Skip to main content

Which one should I use - PrestoDB or Trino ?

By Dinesh

 

First thing to understand is why to use Presto or Trino. 

  • We had been running two clusters specifically Hortonworks (HDP) variant & Cloudera (CDP) variant. 
  • Hive Tables built on HDP were mostly ORC whereas Tables that existed for us on CDP were mostly Parquet.
  • We wanted to add ad-hoc querying functionality to our cluster. And, we came across Apache Impala as an excellent tool for this purposes. 
    • Only CDP supported Apache Impala.
    • Impala had limitation to work with Parquet, Kudu, HBase. Before CDP 6.* there was no support for ORC file format with Impala.
  • Thus, we came to know about PrestoDB, which was built at Facebook, and was an excellent distributed SQL Engine for ad-hoc querying. 
    • It not only supported ORC but has connectors for multiple data sources.

A bit history of Presto - 
  • Developed at Facebook ( 2012)
  • Supported by Presto Foundation establish by Linux Foundation (2019)
    • Original Developers & Linux Foundation get into conflict on naming & branding.
  • Did a hard Fork of PrestoSQL, rebranded it as Trino ( Dec 2020)
    • Supported by non profit org - Trino Software Foundation. 

Initially, we got confused that PrestoSQL is renamed as Trino. But, later we found out that - 
  • Now, there are 2 separate variants-  PrestoDB & Trino.
  • And, certainly have different vision(s).
And, that lead to confusion - as which one to use. Below table depicts, a few high level changes that we could figure out - 

Presto

Trino

Apache License 2.0, supported by The Presto Foundation hosted by Linux Foundation

Apache License 2.0 and supported by the Trino Software Foundation.

Presto on YARN –

https://prestodb.io/presto-yarn/

Apache Slider was supported by HDP but not by CDP https://www.cloudera.com/products/open-source/apache-hadoop/apache-slider.html

Trino on YARN abandoned.

 

https://github.com/trinodb/trino/discussions/6794

https://ahana.io/presto-vs-trino/ - Trino not used at Facebook

False claims of Trino being used at Facebook

https://pandio.com/difference-between-trino-and-prestodb/#:~:text=While%20Trino%20is%20an%20excellent,makes%20Trino%20better%20than%20Presto.

PrestoDB still leading GitHub Stars

PrestoSQL/ Trino matching up with PrestoDB

https://ahana.io/presto-vs-trino/

Ahana still part of Presto Foundation and supporting PrestoDB

Starburst is  also member of Presto Foundation and managing conformance program with other members, to produce enterprise-grade distributions for Presto, which they develop from Trino. 

But, they still suggest that its same software - https://www.starburst.io/blog/prestosql-becomes-trino/

Less inclined towards creating new connectors. Refer - https://prestodb.io/docs/current/connector.html

Seemingly, Trino is more inclined towards creating new connectors, like they already have Atop, Ignite, Kinesis, SingleStore connector which are not there in PrestoDB

Refer - https://trino.io/docs/current/connector.html

 

Presto has worked towards performance

Gains, as listed below –

 

Aria - push down entire expressions to the data source for some file formats like ORC

 

Presto Unlimited - create temporary in-memory bucketed tables

 

dynamic SQL functions

 

Presto-on-Spark to get ETL Fault Tolerance.

 

RaptorX Project for Caching

 

Disaggregated Coordinator  for scaling horizontally.

 

There are more features which are developed by Ahana - https://ahana.io/presto-vs-trino/

Trino Lacks these developments of Presto. But, may be supported by Enterprise Starburst




Conclusion 
Trino seemed to be next popular buzz in the market at this time. It had increasing GitHub stars, more companies were inclined towards using Trino, Community Support is growing , etc. 

But, we choose PrestoDB over Trino, due to - 

  • Reliability and scalability.

  • We were no interested in new connectors, or docker / cloud at this moment. Our interest were with performance gains like RaptorX caching, Aria scan and predicate pushdown, and Presto on Spark ( for reliability and fault tolerance ) 
  • PrestoDB is hosted by Linux Foundation, giving confidence to us on usage. 

Cloudera added ORC support to Impala. It would be good to benchmark PrestoDB (ORC) against Impala (ORC) to see the right fit.  

