FAQ: AC 5.X Professional AI Services Deployment

• How do I create a Resource Group in Azure?

  • Follow these steps:

    • Create a "Resource Group" with the necessary tags.

    • Select "Basics" and provide project and resource details.

    • Mention the name, value, and resource for the tags.

• What is the App Service Plan and how is it created?

  • The App Service Plan is used to host applications. To create it:

    • Mention the name, operating system, and region.

    • Provide necessary tags (name, value, and resource).

• What steps are involved in developing a Web App?

  • The following steps should be followed:

    • Specify the application name and publish type.

    • Choose the runtime stack and Java web server stack.

    • Define OS requirements and select a deployment region.

    • Choose the pricing plan recommended by the product team.

    • Provide GitHub account reference for image retrieval.

    • Configure networking options (public access should be set to "ON").

    • Enable monitoring with Application Insight.

• How do I manage APIs within AC 5.X?

  • Create an "API Management Service" and follow these steps:

    • Add instance details and select the pricing tier.

    • Configure APIs by adding display name, description, web service URL, URL scheme, and API URL suffix.

    • Add and update policies as needed, and develop GET and PUT operations.

• How do I create an AKS (Azure Kubernetes Service) Cluster?

  • To create an AKS Cluster:

    • Go to "Basics" and add cluster details.

    • Configure and deploy the cluster.

• What is Milvus and how is it configured?

  • Milvus is an open-source vector database designed for similarity search and AI applications. The document outlines its configuration, though specifics beyond its mention are not detailed.

• What are the best practices for monitoring web applications?

  • Use the "Monitoring" option within Application Insight to access relevant insights and data for better monitoring of web applications.

• What are the key considerations for networking during deployment?

  • Ensure that public access is enabled ("ON") and disable network injection by setting it to "OFF" during configuration.

Technical Questions: AC 5.X Professional AI Services Deployment

Resource Groups and Tags:

Question: What would be the minimum tags to attach to a Resource Group in Azure while deploying AI services, and why?

Answers: The primary tags are name, value, and resource. These tags will help organize and track resources to manage efficiently across the deployment. Tags will allow for much better categorization, and therefore cost management and reporting.

Question: How does resource tagging affect the management and accumulation of resources deployed?
Answers: Tags enable administrators to pick and classify those resources, which helps in managing assets more easily across development, testing, and production environments. It supports the automation of other tasks like billing and monitoring.

Configure App Service Plan

Questions: How would you choose an appropriate operating system and region in configuring an App Service Plan for AC 5.X?

Answers: This is based on the needs of the OS (stack), for example, Linux for Java-based stacks, and the region based on the need for latency compliance with data sovereignty rules and proximity to the user base for optimal questions performance.

Question: What determines the choice of a pricing plan for an App Service Plan, and how should be deemed necessary in project needs?

Answers: Depending on the expected traffic to the app, performance requirements, scaling needs, and indeed the budget constraint, it differs for a pricing plan. The most important thing for AC 5.X is product recommendation, which balances cost with high scalability and availability.

Web App Deployment:

Question: How is it different for varying runtime stacks to deploy a web app? In what way does your application's performance get affected by the Java Web Server Stack?

Answers: Diverse runtime stacks, like .NET, Python, or Java, optimize the applications differently according to the specific language. However, the Java Web Server Stack is optimized for Java applications with servlet management and JDBC integration; thus, it might enable better load-handling capabilities and database interactions.

Question: What's the instruction for setting public access without compromising the network security of a deployed Web App?

Answers: Make public access "ON" for a user, but enforce firewall rules and secure authentication via OAuth and SSL encryption to make sure that no unauthorized access will take place and data will be transmitted securely.

Question: What do you think would you consider linking a Web App to a GitHub repository to retrieve images?

Answers: First, validate that the right version of the app image is in the repo, and access permissions are correctly set. Continuous Deployment Pipelines should be designed to auto-deploy the changes to new versions with the ability to roll back more capabilities to be rolled back in case of failure.

API Management and Configuration:

Question: What is the role of API Management Service under AC 5.X, and how does this impact achieving the infrastructure to be scalable?

Answers: The API Management Service makes it very easy to expose your backend services through APIs securely and at scale. It supports features like rate limiting, authentication, and monitoring, so you can ensure that your APIs are scaling well through high traffic and conforming to all security policies while being integrated with multiple client applications.

