Well suited: To most of the local run of datasets and non-prod systems - scalability is not a problem at all. Including data from multiple types of data sources is an added advantage. MLlib is a decently nice built-in library that can be used for most of the ML tasks. Less appropriate: We had to work on a RecSys where the music dataset that we used was around 300+Gb in size. We faced memory-based issues. Few times we also got memory errors. Also the MLlib library does not have support for advanced analytics and deep-learning frameworks support. Understanding the internals of the working of Apache Spark for beginners is highly not possible.
If you want to stream high volumes of data, be it for ETL streaming or event sourcing, Google Cloud Pub/Sub is your go-to tool. It's easy to learn, easy to observe its metrics and scales with ease without additional configuration so if you have more producers of consumers, all you need to do is to deploy on k8s your solutions so that you can perform autoscaling on your pods to adjust to the data volume. The DLQ is also very transparent and easy to configure. Your code will have no logic whatsoever regarding orchestrating pubsub, you just plug and play. However, if you are not in the Google Cloud Pub/Sub environment, you might have trouble or be most likely unable to use it since I think it's a product of Google Cloud.
With a pub/sub architecture the consumer is decoupled in time from the publisher i.e. if the consumer goes down, it can replay any events that occurred during its downtime.
It also allows consumer to throttle and batch incoming data providing much needed flexibility while working with multiple types of data sources
A simple and easy to use UI on cloud console for setup and debugging
It enables event-driven architectures and asynchronous parallel processing, while improving performance, reliability and scalability
It serves all of our purposes in the most transparent way I can imagine, after seeing other message queueing providers, I can only attest to its quality.
If the team looking to use Apache Spark is not used to debug and tweak settings for jobs to ensure maximum optimizations, it can be frustrating. However, the documentation and the support of the community on the internet can help resolve most issues. Moreover, it is highly configurable and it integrates with different tools (eg: it can be used by dbt core), which increase the scenarios where it can be used
It has many libraries in many languages, google provides either good guides or they're AI generated code libraries that are easy to understand. It has very good observability too.
1. It integrates very well with scala or python. 2. It's very easy to understand SQL interoperability. 3. Apache is way faster than the other competitive technologies. 4. The support from the Apache community is very huge for Spark. 5. Execution times are faster as compared to others. 6. There are a large number of forums available for Apache Spark. 7. The code availability for Apache Spark is simpler and easy to gain access to. 8. Many organizations use Apache Spark, so many solutions are available for existing applications.
They have decent documentation, but you need to pay for support. We weren't able to answer all our questions with the documentation and didn't have time to setup support before we needed it so I can't give it a higher rating but I think it tends to be a bit slow unless you're a GCP enterprise support customer.
Spark in comparison to similar technologies ends up being a one stop shop. You can achieve so much with this one framework instead of having to stitch and weave multiple technologies from the Hadoop stack, all while getting incredibility performance, minimal boilerplate, and getting the ability to write your application in the language of your choosing.
Having used Amazon Web Services SNS & SQS I can say that even if the latter may offer more features, Google Cloud Pub/Sub is easier to use. On the other hand, usage of SNS & SQS as well as documentation and troubleshooting is easier with the AWS solution. Since we are not using GCP only for Pub/Sub the choice depends on other variables.
You can just plug in consumers at will and it will respond, there's no need for further configuration or introducing new concepts. You have a queue, if it's slow, you plug in more consumers to process more messages: simple as that.
