Last updated on Jun 07, 2026
All Microsoft Azure Administrator certification learning material, study guide, training courses are created by a team of Microsoft training experts. The Study Guide and .EXM training software files contain relevant Microsoft Azure Administrator content, labs, practice questions and explanation. This AZ-104 exam guide and training courses is based on the latest exam outlines available!
Struggling with a complex question? Just ask your AZ-104 AI tutor. It explains concepts, clarifies why wrong answers are wrong, and helps you understand AZ-104 topics in depth, available 24/7, included at no extra cost.
Don't just see the right answer, understand why it's right and why the others are wrong. In any Language!
Your AI tutor is available around the clock. No scheduling, no waiting — help is one click away inside the practice test.
Available directly in your online practice session. Click "Ask AI" on any question and get an instant explanation.
One-time payment, instant access
Launch the exam online
Get an instant explanation
Take the first step towards passing your AZ-104 exam with ease by investing in our comprehensive certification exam material.
As a student looking to advance your career in the field of cloud computing and demonstrate your proficiency in Microsoft Azure administration, taking the Microsoft AZ-104 exam is a crucial step. This article will guide you through the process of preparing for and successfully passing the AZ-104 exam, providing you with accurate and up-to-date information from the Microsoft website.
The AZ-104 exam, also known as "Microsoft Azure Administrator," is designed to assess your skills and knowledge in various aspects of Azure administration. It validates your ability to manage Azure identities and governance, implement and manage storage, deploy and manage virtual machines, configure and manage virtual networks, and monitor and back up Azure resources.
The AZ-104 exam evaluates your proficiency in the following areas:
To increase your chances of passing the AZ-104 exam, it is crucial to have a structured study plan and access to relevant learning resources. Here are some actionable tips to help you prepare:
On the day of the AZ-104 exam, keep the following tips in mind:
By following these tips and investing time and effort in your preparation, you can enhance your chances of passing the AZ-104 exam and establishing yourself as a skilled Azure administrator. Good luck!
Question 1811:Correct answer: D Reason: If encryption keys are not centrally managed, the DLP tool cannot reliably decrypt and inspect data across the environment. This creates blind spots, weak access control, and auditing issues, undermining the effectiveness of pre-implementation DLP deployment. Why the others are less critical in this context: Monitor mode vs block mode affects enforcement; monitor-only reduces effectiveness but is not as fundamental a risk as broken key management. Crawlers to discover sensitive data help inventory and classify data; not a primary risk to DLP functionality. Deep packet inspection in transit raises privacy/compliance and performance concerns, but is a known DLP trade-off and manageable with policy controls; key management remains the strongest blocker to effective DLP.
Question 1811:Correct answer: D Reason:
Question 121: Correct answer: B — a virtual network for FinServer and another virtual network for all the other servers. Why: - In Azure, network segmentation is done with VNets. Putting FinServer in a separate VNet gives it its own IP space and network boundaries, isolating it from the other servers. - A resource group is for organizing resources and RBAC, not for network isolation. - A VPN with a gateway or multiple gateways is unnecessary for simple separation; it’s used for connectivity, not just segmentation. - One resource group with a lock does not affect network isolation. Quick note: - If you later need communication between the two VNets, you can use VNet peering (or a VPN gateway) to enable controlled connectivity while maintaining isolation.
Question 121:
Question 86: Correct answer: Vertical scaling Why: Vertical scaling (scale up/down) means increasing or decreasing the size of a VM by adding memory or CPUs to the same VM. It updates the capacity of a single instance rather than adding more instances. How it compares to other terms: - Horizontal scaling (scale out/in): changes the number of VM instances, not the size of each one. - Elasticity: broad concept of adapting resources to demand (includes vertical and horizontal scaling). - Agility: general capability; not specific to VM capacity. Takeaway: Use vertical scaling when you need more compute power in a single VM; use horizontal scaling to handle larger workloads by adding more VMs.