Question: How do you set up APIs to handle interactions between different systems, and what is the best practice for suggesting API URL schemes and suffixes?

Answers: API is configured through a specification of the base URL with an appropriate suffix, such as /v1/ for versioning, along with operations like GET and PUT on the endpoints. Best practices would be distinct and descriptive names for the URLs, proper documentation, and, most importantly, making use of HTTPS for secure communication.

API Policies and Operations:

Question: How would you configure HTTP(s) endpoints in an API Management Service, and how does policy selection and management impact the behavior of the API?

Answers: A suitable HTTP(s) endpoint is chosen by inputting a URL and applying policies like rate limits, IP filtering, or caching. Policies add a direct impact on performance, security, and scalability, having a handling method of requests as well as ensuring compliance with business rules.

Question: What is the difference between GET and PUT operations in AC 5.X, and how are these used when managing system interactions?

Answers: GET is a retrieval operation, data getting fetched from the system, while PUT is an update operation for any existing data. In AC 5.X, GETs will retrieve system status information; PUTs are typically used for updating configurations or submitting new data for services to backend services.

AKS (Azure Kubernetes Service) Cluster Deployment:

Question: Which cluster information do we need to concern ourselves with in creating an AKS Cluster for AI Services and how does this provisioning impact the scalable nature of the system?

Answers: Such important features engaged are the cluster name, region, node size, and count. All these parameters affect the way how the cluster can scale up with the demand at hand. The larger the size of nodes, and the closer is the region from the users, reduces latency and performs better under huge loads.

Question: In what way does adding AKS clusters enhance the deployment and orchestration of AI services in production environments?

Answers: AKS clusters do provide auto-scaling, load balancing, and orchestration of AI workloads so that the services may be able to handle variable traffic without losing uptime. They make it much easier to manage containerized applications in a production environment.

Milvus Configuration:

Question: What are the major configuration steps of Milvus for vector similarity search, and how does this configuration integrate with AC 5.X for AI-based applications?

Answers: The steps mainly include defining the data structure, determining the indexing method-for example, IVF or HNSW-and optimizing query parameters to ensure efficient similarity search. In AC 5.X, Milvus adds integration to index large datasets for near-real-time search and retrieval.

Question: How do the capabilities of Milvus affect the retrieval and processing of data in AI systems?

Answers: Milvus's performance, especially in indexing and querying, affects the speed of retrieving data. Efficient indexing minimizes time for AI models to search, thus improving the responsiveness of applications that handle large-scale similarity searches.

Application Insight for Monitoring:

Question: What is Application Insight and how can it be used for monitoring deployed web applications, and what specific metrics should be tracked for optimal performance?

Answers: Application Insight can track metrics such as request response times, failed requests, CPU and memory usage, and dependency call failures. Tracking such ensures that bottlenecks are identified, errors are reduced, and performance is optimized.

Question: How do you configure alerting and diagnostic settings in Application Insight to preemptively detect and resolve issues?

Answers: Alert thresholds may be configured for critical metrics like when response time exceeds a limit. Log-based diagnostics might also be set up to trace detailed error reports. Emails and SMS can be enabled to intervene in issues that may occur in a timely manner.

Networking and Security Considerations:

Question: A web application has its public access set at "ON." What is the security implication of this action, and how might you mitigate such risks without disabling public access?

Answers: This increases the vulnerability as access to the application is provided to everyone who has it. Mitigation of this risk can be done using secure authentication, like OAuth, IP whitelisting, and enforcing SSL/TLS for encrypted communication.

Question: Does "OFF" in-network injection affect the deployment and performance of AI services in AC 5.X?

Answers: Network injection disabling reduces the exposure of the internal network and isolates the service from the non-essential network traffic; therefore, it will enhance security and stability. It could have a minor performance impact because certain services depend on network resources but mostly enhance resiliency.

Scaling and Performance:

Questions: How do you optimize the deployment of AC 5.X AI services for optimal scaling in a production environment?

Answers: Auto-scale your AKS cluster, load balancing for web applications, and monitor resource utilization in Application Insights to trigger scaling actions based on demand. Preemptive scaling based on traffic pattern ensures smooth performance.
Question: Which of the following is the most appropriate set of performance metrics to be monitored to ensure that deployed AI services are within their respective SLA parameters?

Answers: Monitor response times, error rate, CPU, memory usage, and uptime. Low latency, minimal downtime, and maximal utilization of resources will achieve the goal of SLAs.