Forem: Marius

Meet AI: Introduction to the World of Artificial Intelligence

Marius — Sun, 12 Oct 2025 09:41:26 +0000

What is AI

Artificial Intelligence, or AI, is a field of science about building machines that can think and learn like we do. Essentially, it's about teaching computers to do things that normally require human intelligence: understanding speech, recognizing faces, solving problems, or making decisions.

For example, you've probably seen AI in action when you talk to virtual assistants like Siri or Alexa, when Netflix recommends a show you might like, or when your phone automatically fixes your spelling mistakes. Beyond these everyday uses, programmers are using AI assistants to write and debug code faster, and artists are using AI tools like Midjourney to create book illustrations in hours instead of weeks.

Understanding AI fundamentals will help you make informed decisions and participate in one of today's most important conversations. AI isn't just one thing—it's a collection of different techniques and approaches, and this guide will help you understand the key distinctions.

In this guide:

Machine Learning
Deep Learning
Discriminative vs. Generative AI
Why Generative AI is Rising
Understanding AI's Limitations
Your AI Journey Starts Here

Machine Learning

Machine Learning is a subset of AI that lets computers learn from data and improve at specific tasks over time, without requiring explicit programming for each scenario. Instead of following fixed instructions, a machine uses patterns and examples from past data to make predictions or decisions on its own.

At its core, Machine Learning focuses on recognizing patterns. By analyzing extensive datasets, machine learning algorithms learn how to spot those patterns, allowing them to make predictions or decisions based on what they’ve learned.

When we say a machine learning system is 'trained,' we mean it's shown thousands or millions of examples with the correct answers.

For instance, to train an image recognition system to identify cats, we'd show it thousands or millions of photos labeled 'cat' or 'not cat.' The system adjusts its internal parameters with each example until it can accurately identify cats in new photos it's never seen before.

Deep Learning

Deep Learning takes Machine Learning further by using artificial neural networks (loosely based on how neurons connect in the brain but simplified) to learn from enormous volumes of data. These networks can automatically learn different levels of patterns and features from raw data. For example, when learning to recognize faces, the first layer might detect simple edges, the next layer combines those into facial features like eyes or noses, and deeper layers recognize complete faces. This happens automatically—we don't have to tell the system what features to look for.

Deep Learning is especially useful for complex tasks like understanding language, recognizing images, and even creating new content.

Discriminative vs. Generative AI

Within Machine Learning and Deep Learning, a key distinction is between discriminative and generative approaches.

Discriminative AI analyzes and classifies existing data. These systems look at data and answer questions like "Is this X or Y?" or "What category does this belong to?". This powers everyday applications: flagging spam emails, detecting fraudulent transactions, recognizing your face to unlock your phone, or fixing your typos when autocorrect changes 'teh' to 'the'.

Generative AI takes things further by actually creating new content, conversations, and ideas - teaching machines to generate text, images, music, or videos that look or sound like something a human made. It has creative potential and broad applicability across industries.

While both approaches have existed for decades, generative AI is the most discussed area of Machine Learning and Deep Learning right now.

The diagram below illustrates how these different AI concepts relate to each other, showing how Machine Learning and Deep Learning fit within the broader field of AI, and where Generative AI sits within that landscape.

This visualization shows the relationship between AI, Machine Learning, Deep Learning, and their various applications, helping you see how the concepts we've discussed fit together.

Why Generative AI is Rising

While AI research began in the 1950s, the recent explosion in capabilities—especially in generative AI—has happened in just the last few years. What changed? The three factors we'll explore below:

1. Computing power

We have significantly more of it now thanks to GPUs and specialized AI chips that can process massive datasets at incredible speeds. This computing power is what makes training and running these big models possible.

2. Algorithmic breakthroughs

We've had major advances in Machine Learning, especially with new algorithms and transformer architectures (a type of model that helps AI understand and generate human-like content better than older approaches).

3. Data

The internet has created a massive amount of information for AI to learn from. From text and articles to photos and videos shared online, there's now significantly more training material available than ever before.

When you combine all this—powerful hardware, better algorithms, and vast amounts of data—you get the perfect conditions for generative AI. This convergence is why we're seeing AI capabilities that seemed impossible just a few years ago.

However, with all this excitement around AI's rapid progress, it's equally important to understand its boundaries. Knowing what AI can and can't do helps you use it more effectively, set realistic expectations, and think critically about where human judgment still matters most.

