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.
Ideal if the company is already a Microsoft shop, so chances are that it is free with SQL Server. Also, good for moving data between on premise systems. Not ideal for moving data to the cloud. No functionality out of the box to work with REST APIs. Stable product but definitely not the future
Connection managers for online data sources can be tricky to configure.
Performance tuning is an art form and trialing different data flow task options can be cumbersome. SSIS can do a better job of providing performance data including historical for monitoring.
Mapping destination using OLE DB command is difficult as destination columns are unnamed.
Excel or flat file connections are limited by version and type.
Some features should be revised or improved, some tools (using it with Visual Studio) of the toolbox should be less schematic and somewhat more flexible. Using for example, the CSV data import is still very old-fashioned and if the data format changes it requires a bit of manual labor to accept the new data structure
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
SSIS is a great tool for most ETL needs. It has the 90% (or more) use cases covered and even in many of the use cases where it is not ideal SSIS can be extended via a .NET language to do the job well in a supportable way for almost any performance workload.
SQL Server Integration Services performance is dependent directly upon the resources provided to the system. In our environment, we allocated 6 nodes of 4 CPUs, 64GB each, running in parallel. Unfortunately, we had to ramp-up to such a robust environment to get the performance to where we needed it. Most of the reports are completed in a reasonable timeframe. However, in the case of slow running reports, it is often difficult if not impossible to cancel the report without killing the report instance or stopping the service.
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.
The support, when necessary, is excellent. But beyond that, it is very rarely necessary because the user community is so large, vibrant and knowledgable, a simple Google query or forum question can answer almost everything you want to know. You can also get prewritten script tasks with a variety of functionality that saves a lot of time.
The implementation may be different in each case, it is important to properly analyze all the existing infrastructure to understand the kind of work needed, the type of software used and the compatibility between these, the features that you want to exploit, to understand what is possible and which ones require integration with third-party tools
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.
I had nothing to do with the choice or install. I assume it was made because it's easy to integrate with our SQL Server environment and free. I'm not sure of any other enterprise level solution that would solve this problem, but I would likely have approached it with traditional scripting. Comparably free, but my own familiarity with trad scripts would be my final deciding factor. Perhaps with some further training on SSIS I would have a different answer.
