: 1 - name: ingress properties: - name: oam_texture value: aks traits: - name: manual-scaler properties: replicaCount: 1 - name: ingress.core

using below code: x_train_scaled = preprocessing.scale(x_train) scaler = preprocessing.StandardScaler().fit(x_train) x_test_scaled = scaler.transform(x_test) Here, we have normalized the training data using

e> channelsPerPixel); FactoryImpl class has the only method. This method creates the scaler object and returns it via the abstract interface. Pay attention that create() method has no generic each one # with the corresponding compiler flags ko_compile_for_all_implementations(__per_arch_rgb_scaler_factory_objs KoOptimizedPixelDataScalerU8ToU16FactoryImpl.cpp) else() # in case XSIMD is not options (x86_64) set(__per_arch_rgb_scaler_factory_objs KoOptimizedPixelDataScalerU8ToU16FactoryImpl.cpp) endif() # ... set(kritapigment_SRCS # ... ${__per_arch_rgb_scaler_factory_objs} # ... ) Now we have

Number of CPU units for the service, like 0.5(0.5 CPU core),1(1 CPU core) string false $ vela show scaler /Users/zhouzhengxi vela show webservice NAME DESCRIPTION TYPE REQUIRED DEFAULT cmd Commands

metadata: name: oam-helloworld-app spec: - name: oamfrontend - name: oambackend traits: - name: scaler parameterValues: value: 1 ## scopes: path: /metrics protocol: https ## Trait For assigning operational