Understanding AI's limitations

While AI capabilities are impressive, it's important to understand what AI can't do and where it struggles. This helps you use it effectively and avoid common pitfalls.

AI doesn't truly "understand" like humans do

When ChatGPT writes a poem or answers a question, it's recognizing patterns in language and predicting what words should come next, without actually comprehending meaning the way you do. This is why AI can sometimes generate responses that sound confident but are completely wrong, a problem often called "hallucination".

AI is only as good as its training data

If an AI system is trained mostly on data from one group or perspective, it will reflect those biases. For example, early facial recognition systems performed poorly on people with darker skin tones because they were primarily trained on images of lighter-skinned faces.

AI lacks common sense and context

Ask an AI to create "a photo of an astronaut riding a horse on the moon" and you might get a beautiful image with details that make no physical sense, like shadows in impossible directions, because AI doesn't have real-world knowledge about how gravity or physics actually work.

AI can't explain its own reasoning

When a deep learning system makes a decision, it's often a "black box". Even the engineers who built it can't always explain why it chose a particular answer, which is concerning for high-stakes decisions like medical diagnoses or loan approvals.

Security and privacy concerns

AI systems can be tricked by small changes invisible to humans, and AI models trained on personal data raise serious questions about privacy and how that information is used.

AI is a powerful tool, but it's not infallible or intelligent in the human sense. Treat it as an assistant that needs oversight, not an authority that's always right.

Your AI Journey Starts Here

Many AI tools are now freely accessible. ChatGPT, Google's Gemini, and various image generators offer free tiers, making AI experimentation available to anyone with internet access.

The most effective way to understand AI is to experiment with it yourself. Here's how you can get started:

Try conversational AI

Conversational AIs like ChatGPT by OpenAI, Google Gemini, Anthropic Claude, and others let you explore how natural language models understand, reason, and respond. You can use the free versions, explore different prompts, compare responses.

Experiment with image generation

Image generation AIs let you see how text-to-image models convert language into visuals. Tools you can try: DALL-E, Midjourney, Leonardo.ai.

Audio, Music, Video and Animation

AI tools can now create or enhance sound, music, and video.
Try ElevenLabs or Play.ht for realistic voice, Suno AI or Udio for song generation, and Runway ML or Pika Labs for AI video creation.

Explore AI Ethics, Bias, and Society

Ask different AIs moral or opinion questions, and compare how they reason or hedge. Understanding how AI affects people is just as important as using it.

Think critically about AI's role

As you experiment, consider these questions: Where does AI provide genuine value? Where might it be unnecessary or concerning? Which decisions should remain in human hands? When you see AI-generated content, consider questions about creativity, authorship, and authenticity. Forming your own opinions about AI's societal impact is just as important as understanding how it works.

The AI field is evolving rapidly, and you don't need to be a computer scientist to participate in the conversation or benefit from the technology. Your perspective as someone learning about AI for the first time is valuable.

Photo by Igor Omilaev on Unsplash

I didn't know how to engage in Deep Work

Marius — Sat, 31 Aug 2024 08:09:24 +0000

For many years, I struggled with a tendency to jump from one project or learning opportunity to another, only to abandon them a little bit later when something new came along. I found myself switching multiple projects or codebases simultaneously, often working in a reactive environment where I was constantly responding to client requests or changes in project scope. While this might sound productive, I soon realized that it led to a lack of deep expertise and mastery.

Shallow Work - common among developers

Shallow work is quite common among developers. It consists of simple, easy tasks often performed while distracted, such as checking emails, browsing the web, or handling minor administrative duties. When developers spend too much time on these activities, they have less time for focused, deep work, such as writing code or solving complex problems. This gives the illusion of productivity but actually reduces overall efficiency and slows progress. Other examples of shallow work include chat messages, meeting requests, and ad-hoc discussions. With the rise of remote work, the line between shallow work and personal tasks has become increasingly blurred.

Deep Work: rare and valuable

In his book Deep Work, Cal Newport discusses the rarity and value of the ability to focus deeply in our distracted, technology-driven world. Deep work is not only crucial for professional success but also for personal fulfillment. It allows for sustained attention and concentration, leading to significant growth and mastery.
Deep work is a method for being present on the work that matters, and for working with impact. Deep work reclaims productive focus.
Here are some practical strategies that helped me improved my work habits:

minimize Shallow Work: reduce the amount of low-value, non-cognitive tasks by automating or delegating them. For instance, use email filters to manage routine messages or delegate administrative tasks to support staff.
time blocking: allocate specific blocks of time for deep work and protect these periods from interruptions. For example, you might block out 2 hours for coding, 1 hour for meetings, and 30 minutes for email each day.
fixed-schedule productivity: set a specific number of hours you will work each day and avoid working beyond this time frame, even if you feel you could accomplish more. This encourages focused, efficient work within the set hours.

