ACA-CloudNative ACA Cloud Native Certification Exam Questions and Answers

True

False

True

False

Virtual machine

Big data analytics

Containers

Cloud provider

RDBMS

True

False

Logs

Metrics

APM tracing

All above

Duration of transaction

Flow of transaction

Anomalies

Transaction logs

Data Schema

ACR is able to replicate images across regions

ACR provides vulnerability scanning for uploaded images

ACR can manage helm charts

ACR's standard version includes cross-region image replication

Composable workflow, support continuous training pipelines,

Scale out training out to hundreds of nodes

Support heterogeneous resources management, such as GPU, FPGA, RDMA.

All resources are free

Accelerates remote data reading for AI training jobs, and reduces GPU/CPU waiting time

Supports different storage backends including OSS/HDFS/NAS

Multiple cache layers, including RAM/SSD/HDD

Supports preload once, and reads by many jobs

Supports horizontal scaling

Replicates 3 copies of local data to avoid data loss

Because need to load all used & unused modules of applications

Because need to load only all used modules of applications

Because need to load all unused modules of applications

Because need waiting approval from Project Manager

High availability with multi-zone redundancy to minimize business disruptions

Easy to scale up or down (vertical or horizontal)

Set up infrastructure and provision resources in less than 1 day

No need for expert Database Administrators for setup & maintenance of database

Fully monitor your infrastructure in real-time

All resources that you used become free

ACR EE allows you to perform security scans on all Linux-based images

You can receive vulnerability assessments and recommendations

With the cloud-native delivery chain, the system will automatically block the deployment of

high-risk images

With the cloud-native delivery chain, an attacker can deploy an unsigned image on the ACK

cluster E. By default, a newly created ACR EE is connected to all networks

Al

DeepLearning

Batching compute

HPC

Domain-Driven Design

Decompose by Business capability

Centralize all components(Authorization, Presentation, Business logic, Database layer) use single-tiered software application combined into a single program from a single platform

Divide and conquer

Decompose by Subdomain

The development of a single application becomes a centralized, including all dependencies on bare metal.

Developing a single application while keeping data consistency with transaction control (like row locking then commit, will rollback when failed)

Developing a single application as a suite of small and independent services that are running in their own processes, can develop and deploy independently.

About developing a single application become modular in the level of method/function, but keep only in 1 object app because we need to manage it easily.

Heterogeneous computing is supported.

All major machine learning frameworks are supported.

Optimized K8S scheduling is provided for jobs running on the GPU. For example, it can ensure

pods are scheduled on the GPU node, provide shared GPU resources, isolate GPU resources, and do

co-scheduling.

GPU and shared GPU utilization can be monitored via Prometheus on Alibaba Cloud.

The "Arena" tool is developed by the community but integrated with Alibaba Cloud ACK.

Serverless cluster is nodeless

Serverless cluster supports GPU and spot instance

Serverless cluster can add ECS instance as worker node

Kubernetes Cluster Federation

Auto Scaling to Serverless Kubernetes

Register External Kubernetes Cluster to be managed from Alibaba Cloud Container Service Console

Manage every clusters manually but to connect to the same image repository

Cloud Disk

NAS

OSS

Local Disk

Elasticsearch

Kibana

Grafana

Logstash