Spark has primarily replaced my use of writing pure Hadoop MapReduce or Apache Pig jobs for processing data. I like the fact that I can alternate between the main programming languages that I know - Java and Python - and use those to learn the Scala API. Spark also can be …
Sqoop comes preinstalled on the major Hadoop vendor distributions as the recommended product to import data from relational databases. The ability to extend it with additional JDBC drivers makes it very flexible for the environment it is installed within.
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.
Sqoop is great for sending data between a JDBC compliant database and a Hadoop environment. Sqoop is built for those who need a few simple CLI options to import a selection of database tables into Hadoop, do large dataset analysis that could not commonly be done with that database system due to resource constraints, then export the results back into that database (or another). Sqoop falls short when there needs to be some extra, customized processing between database extract, and Hadoop loading, in which case Apache Spark's JDBC utilities might be preferred
Sqoop2 development seems to have stalled. I have set it up outside of a Cloudera CDH installation, and I actually prefer it's "Sqoop Server" model better than just the CLI client version that is Sqoop1. This works especially well in a microservices environment, where there would be only one place to maintain the JDBC drivers to use for Sqoop.
The only thing I dislike about spark's usability is the learning curve, there are many actions and transformations, however, its wide-range of uses for ETL processing, facility to integrate and it's multi-language support make this library a powerhouse for your data science solutions. It has especially aided us with its lightning-fast processing times.
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.
All the above systems work quite well on big data transformations whereas Spark really shines with its bigger API support and its ability to read from and write to multiple data sources. Using Spark one can easily switch between declarative versus imperative versus functional type programming easily based on the situation. Also it doesn't need special data ingestion or indexing pre-processing like Presto. Combining it with Jupyter Notebooks (https://github.com/jupyter-incubator/sparkmagic), one can develop the Spark code in an interactive manner in Scala or Python
Sqoop comes preinstalled on the major Hadoop vendor distributions as the recommended product to import data from relational databases. The ability to extend it with additional JDBC drivers makes it very flexible for the environment it is installed within.
Spark also has a useful JDBC reader, and can manipulate data in more ways than Sqoop, and also upload to many other systems than just Hadoop.
Kafka Connect JDBC is more for streaming database updates using tools such as Oracle GoldenGate or Debezium.
Streamsets and Apache NiFi both provide a more "flow based programming" approach to graphically laying out connectors between various systems, including JDBC and Hadoop.
Faster turn around on feature development, we have seen a noticeable improvement in our agile development since using Spark.
Easy adoption, having multiple departments use the same underlying technology even if the use cases are very different allows for more commonality amongst applications which definitely makes the operations team happy.
Performance, we have been able to make some applications run over 20x faster since switching to Spark. This has saved us time, headaches, and operating costs.
When combined with Cloudera's HUE, it can enable non-technical users to easily import relational data into Hadoop.
Being able to manipulate large datasets in Hadoop, and them load them into a type of "materialized view" in an external database system has yielded great insights into the Hadoop datalake without continuously running large batch jobs.
Sqoop isn't very user-friendly for those uncomfortable with a CLI.