How developers can benefit from Deep Work

When developers engage in deep work, they are less likely to make errors or overlook details, leading to cleaner, higher-quality code. Deep work also enables developers to learn and master new programming languages, frameworks, or tools more quickly. Concentrated work with fewer interruptions accelerates development cycles and reduces stress, resulting in a greater sense of achievement and job satisfaction.

A lack of focus prevents developers from committing deeply to any single task or project long enough to develop true expertise or achieve mastery. Without focus, it is challenging to stick with a task through its more difficult or less exciting phases. Deep work, which requires sustained attention and concentration, is crucial for developing a high level of skill and knowledge in any field.

Conclusion

It's not that programming is super hard or that there are way too many technologies to learn, or even that you don’t have enough experience.

What’s really challenging is understanding that being great at software development isn’t just about knowing how to code or learning all the latest frameworks. It’s more about having the discipline to stick with it and keep improving.

This might not seem like a big deal now, but if you want to level up as a software developer—whether it’s to get better at what you do, feel more confident, or make more money—then it’s something to think about.

I’d love to hear from our tech-savvy community! Please share your thoughts in the comments below.

Photo by Nick Morrison on Unsplash

What developers have to face: Navigating the Multi-Dimensional Interview Process

Marius — Thu, 29 Aug 2024 10:50:22 +0000

In the competitive landscape of software development, the interview process evaluates candidates from multiple angles. As developers, we face a comprehensive assessment that goes beyond just the technical capabilities. Sure, problem-solving skills are valuable, basic data structure and algorithm knowledge is useful but companies are increasingly looking for versatile individuals who can demonstrate expertise in a variety of areas critical to their success.

Evaluating Analytical Skills

First and foremost, our analytical skills are evaluated. Interviewers look at how we approach complex problems, the time it takes us to arrive at a solution, how we structure the problem, and whether we have considered potential trade-offs, all while under high-pressure conditions.

Testing Coding Skills

Next, our coding skills are tested. Interviewers examine whether we can effectively translate an algorithm into clean, well-structured, optimized, and scalable code, as well as our ability to anticipate potential issues. They often use live coding challenges or ask us to complete take-home assignments to evaluate these skills.

Technical Knowledge and Experience

Our technical knowledge and experience are also under scrutiny. Interviewers want to see a deep understanding of software development principles, how familiar we are with various tools and technologies, and practical experience applying this knowledge in past projects. This assessment often includes questions about system design, architecture, and the technologies we've worked with.

Communication Skills and Cultural Fit

Equally important are our communication skills and our ability to align with the organization's values. We must collaborate effectively with other team members and sometimes clients and clearly express our thoughts. The interviewers want to ensure that we will contribute positively to the team.

Seeking Versatile Candidates

By evaluating developers on these various dimensions, companies seek to identify versatile candidates. It's essential for use to possess not only the right technical skills but also the necessary soft skills and mindset in order to be a good fit for the company's culture and long-term goals.

Embracing Continuous Improvement

It might seem overwhelming for one person to meet all these requirements and it is important to recognize that companies are looking for a combination of skills and qualities rather then expecting perfection in every aspect.
As developers, we would need to continuously work on both technical and soft skills. Highlighting our strengths and being open to learning and growth is crucial.

I’d love to hear from our tech-savvy community! Please share your thoughts in the comments below.

Photo by Lindsey LaMont on Unsplash

Exploring Azure Storage Table

Marius — Wed, 27 Mar 2024 20:02:50 +0000

Azure Table Storage is a NoSQL storage solution provided by Microsoft Azure that is well-suited for a variety of real-world scenarios where scalability, flexibility, and cost-effectiveness are key considerations.

Azure Table Storage uses the key-value model meaning each item stored in the database is stored as a key-value pair. Items are referred to as rows and fields as columns making Azure Table Storage to be seen as a table in a relational database. However, Azure Storage Table allows us to store semi-structured data, this flexibility being particularly useful when there is data that doesn't match to a relational database schema or when you want to avoid the overhead of managing a complex schema. Azure Table Storage tables have no concept of relationships, stored procedures, secondary indexes or foreign keys. The data is available in a single row without requiring that we perform joins across relationships.

