What Is Apache Kafka and How Do You Monitor It?

What Is Apache Kafka and How Do You Monitor It?

Apache Kafka is known for its ability to handle real-time streaming data with speed and efficiency. It’s also known for being scalable and durable, which makes it ideal for complex, enterprise-grade applications. Of course, those new to the concept behind Kafka may find that it takes some time to understand how it works. 

Thanks to its unique combination of messaging, storage, and stream processing features, Kafka is well suited for both real-time and historical data analysis. So, let’s dive into what you need to know about this platform and the process of monitoring it. 


What Is Apache Kafka?

Apache Kafka is a type of distributed data store, but what makes it unique is that it’s optimized for real-time streaming data. Streaming data refers to data that is being simultaneously and constantly generated by multiple (e.g., thousands) data sources at once. A special platform like Apache Kafka is necessary to handle these massive streams of data and process them efficiently.

Due to its ability to efficiently handle real-time streaming data, Apache Kafka is the perfect underlying infrastructure for pipelines and applications that deal with this kind of data. Many businesses also use Apache Kafka as a message broker platform to help applications communicate with each other. 

How Does Apache Kafka Work?

A crucial element that sets Kafka apart from the rest is how it has stitched together two messaging models to create its partitioned log model. The partitioned log model used by Kafka combines the best of two models: queuing and publish-subscribe.

Queuing is a widely used model because it allows for multiple consumer instances to handle data processing, creating a distributed solution. However, there can only be one subscriber for a traditional queue. Meanwhile, the publish-subscribe model offers a multi-subscriber solution, but it does not allow for work distribution because all subscribers get all messages.

To help solve these downsides, Kafka stitched these models together. In the partitioned log model used by Kafka, a log represents an orderly sequence of records, which can be partitioned to allow for certain records to go straight to certain subscribers. In other words, Kafka’s model allows for a multi-subscriber design but improves scalability by allowing for logs to be segmented or partitioned to distribute work more effectively. 

Additionally, Kafka’s model also creates replayability, which allows applications to work independently of one another as they read the streaming data, each working at its own rate without missing information that’s already been processed by another app. 

Apache Kafka offers a unique solution thanks to its partitioned log model that combines the best of traditional queues with the best of the publish-subscribe model. Additionally, it’s one of the few data storage solutions on the market that’s able to handle real-time streaming data with such efficiency.

Overall, there are three advantages that make Kafka so popular, and those are its speed, scalability, and durability. By decoupling data streams, Kafka creates an extremely fast solution with very low latency. Additionally, its unique model allows users to distribute workloads across multiple servers, which makes it immensely scalable.

Lastly, the partitioning method employed by Kafka allows for distributable and replicable work, and since all data is written to disk, Kafka provides protection against server failure, making it a highly durable, fault-tolerant solution. 

Apache Kafka Use Cases

Kafka’s features offer countless benefits for businesses working with real-time streaming data and/or massive amounts of historical data. However, there are some instances when you might not want to choose Kafka. Here’s a look at when you should use Kafka along with some circumstances when you should consider looking elsewhere. 

When To Use

Thanks to its versatile set of features, there are many use cases for Apache Kafka, including:

  • Messaging: Message brokers have a number of use cases in and of themselves, and if you’re looking for a reliable messaging solution, it’s hard to beat the throughput, replication, partitioning, and fault-tolerance of Kafka’s highly capable platform. 
  • Activity Tracking: By creating a topic for each type of activity, such as website page views or searches, Kafka is highly capable of rebuilding a pipeline to which multiple processing, monitoring, and storage apps can subscribe as needed. 
  • Metrics: With its ability to aggregate statistics from distributed applications, Kafka is great for operational monitoring, as it’s able to produce a centralized feed of multiple data sources. 
  • Log Aggregation: Kafka offers a lower-latency alternative to traditional log aggregation by abstracting log and event data into a stream of messages. Kafka also benefits from improved durability thanks to replication. 
  • Stream Processing: The ability to process raw streaming data with Kafka and aggregate, enrich, or transform it using custom data pipelines is one of the most valuable use cases for the Kafka platform. 
  • Event Sourcing: When designing an application using event sourcing, businesses require infrastructure able to support very large amounts of stored data, which makes Kafka the perfect choice. 
  • Commit Log: Distributed systems that require external commit logs could make use of Kafka thanks to its replication and log compaction features.  

