Re: Open sourcing Spindle by Adobe Research, a web analytics processing engine in Scala, Spark, and Parquet.
\cc David Tompkins and Jim Donahue if they have anything to add. \cc My school email. Please include bamos_cmu.edu for further discussion. Hi Deb, Debasish Das wrote Looks very cool...will try it out for ad-hoc analysis of our datasets and provide more feedback... Could you please give bit more details about the differences of Spindle architecture compared to Hue + Spark integration (python stack) and Ooyala Jobserver ? Does Spindle allow sharing of spark context over multiple spark jobs like jobserver ? Great point, I think these jobservers would work well with Spindle on larger clusters. I've added the following portion to the README to mention this as an area of future work. Regards, Brandon. --- ## Future Work - Utilizing Spark job servers or resource managers. Spindle's architecture can likely be improved on larger clusters by utilizing a job server or resource manager to maintain a pool of Spark contexts for query execution. [Ooyala's spark-jobserver][spark-jobserver] provides a RESTful interface for submitting Spark jobs that Spindle could interface with instead of interfacing with Spark directly. [YARN][yarn] can also be used to manage Spark's resources on a cluster, as described in [this article][spark-yarn]. However, allocating resources on the cluster raises additional questions and engineering work that Spindle can address in future work. Spindle's current architecture coincides HDFS and Spark workers on the same nodes, minimizing the network traffic required to load data. How much will the performance degrade if the resource manager allocates some subset of Spark workers that don't coincide with any of the HDFS data being accessed? Furthermore, how would a production-ready caching policy on a pool of Spark Contexts look? What if many queries are being submitted and executed on different Spark Contexts that use the same data? Scheduling the queries on the same Spark Context and caching the data between query executions would substantially increase the performance, but how should the scheduler be informed of this information? [spark-jobserver]: https://github.com/ooyala/spark-jobserver [yarn]: http://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/YARN.html [spark-yarn]: http://blog.cloudera.com/blog/2014/05/apache-spark-resource-management-and-yarn-app-models/ -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Open-sourcing-Spindle-by-Adobe-Research-a-web-analytics-processing-engine-in-Scala-Spark-and-Parquet-tp12203p12731.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
Re: Open sourcing Spindle by Adobe Research, a web analytics processing engine in Scala, Spark, and Parquet.
\cc David Tompkins and Jim Donahue if they have anything to add. \cc My school email. Please include bamos_cmu.edu for further discussion. Hi Soumya, ssimanta wrote The project mentions - process petabytes of data in real-time. I'm curious to know if the architecture implemented in the Github repo was used to process petabytes? If yes, how many nodes did you use for this and did you use Spark standalone cluster or with YARN/Mesos ? I'm also interested to know what issues you had with Spray and Akka working at this scale. Great question, I've added the following portion to the README's intro section to make it clear that Spindle is not yet ready for processing petabytes of data in real-time. Also, I'd be interested in seeing how Spray/Akka at larger scales compares to using job or resource managers. We're currently running Spindle on the standalone cluster. Regards, Brandon. --- This repo contains the Spindle implementation and benchmarking scripts to observe Spindle's performance while exploring Spark's tuning options. Spindle's goal is to process petabytes of data on thousands of nodes, but the current implementation has not yet been tested at this scale. Our current experimental results use six nodes, each with 24 cores and 21g of Spark memory, to query 13.1GB of analytics data. The trends show that further Spark tuning and optimizations should be investigated before attempting larger scale deployments. -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Open-sourcing-Spindle-by-Adobe-Research-a-web-analytics-processing-engine-in-Scala-Spark-and-Parquet-tp12203p12706.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
Re: Open sourcing Spindle by Adobe Research, a web analytics processing engine in Scala, Spark, and Parquet.
Thanks for sharing this, Brandon! Looks like a great architecture for people to build on. Matei On August 15, 2014 at 2:07:06 PM, Brandon Amos (a...@adobe.com) wrote: Hi Spark community, At Adobe Research, we're happy to open source a prototype technology called Spindle we've been developing over the past few months for processing analytics queries with Spark. Please take a look at the repository on GitHub at https://github.com/adobe-research/spindle, and we welcome any feedback. Thanks! Regards, Brandon. -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Open-sourcing-Spindle-by-Adobe-Research-a-web-analytics-processing-engine-in-Scala-Spark-and-Parquet-tp12203.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org
Re: Open sourcing Spindle by Adobe Research, a web analytics processing engine in Scala, Spark, and Parquet.
Hi Brandon, Looks very cool...will try it out for ad-hoc analysis of our datasets and provide more feedback... Could you please give bit more details about the differences of Spindle architecture compared to Hue + Spark integration (python stack) and Ooyala Jobserver ? Does Spindle allow sharing of spark context over multiple spark jobs like jobserver ? Thanks. Deb On Sat, Aug 16, 2014 at 12:19 AM, Matei Zaharia matei.zaha...@gmail.com wrote: Thanks for sharing this, Brandon! Looks like a great architecture for people to build on. Matei On August 15, 2014 at 2:07:06 PM, Brandon Amos (a...@adobe.com) wrote: Hi Spark community, At Adobe Research, we're happy to open source a prototype technology called Spindle we've been developing over the past few months for processing analytics queries with Spark. Please take a look at the repository on GitHub at https://github.com/adobe-research/spindle, and we welcome any feedback. Thanks! Regards, Brandon. -- View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Open-sourcing-Spindle-by-Adobe-Research-a-web-analytics-processing-engine-in-Scala-Spark-and-Parquet-tp12203.html Sent from the Apache Spark User List mailing list archive at Nabble.com. - To unsubscribe, e-mail: user-unsubscr...@spark.apache.org For additional commands, e-mail: user-h...@spark.apache.org