Partitioning and partition key

Partitioning is a mechanism for grouping related rows based on a common property or partition key. Rows that share the same partition key will be stored together. If an application adds a new row to a table, Azure ensures that the row is placed in the correct position in the table, with respect to the two elements of the key in Azure Table Storage (partition key and row key).
The partition key identifies the partition containing the row, and the row key uniquely identifies each row in the same partition. Items in the same partition are stored in row key order. This ensures fast access. The partition key can be used in a search criterion to narrow down the volume of data.
We can add as many partitions as we want, they are independent one of each other and they can grow or shrink depending on the number of rows.
In conclusion, partitions are a great way to organize data, improve performance and scalability in Azure Table Storage.

Advantages and disadvantages

There are a few advantages for using Azure Table Storage:

simple to scale
storing semi-structured data
no need to maintain complex relationship
row insertion and data retrieval is fast (when specifying partition)

There are disadvantages in storing data using Azure Table Storage:

difficult to filter and sort non-key data
search could result in scanning the entire table
manage relationships between rows within your application logic (externally)
full transactional support across multiple entities isn't provided, meaning that updates may not be atomic or isolated.

Examples of scenarios

Here are some examples of scenarios where Azure Table Storage can be an appropriate storage solution:

Internet of Things (IoT) Data Storage: it could store vast amounts of IoT device telemetry data
Web Application Data: it could store user profiles, session data, preferences, and other semi-structured data
Logging and Analytics: it could store logs, audit trails, and analytics data generated by applications and services

In summary, Azure Table Storage is a versatile storage solution that can be applied to a wide range of use cases across industries.

Photo by Mika Baumeister on Unsplash

I would love hearing from our tech-savvy community! Have insights, tips, or burning questions? Don't keep them to yourself! Drop your thoughts in the comments below. Let's spark a conversation and learn from each other's experiences in the dynamic world of Azure.

Examples of Cloud Computing Roles

Marius — Sun, 24 Mar 2024 17:09:00 +0000

Here are some cloud computing roles we can land in cloud.

Solutions Architect

This role would be responsible of designing and implementing cloud-based solutions that meet the needs of an organization, considering factors like scalability, security, and cost-effectiveness. A lot of people learning cloud aspire to attain this position.

Cloud Engineer

Basically, having this role you will simply manage the cloud. This position would be responsible for tasks such as provisioning and configuring cloud resources, maintaining system availability and performance, monitoring and troubleshooting issues, and implementing automation and optimization strategies.

Cloud Operations Engineer

This role is seen as an entry level to cloud. It typically focuses on the day-to-day management and maintenance of cloud infrastructure and services.

Sales Engineer

A Sales Engineer is a technical expert who works in conjunction with the sales team to sell complex technical products and services. Their primary role involves understanding the customer's requirements, demonstrating how a product or service can meet those needs, and addressing any technical questions or concerns that may arise during the sales process.

DevOps Engineer

This role would manage tools and services taking into consideration the DevOps culture. They are considered to play a critical role in enabling organizations to deliver high-quality software faster and more efficiently by integrating development and operations processes, automating repetitive tasks, and fostering a culture of collaboration and continuous improvement.

Cloud Support

This role would require to provide technical assistance and support to customers or internal teams for a lot of services in cloud.

I am sure there are a lot more.

Photo by Annie Spratt on Unsplash

Short definition of FaaS

Marius — Fri, 22 Mar 2024 14:57:32 +0000

FaaS is described as a cloud computing service that allows you to execute code in response to events.

FaaS is using a serverless framework. Let's explain that really quickly. A server is a computer or a device that provides programmable service to another computer program. A few server types: database, app or web. So, in the most basic definition possible, serverless computing is a cloud computing execution model where the cloud provider dynamically manages the allocation of machine resources, abstracting away the need for developers to manage servers.

Let's continue with the definition of Faas and rembember that a function in programming is a reusable block of code that performs a specific task or computation. It takes data, process it and responds. This happens the same way using FaaS, developers create code that responds to specific events or triggers, and the cloud provider dynamically allocates resources to execute that code in response to those events. Devs will only need to manage the application and focus only on code, everything else (data, servers, networking, etc.) will be handled by the cloud provider. Developers don't have to deal with managing servers directly nor any other resource.

In Azure we have Azure Functions while Amazon has AWS Lambda and Google has Google Cloud Functions.

