Apache Airflow vs. HashiCorp Nomad

For a quick job scanning of status and deep-diving into job issues, details, and flows, AirFlow does a good job. No fuss, no muss. The low learning curve as the UI is very straightforward, and navigating it will be familiar after spending some time using it. Our requirements are pretty simple. Job scheduler, workflows, and monitoring. The jobs we run are >100, but still is a lot to review and troubleshoot when jobs don't run. So when managing large jobs, AirFlow dated UI can be a bit of a drawback.

Incentivized

Nomad is well suited for organizations who wish to tackle the problem of cloud computing with as little opinion as possible. Where competing tools like Kubernetes limit the concept of "batteries included," Nomad relies on engineers understanding the missing components and filling them in as necessary. The benefit of Nomad is the ability to build a system out of small pieces with the cost of having more complexity at a system level compared to alternatives.

Incentivized

• Apache Airflow is one of the best Orchestration platforms and a go-to scheduler for teams building a data platform or pipelines.
• Apache Airflow supports multiple operators, such as the Databricks, Spark, and Python operators. All of these provide us with functionality to implement any business logic.
• Apache Airflow is highly scalable, and we can run a large number of DAGs with ease. It provided HA and replication for workers. Maintaining airflow deployments is very easy, even for smaller teams, and we also get lots of metrics for observability.

Incentivized

• Nomad is incredibly simple by nature, following the Linux philosophy of doing one thing great. That one thing for Nomad is job scheduling.
• Nomad is a modern tool, written in Go with a large community and maintained by HashiCorp.
• Implementation of Nomad is very simple since it is a single binary.

Incentivized

• A local "dry run" or IDE plugin that can validate and simulate DAG execution without needing a full environment.
• Better feedback on DAG parse errors in the UI or CLI.
• Navigating large DAGs with hundreds of tasks can be slow and hard to understand visually.

Incentivized

• Nomad only handles one part of a full platform. Expertise and vision are required in implementing an entire system that is functional enough for an organization to rely on. This includes other tools to handle things like secrets, service discovery, network routing, etc.
• Nomad is delayed in some modern functionality, like features for service-mesh and open tracing. These features are on the tool's roadmap, but there's currently no native support. These paradigms can be established still, but require more expertise outside of Nomad itself.
• Nomad is not the leading tool for this space, and as such risks being left behind by tools with much greater support, such as Kubernetes.

Incentivized

For its capability to connect with multicloud environments. Access Control management is something that we don't get in all the schedulers and orchestrators. But although it provides so many flexibility and options to due to python , some level of knowledge of python is needed to be able to build workflows.

Incentivized

Apache Airflow is suited for a much wider set of use cases compared to Databricks. You can run it anywhere, and there is also no vendor lock-in. With Airflow, we can utilize almost any compute engine. Same thing we want to do with Databricks. There might be some level of difficulty based on the support.

Incentivized

Nomad's primary competitor is Kubernetes, specifically its scheduling component. Kubernetes is a much more complete system that will handle more things than job scheduling, including service discovery, secrets management, and service routing. There also exists a much larger community support for Kubernetes vs Nomad. One might say Kubernetes is the safer choice between the two. Kubernetes is the complete "operating system" for cloud computing, but with it includes complexities that are "Kubernetes" specific. The decision really comes down to a mindset of monolith vs components. With Kubernetes, I would argue you choose the entire system as a whole. With Nomad, you design your system piece by piece. There is no wrong answer.

Incentivized

• Most of the ETL processes were automated, cutting down on human labor.
• Apache Airflow's user interface (UI) was very informative and straightforward.
• Since ETL processes were providing data via airflow, we were able to gain a deeper comprehension of the data at hand.

Incentivized

• Nomad has allowed our organization to deploy quicker and more frequently with a lower failure rate.
• Nomad has brought in consistency from an operations perspective.
• Nomad's performance allows us to scale infinitely while providing functionality that reduces mean time to repair (canary deploys, versioning, rollbacks, etc).

Incentivized