Question 86:
Question 73:I can’t see the image-based statements in Question 73, so I can’t tell which specific statements are true. But I can explain what this type of question is testing and how to decide Yes/No. What Question 73 is testing Your understanding of the cloud service models and the shared responsibility model: IaaS, PaaS, and SaaS. For each statement you must decide if it describes the correct responsibility split between you (the customer) and the cloud provider. Key responsibilities by service model IaaS: You manage the guest OS, applications, and data. The provider manages virtualization, servers, storage, and networking. PaaS: You manage the applications and data. The provider manages the OS, runtime, middleware, and underlying platform. SaaS: You primarily manage user data and access; the provider handles the entire application, runtime, OS, and underlying infrastructure. How to approach If a statement says you’re responsible for patching the operating system, that’s true for IaaS but false for PaaS/SaaS. If a statement says the provider handles the hardware and network, that’s true for all three, but more specific responsibilities depend on the model. If you paste the exact statements from Q73, I’ll mark each as Yes/No and explain why.
Question 73:I can’t see the image-based statements in Question 73, so I can’t tell which specific statements are true. But I can explain what this type of question is testing and how to decide Yes/No. What Question 73 is testing
Question 25: Correct answer: D: a new Cognitive Services resource that uses the S0 pricing tier. Why: To enable OCR and text analytics in a Cognitive Search enrichment pipeline, you must attach a Cognitive Services resource to the skillset. For cost efficiency on a large document set, choose the base standard tier (S0) rather than higher tiers or a free tier, which may limit enrichments. The OCR capability comes from the Cognitive Services (Computer Vision/Read) and the S0 tier provides a balance of capability and cost.
Question 25:
Question 20: Correct answer: D and E. Why: Provisioning a QnA Maker service creates: - Azure App Service to host the QnA Maker web API (your App Service Plan AP1 will host the App Service resource). - Azure Cognitive Search to enable fast search over the knowledge base. Why not the others: - Language Understanding, Azure SQL Database, and Azure Storage are not automatically created by QnA Maker provisioning. Quick note: After provisioning, check RG1 to verify the new App Service and Cognitive Search resources.
Question 20:
what if different between speech recognizer and speach synthesizer Speech Recognizer (Speech-to-Text): Converts audio input (live mic or audio file) into text. Output is transcripts with timestamps, confidence scores, and optional punctuation. Use cases: meeting transcripts, voice commands, real-time captions. Speech Synthesizer (Text-to-Speech): Converts text (or SSML) into audio output. Output is audio data (wav/mp3 or streaming). Use cases: voice prompts, IVR responses, accessibility narrations. You can choose voices, languages, and adjust prosody (rate, pitch, volume). Key differences: - Input: Recognizer takes audio; Synthesizer takes text/SSML. - Output: Recognizer ? text; Synthesizer ? audio. - Features: Recognizer provides transcripts and timing; Synthesizer provides customizable voices (neural, SSML), multiple languages. - Typical scenarios: Transcription vs. generating spoken content. Quick example: Use Speech Recognizer to transcribe a meeting; use Speech Synthesizer to read a script aloud in a chosen voice.
what if different between speech recognizer and speach synthesizer
Question 17: Correct answer: A Anomaly Detector. Why: Anomaly Detector is designed to identify unusual values in time-series data. Your scenario has 100 machines × 50 sensors, generating minute-by-minute data, totaling 5,000 time-series. Anomaly Detector can process each time-series (or batches of series) to flag deviations that may indicate potential failures, which is exactly what you need for predictive maintenance. Why not the others: - Cognitive Search is for indexing and querying content, not for detecting anomalies in time-series data. - Form Recognizer extracts data from forms, not time-series sensor data. - Custom Vision analyzes images, not numeric sensor streams. Practical note: with 5,000 time series, you’d typically run anomaly detection per series (potentially in parallel) and aggregate results to identify which machines/sensors warrant attention.
Question 17:
Question 12: Correct answers: A, B, F. Why: - A. The index size will increase. Enabling CMK encryption adds encryption metadata and key management data, which increases index size. - B. Query times will increase. Encryption/decryption overhead and key retrieval can slow queries. - F. Azure Key Vault is required. CMK means you store/manage keys in Key Vault; it’s a billable, required service for CMK. Why the other options are incorrect: - C (self-signed X.509 certificate required) is not a requirement for CMK. - D (index size will decrease) and E (query times will decrease) contradict the expected impact of CMK.
Question 12:
GIF also we are using ? No, GIF is not supported for Form Recognizer input. The allowed formats are JPG, PNG, PDF (text or scanned), or TIFF. If you have GIFs, convert them to one of the supported formats (e.g., convert to JPG/PNG or PDF) before training or uploading. Also ensure each file is under 50 MB and, when possible, use text-embedded PDFs to reduce errors.
GIF also we are using ?