Photo by Growtika on Unsplash

Concepts of Cosmos DB: Data Consistency Levels

Marius — Wed, 14 Feb 2024 20:59:36 +0000

Azure Cosmos DB offers multiple levels of data consistency but before going further and exploring these levels, let's review what consistency means in the context distributed NoSQL databases.

In a distributed environment, consistency describes the uniformity of data that is replicated across multiple nodes (sometimes physically separated by thousands of kilometers) for redundancy. In this case, consistency becomes a challenge due to factors such as network latency, node failures, and concurrent updates.
Consistency models in distributed NoSQL databases define clear rules and guarantees regarding how data is read and written across multiple replicas or partitions. These models aim to strike a balance between providing strong guarantees of data consistency and availability and maintaining system performance and scalability.

Levels

Let's take an example of a student database and suppose that we are in Europe and we will update the email address of a student. After the update, someone from North America queries that specific data. What value is he going to see?

Eventual Consistency

Eventual consistency is the weakest form of consistency because a client may read the values that are older than the ones it read in the past. Eventual consistency is ideal where the application doesn't require any ordering guarantees. In our example, there are chances the person from North America will not yet see the new value. Over time, the value will get replicated out to all of the different nodes and replica sets and *eventually * the data will be consistent. Therefore, it is suitable for applications that do not require guaranteed ordering. For example, count of retweets, non-threaded comments, likes, etc.

Consistent Prefix

This level applies to a special case when we need to be sure that the updates are read in order. However, data is not always current. In our case, the person from NA might not see the latest version. The consistent prefix provides read consistency to a specific point in time. It is suitable for the models that can afford the lag but requires high availability with low latency.

Session Consistency

In Azure Cosmos DB, when the consistency level is set to Session it means that the reads will be consistent within a single session. This will guarantee that a client session will read its own writes. It is the default consistency level for Azure Cosmos DB databases and collections and it is suitable for e-commerce applications, social media apps, and applications that require persistent user connections. In our scenario, the person from North America queries the data within the same session, she will see the updated email address. This is because the session maintains consistency, ensuring that all reads within the same session reflect the effects of all previous writes within that session.

Bounded Staleness

This doesn't guarantee we will get the most recent version but it guarantees that we will get a fairly recent version. It trades delays for strong consistency. We can specify maximum lag (time) or maximum lag (operations). In our scenario, they may see the updated email address, provided that the staleness window has not been exceeded. If the staleness window hasn't elapsed since the update, the query from North America will reflect the most recent consistent state of the data, including the updated email address. However, if the staleness window has passed since the update, and the lag in replication from Europe to North America exceeds this staleness window, the query from North America might not reflect the most recent update. Instead, it may reflect a state of the data that is consistent within the configured staleness window, but not necessarily including the most recent update from Europe.

Strong Consistency

This level guarantees reads to the most recent committed version of the item. The strong consistency is suitable for applications that cannot tolerate any data loss due to downtime. In our scenario, if we update the email address of a student in Europe and then someone from North America queries the data with a consistency level set to Strong, they will definitely see the updated email address. This is because this consistency level ensures that reads always reflect the most recent write, regardless of the location of the reader or the writer. Therefore, the user in North America querying the data with Strong consistency will see the updated email address of the student, regardless of where the update was made.

The consistency levels in Azure Cosmos DB determine how quickly we can read data that has been written or inserted into the database. In other words, when we write data to Cosmos DB, the consistency level determines how long it will take for that data to be readable by other operations while providing a trade-off between consistency and performance.

Photo by Shubham's Web3 on Unsplash

P.S.

Concepts of Cosmos DB: Throughput

Marius — Mon, 12 Feb 2024 09:31:47 +0000

In Azure Cosmos DB, throughput is an important concept that defines the capacity and performance of your database operations. It helps you mange the RUs and, essentially, it determines how much data can be read from or written to your Cosmos DB account within a given unit of time, typically measured in requests per second (RUs).
Before continuing reading this article, I suggest revisiting Concepts of Cosmos DB: Request Unit

Cosmos DB offers different models for managing throughput.

Provisioned throughput

Standard

You can set a certain amount of RU and it can be configured at the database or container level but you can combine the two models.
Provisioned throughput is great for databases that receive sustained traffic. For example, if you decide to configure the database at 5000 RU/s, it is guaranteed that this amount will be available, but once the RU's are consumed, future requests will be rate limited because by default, it is required a manual scaling to acquire more RU's.