What Apache Kafka Can’t Do

In certain circumstances, you might want to avoid Apache Kafka, such as when applied to:

  • Internet of Things: Kafka is considered a real-time solution for IT projects, but it does not offer the same real-time reliability necessary for internet of things (IoT) solutions. Instead, those projects require a hard real-time solution, which means no latency, no spikes, and a deterministic network. Kafka simply doesn’t meet those requirements.
  • Safety-Related Data: Just as Kafka does not work with IoT devices because it lacks a zero latency, deterministic network, it cannot be used for safety-related data, either. This would include modern applications like robotics and vehicle safety systems.
  • Function Like a Blockchain: While the distributed log model utilized by Kafka is similar to the concept of blockchain architecture, Kafka should not be used to replace the blockchain. With that said, Kafka is often the simpler and better-suited alternative when the “trustless” features of a blockchain aren’t necessary. 
  • Replace Other Databases: While Kafka is a database, it is rarely suitable to replace other databases your business may rely on. If you’re trying to replace a database, it’s a good idea to take a few steps back and understand what you need versus what Kafka is intended to do.

How to Monitor Kafka

Given the high-volume workloads that most Kafka users will have on their hands, monitoring Kafka to keep tabs on performance (and continuously improve it) is crucial to ensuring long-term useability and reliability. With that said, there are a handful of metrics you should focus on, such as:

  • How many messages are flowing in and out? What is the in and out rate for the host network?
  • What is the idle time for the network handler, request handler, and CPU?
  • Are there under-replicated partitions? What is the leader election rate? 
  • Which consumers are lagging behind?

While your exact use case and requirements will change how you monitor Kafka, this list provides a good starting point to get you going if you are unsure about which metrics you should look to measure and track over time. 

Aside from establishing baselines and watching when things deviate, which can alert you to new bottlenecks and other emerging issues, monitoring can also help you continuously improve performance by using the information to optimize your Kafka environment and understand how the changes you make impact it. 

Apache Kafka + Microservices

Microservices architecture is being widely implemented across the world of business thanks to its ability to help break down monoliths and steer development teams in the direction of simple, independent features or “services.” The biggest benefit of microservices is that each service can be bundled up with others to create different applications and solutions, all while independent features can be removed or updated without dependencies on each other.

The scalability and reusability of microservices are undeniable, but when it comes to actually executing microservices architecture, one of the most crucial design decisions is deciding whether services should communicate directly with each other or if a message broker should act as the middleman. The latter is often considered more flexible, and it offers a level of failure resistance.

Of all of the businesses that choose to use a message broker as an intermediary in their microservices architecture, many will turn to Kafka to help them fill that role. This is because Apache Kafka is an obvious choice thanks to its distributed partitioned log model and its unique messaging features that help it work more efficiently. 

Here are some reasons why you might choose Kafka for this purpose:

  • Kafka easily connects with other systems, helping you integrate it into your existing environment with ease. Kafka can be used to transport some or all of your data and to create backward compatibility with legacy systems.
  • Kafka provides a centralized management system to control who can access various types of data. This is excellent for data governance and compliance standards, and it helps to simplify the burden of securing your data.
  • Kafka’s clustered design helps provide fault tolerance and scalability, both of which are crucial functions of microservices environments. When the number of consumers changes or the number of messages increases, Kafka can rebalance the load automatically, which is essential to maintaining uptime and performance.
  • Kafka’s ability to process any type of data makes it highly flexible for a microservices environment. Choosing another messaging solution could pose limitations in the future if you begin to work with other types of data that the solution doesn’t support.

All in all, Kafka is considered a highly powerful solution for use in microservices environments. Of course, choosing a messaging solution is far from the only step in designing microservices architecture. It is critical for you to consider all of the complexities that come along with it and decide if it’s the right way forward for your business. 

Apache Kafka + Kubernetes

Kafka is known for its flexibility, but Kubernetes promises to maximize that flexibility by providing a container management system to help automate the deployment, scalability, and operation of containers. Kafka and Kubernetes together offer a powerful solution for cloud-native development projects by providing a distributed, independent service with loose coupling and highly scalable infrastructure.

By far, the biggest benefit of choosing Kubernetes for your Apache Kafka installation is the ability to achieve infrastructure abstraction. Since you can configure things once and then run it anywhere, Kubernetes allows assets to be pooled together to better allocate resources while providing a single environment for ops teams to easily manage all of their instances.

If you are considering using Kubernetes to run Kafka, it’s important to understand how it works. To put it simply, Kafka will run as a cluster of brokers, which you can deploy on Kubernetes using different nodes. Kubernetes can then recover nodes as needed, helping to ensure optimal resource utilization. This approach also supports the fault-tolerance that Kafka is known for. 


Apache Kafka is a flexible solution for businesses seeking a platform to help process real-time streaming data with grace. The fault-tolerance, distribution, and replication features offered by Kafka make it suitable for a variety of use cases. Plus, it can even work as the messaging solution for your microservices architecture, providing you with a solid backing for pursuing a new approach to development and business offerings.

With all of those things in mind, there are instances where Apache Kafka simply isn’t suitable. For example, when working with IoT devices, safety-related data, or any instance where you need a truly zero-latency, hard real-time solution, you should look elsewhere as that simply isn’t what Kafka is built to do.
Given that information, now is a good time to explore all that Kafka offers and see if you can find examples of your unique use case. Head to the Kafka project website for more information.