beam
Scio
Our great sponsors
beam | Scio | |
---|---|---|
30 | 7 | |
7,508 | 2,520 | |
1.5% | 0.4% | |
10.0 | 9.6 | |
5 days ago | 4 days ago | |
Java | Scala | |
Apache License 2.0 | Apache License 2.0 |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
beam
-
Ask HN: Does (or why does) anyone use MapReduce anymore?
The "streaming systems" book answers your question and more: https://www.oreilly.com/library/view/streaming-systems/97814.... It gives you a history of how batch processing started with MapReduce, and how attempts at scaling by moving towards streaming systems gave us all the subsequent frameworks (Spark, Beam, etc.).
As for the framework called MapReduce, it isn't used much, but its descendant https://beam.apache.org very much is. Nowadays people often use "map reduce" as a shorthand for whatever batch processing system they're building on top of.
-
beam VS quix-streams - a user suggested alternative
2 projects | 7 Dec 2023
-
How do Streaming Aggregation Pipelines work?
Apache Beam is one of many tools that you can use
-
Releasing Temporian, a Python library for processing temporal data, built together with Google
Flexible runtime ☁️: Temporian programs can run seamlessly in-process in Python, on large datasets using Apache Beam.
-
Kafka cluster loses or duplicates messages
To perform the tests I'm using a Kafka cluster on Kubernetes from the Beam repo (here).
- Apache Beam
-
Real Time Data Infra Stack
Apache Beam: Streaming framework which can be run on several runner such as Apache Flink and GCP Dataflow
-
Google Cloud Reference
Apache Beam: Batch/streaming data processing 🔗Link
-
Composer out of resources - "INFO Task exited with return code Negsignal.SIGKILL"
What you are looking for is Dataflow. It can be a bit tricky to wrap your head around at first, but I highly suggest leaning into this technology for most of your data engineering needs. It's based on the open source Apache Beam framework that originated at Google. We use an internal version of this system at Google for virtually all of our pipeline tasks, from a few GB, to Exabyte scale systems -- it can do it all.
-
Pub/Sub parallel processing best practices
That being said, there is a learning curve in understanding how Apache Beam works. Take a look at the beam website for more information.
Scio
- Are there any openly available data engineering projects using Scala and Spark which follow industry conventions like proper folder/package structures and object oriented division of classes/concerns? Most examples I’ve seen have everything in one file without proper separation of concerns.
-
For the DE's that choose Java over Python in new projects, why?
I doubt it is possible because I suspect that GIL would like a word. So I could spend nights trying to make it work in Python (and possibly, if not likely, fail). Or I could just use this ready made solution.
-
what popular companies uses Scala?
Apache Beam API called Scio. They open sourced it https://spotify.github.io/scio/
-
Scala or Python
Generally Python is a lingua franca. I have never met a data engineer that doesn't know Python. Scala isn't used everywhere. Also, you should know that in Apache Beam (data processing framework that's gaining popularity because it can handle both streaming and batch processing and runs on spark) the language choices are Java, Python, Go and Scala. So, even if you "only" know Java, you can get started with Data engineering through apache beam.
-
Wanting to move away from SQL
I agree 100%. I haven't used SQL that much in previous data engineering roles, and I refuse to consider jobs that mostly deal with SQL. One of my roles involved using a nice Scala API for apache beam called Scio and it was great. Code was easy to write, maintain, and test. It also worked well with other services like PubSub and BigTable.
-
ETL Pipelines with Airflow: The Good, the Bad and the Ugly
If you prefer Scala, then you can try Scio: https://github.com/spotify/scio.
-
ELT, Data Pipeline
To counter the above mentioned problem, we decided to move our data to a Pub/Sub based stream model, where we would continue to push data as it arrives. As fluentd is the primary tool being used in all our servers to gather data, rather than replacing it we leveraged its plugin architecture to use a plugin to stream data into a sink of our choosing. Initially our inclination was towards Google PubSub and Google Dataflow as our Data Scientists/Engineers use Big Query extensively and keeping the data in the same Cloud made sense. The inspiration of using these tools came from Spotify’s Event Delivery – The Road to the Cloud. We did the setup on one of our staging server with Google PubSub and Dataflow. Both didn't really work out for us as PubSub model requires a Subscriber to be available for the Topic a Publisher streams messages to, otherwise the messages are not stored. On top of it there was no way to see which messages are arriving. During this the weirdest thing that we encountered was that the Topic would be orphaned losing the subscribers when working with Dataflow. PubSub we might have managed to live with, the wall in our path was Dataflow. We started off with using SCIO from Spotify to work with Dataflow, there is a considerate lack of documentation over it and found the community to be very reserved on Github, something quite evident in the world of Scala for which they came up with a Code of Conduct for its user base to follow. Something that was required from Dataflow for us was to support batch write option to GCS, after trying our hand at Dataflow to no success to achieve that, Google's staff at StackOverflow were quite responsive and their response confirmed that it was something not available with Dataflow and streaming data to BigQuery, Datastore or Bigtable as a datastore was an option to use. The reason we didn't do that was to avoid high streaming cost to these services to store data, as majority of our jobs from the data team are based on batched hourly data. The initial proposal to the updated pipeline is shown below.
What are some alternatives?
Apache Arrow - Apache Arrow is a multi-language toolbox for accelerated data interchange and in-memory processing
Apache Spark - Apache Spark - A unified analytics engine for large-scale data processing
Apache Hadoop - Apache Hadoop
Apache Flink - Apache Flink
Apache Kafka - Mirror of Apache Kafka
Airflow - Apache Airflow - A platform to programmatically author, schedule, and monitor workflows
Reactive-kafka - Alpakka Kafka connector - Alpakka is a Reactive Enterprise Integration library for Java and Scala, based on Reactive Streams and Akka.
Apache Hive - Apache Hive
metorikku - A simplified, lightweight ETL Framework based on Apache Spark
Apache Accumulo - Apache Accumulo
Scoobi - A Scala productivity framework for Hadoop.