Autoscale

This is when autoscaling can be used. You can specify a maximum RU throughout amount and Cosmos DB will ensure that the data will be available. The minimum throughout will be calculated as 10% of the maximum. So, following the above example, if you set the maximum to 5000 RU/s, Azure Cosmos DB will automatically scale your provisioned throughput up and down to accommodate the requirements of the application within this range (at least 500 RU/s - maximum 5000 RU/s). This applies best when the application have unpredictable traffic patterns.

Next, we will see how can we accommodate to a type of workload that have times when it will be not used at all (unused capacity).

Serverless throughput

When choosing serverless we will pay only for the RU's that are consumed and the used storage.
This can be view as a pure consumption model. It applies to development/testing activities or if you just want to get starting with Azure Cosmos DB. Taking into the consideration that the cost become less predictable you can monitor the consumption via a chart.

Despite these options, in some situations, it may be unclear whether provisioned throughput or serverless should be chosen for a given workload. In this case, Microsoft recommends us to estimate the overall expected consumption, or the total number of RUs we may consume over a month.

By effectively managing throughput in Cosmos DB, you can ensure that your applications meet performance requirements while also optimizing costs based on actual usage patterns.

Photo by Shubham's Web3 on Unsplash

P.S.

Azure Developer Certification (AZ-204): A practice Scenario #6

Marius — Sun, 11 Feb 2024 12:38:39 +0000

In this blog article, we explore a practical scenario where a company needs to find a solution to reduce the latency between their VMs.

Your company has multiple VMs deployed in a VNet. All the VMs are running on Windows. The company needs high-speed communication to analyze shared streaming data between different virtual machines. Currently, they experience high lag between the VMs and they want to reduce it.

Which solution could reduce the latency between VMs?

Photo by Ben Mullins on Unsplash

P.S.

Azure Developer Certification (AZ-204): A practice Scenario #5

Marius — Thu, 08 Feb 2024 09:28:18 +0000

In this blog article, we explore a practical scenario where a developer needs to follow the correct steps for deploying a web application.

Your company is deploying a new web application using App Service and you will be responsible of it. You want to use the Azure CLI and you will have at your disposal a brand new (empty) account.

What are the steps you will need to follow?

Photo by Ben Mullins on Unsplash

P.S.

Concepts of Cosmos DB: Request Unit

Marius — Mon, 05 Feb 2024 13:32:25 +0000

After the introduction of Cosmos DB and understanding the various way of scaling the "traditional" databases I would like to continue writing about Request Unit, a fundamental topic for pricing and capacity planning in Cosmos DB.

What is a Request Unit (RU)?

According to Microsoft, a Request Unit is a performance currency abstracting the system resources such as CPU, IOPS, and memory that are required to perform the database operations supported by Azure Cosmos DB. Whether the database operation is a write, point read, or query, operations are always measured in RUs.

In simpler terms, RU in Azure Cosmos DB is a measure of the amount of resources required to perform operations on the database. It is a unit that combines the cost of reading and writing data, as well as other database operations like indexing. This is how Azure Cosmos DB charges for database operations. Microsoft defines 1 RU = 1 point read of 1kb.

So, when you make a request to the Cosmos DB database, the number of Request Units needed depends on the complexity and amount of data involved in that operation. For example, a simple read operation might cost fewer Request Units than a complex query or a write operation that involves a lot of data.

Monitoring and Optimization

The Azure Portal provides tools for monitoring the consumption of Request Units for your Cosmos DB account. You could use Azure Monitor Metrics or Diagnostic Logs. Efficient monitoring of these resources guarantees availability, performance (identify and address bottlenecks and inefficiencies in real time) and functionality. Monitoring can be used to optimize queries, indexes, and data models for better performance and cost efficiency (resources are allocated efficiently, reducing unnecessary costs).

In summary, Request Units are a way to quantify the cost of operations in Azure Cosmos DB, helping you understand and manage the resource consumption of your database.

In the next article, I will cover another import concept of Cosmos DB: Provisioned Throughput. You can provision a certain amount of RU/s (Request Units per second) for a container based on your anticipated workload.

Photo by Shubham's Web3 on Unsplash

P.S.

Azure Developer Certification (AZ-204): A practice Scenario #4

Marius — Sat, 03 Feb 2024 07:25:29 +0000

In this blog article, we explore a practical scenario where a developer is feeling anxious about moving a process from a VM to an Azure function.

Your company runs a e-learning platform. Currently, they perform multiple actions on a VM when an user is added. You want to move this to an Azure function but you are afraid of receiving a potential timeout.

Could this be solved using Durable Functions?

Photo by Ben Mullins on Unsplash

P.S.