Comments

Post a Comment

Popular posts

Spark MongoDB Connector Not leading to correct count or data while reading

By Dinesh
  We are using Scala 2.11 , Spark 2.4 and Spark MongoDB Connector 2.4.4 Use Case 1 - We wanted to read a Shareded Mongo Collection and copy its data to another Mongo Collection. We noticed that after Spark Job successful completion. Output MongoDB did not had many records. Use Case 2 -  We read a MongoDB collection and doing count on dataframe lead to different count on each execution. Analysis,  We realized that MongoDB Spark Connector is missing data on bulk read as a dataframe. We tried various partitioner, listed on page -  https://www.mongodb.com/docs/spark-connector/v2.4/configuration/  But, none of them worked for us. Finally, we tried  MongoShardedPartitioner  this lead to constant count on each execution. But, it was greater than the actual count of records on the collection. This seems to be limitation with MongoDB Spark Connector. But,  MongoShardedPartitioner  seemed closest possible solution to this kind of situation. But, it per...
Post a Comment
Read more




Scala Spark building Jar leads java.lang.StackOverflowError

By Dinesh
  Exception -  [Thread-3] ERROR scala_maven.ScalaCompileMojo - error: java.lang.StackOverflowError [Thread-3] INFO scala_maven.ScalaCompileMojo - at scala.collection.generic.TraversableForwarder$class.isEmpty(TraversableForwarder.scala:36) [Thread-3] INFO scala_maven.ScalaCompileMojo - at scala.collection.mutable.ListBuffer.isEmpty(ListBuffer.scala:45) [Thread-3] INFO scala_maven.ScalaCompileMojo - at scala.collection.mutable.ListBuffer.toList(ListBuffer.scala:306) [Thread-3] INFO scala_maven.ScalaCompileMojo - at scala.collection.mutable.ListBuffer.result(ListBuffer.scala:300) [Thread-3] INFO scala_maven.ScalaCompileMojo - at scala.collection.mutable.Stack$StackBuilder.result(Stack.scala:31) [Thread-3] INFO scala_maven.ScalaCompileMojo - at scala.collection.mutable.Stack$StackBuilder.result(Stack.scala:27) [Thread-3] INFO scala_maven.ScalaCompileMojo - at scala.collection.generic.GenericCompanion.apply(GenericCompanion.scala:50) [Thread-3] INFO scala_maven.ScalaCompile...
Post a Comment
Read more




MongoDB Chunk size many times bigger than configure chunksize (128 MB)

By Dinesh
  Shard Shard_0 at Shard_0/xyz.com:27018 { data: '202.04GiB', docs: 117037098, chunks: 5, 'estimated data per chunk': '40.4GiB', 'estimated docs per chunk': 23407419 } --- Shard Shard_1 at Shard_1/abc.com:27018 { data: '201.86GiB', docs: 116913342, chunks: 4, 'estimated data per chunk': '50.46GiB', 'estimated docs per chunk': 29228335 } Per MongoDB-  Starting in 6.0.3, we balance by data size instead of the number of chunks. So the 128MB is now only the size of data we migrate at-a-time. So large data size per chunk is good now, as long as the data size per shard is even for the collection. refer -  https://www.mongodb.com/community/forums/t/chunk-size-many-times-bigger-than-configure-chunksize-128-mb/212616 https://www.mongodb.com/docs/v6.0/release-notes/6.0/#std-label-release-notes-6.0-balancing-policy-changes
Post a Comment
Read more




AWS EMR Spark – Much Larger Executors are Created than Requested

By Dinesh
  Starting EMR 5.32 and EMR 6.2 you can notice that Spark can launch much larger executors that you request in your job settings. For example - We started a Spark Job with  spark.executor.cores  =   4 But, one can see that the executors with 20 cores (instead of 4 as defined by spark.executor.cores) were launched. The reason for allocating larger executors is that there is a AWS specific Spark option spark.yarn.heterogeneousExecutors.enabled (exists in EMR only, does not exist in Open Source Spark) that is set to true by default that combines multiple executor creation requests on the same node into a larger executor container. So as the result you have fewer executor containers than you expected, each of them has more memory and cores that you specified. If you disable this option (--conf "spark.yarn.heterogeneousExecutors.enabled=false"), EMR will create containers with the specified spark.executor.memory and spark.executor.cores settings and will not co...
Post a Comment
Read more




Hive Count Query not working

By Dinesh
Hive with Tez execution engine -  count(*) not working , returning 0 results.  Solution -  set hive.compute.query.using.stats=false Refer -  https://cwiki.apache.org/confluence/display/Hive/Configuration+Properties hive.compute.query.using.stats Default Value:  false Added In: Hive 0.13.0 with  HIVE-5483 When set to true Hive will answer a few queries like min, max, and count(1) purely using statistics stored in the metastore. For basic statistics collection, set the configuration property  hive.stats.autogather   to true. For more advanced statistics collection, run ANALYZE TABLE queries.
Post a Comment
Read more