Need advice about which tool to choose?Ask the StackShare community!
AWS Glue vs Apache Spark vs Presto: What are the differences?
Introduction:
AWS Glue, Apache Spark, and Presto are three popular tools used for big data processing and analysis. Each tool has its unique features and capabilities that set them apart from each other.
1. Scalability: AWS Glue is a fully managed ETL service that automatically scales based on the size of your data. Apache Spark, on the other hand, requires manual configuration and management of resources for scalability. Presto is designed for interactive queries and can scale horizontally by adding more nodes to the cluster.
2. Performance: Apache Spark is known for its in-memory processing capabilities, which make it faster than traditional disk-based processing systems. AWS Glue is optimized for Amazon's cloud infrastructure and may not perform as well on other platforms. Presto excels in executing ad-hoc SQL queries with high performance.
3. Cost: AWS Glue is a fully managed service, which means you only pay for the resources you use without worrying about infrastructure management costs. Apache Spark and Presto require more hands-on management, which may result in higher operational costs.
4. Data Sources: AWS Glue is tightly integrated with other AWS services, making it easy to access data stored in S3, RDS, DynamoDB, and Redshift. Apache Spark supports a wide range of data sources and formats, including HDFS, S3, JDBC, Avro, and Parquet. Presto also supports various data sources, such as Hive, HDFS, RDBMS, and Cassandra.
5. Language Support: Apache Spark provides APIs in Scala, Java, Python, and R, giving users flexibility in choosing their preferred programming language. Presto, on the other hand, primarily uses SQL for querying data but also supports some functions in Java. AWS Glue offers support for Python scripting for ETL jobs.
6. Use Cases: AWS Glue is suitable for data integration and transformation tasks, making it ideal for ETL processes in data pipelines. Apache Spark is commonly used for data processing, machine learning, and real-time analytics applications. Presto is preferred for querying large datasets stored in data lakes and data warehouses.
In Summary, AWS Glue, Apache Spark, and Presto differ in scalability, performance, cost, data sources, language support, and use cases, catering to different needs in the big data processing landscape.
We need to perform ETL from several databases into a data warehouse or data lake. We want to
- keep raw and transformed data available to users to draft their own queries efficiently
- give users the ability to give custom permissions and SSO
- move between open-source on-premises development and cloud-based production environments
We want to use inexpensive Amazon EC2 instances only on medium-sized data set 16GB to 32GB feeding into Tableau Server or PowerBI for reporting and data analysis purposes.
You could also use AWS Lambda and use Cloudwatch event schedule if you know when the function should be triggered. The benefit is that you could use any language and use the respective database client.
But if you orchestrate ETLs then it makes sense to use Apache Airflow. This requires Python knowledge.
Though we have always built something custom, Apache airflow (https://airflow.apache.org/) stood out as a key contender/alternative when it comes to open sources. On the commercial offering, Amazon Redshift combined with Amazon Kinesis (for complex manipulations) is great for BI, though Redshift as such is expensive.
You may want to look into a Data Virtualization product called Conduit. It connects to disparate data sources in AWS, on prem, Azure, GCP, and exposes them as a single unified Spark SQL view to PowerBI (direct query) or Tableau. Allows auto query and caching policies to enhance query speeds and experience. Has a GPU query engine and optimized Spark for fallback. Can be deployed on your AWS VM or on prem, scales up and out. Sounds like the ideal solution to your needs.
We have a Kafka topic having events of type A and type B. We need to perform an inner join on both type of events using some common field (primary-key). The joined events to be inserted in Elasticsearch.
In usual cases, type A and type B events (with same key) observed to be close upto 15 minutes. But in some cases they may be far from each other, lets say 6 hours. Sometimes event of either of the types never come.
In all cases, we should be able to find joined events instantly after they are joined and not-joined events within 15 minutes.
The first solution that came to me is to use upsert to update ElasticSearch:
- Use the primary-key as ES document id
- Upsert the records to ES as soon as you receive them. As you are using upsert, the 2nd record of the same primary-key will not overwrite the 1st one, but will be merged with it.
Cons: The load on ES will be higher, due to upsert.
To use Flink:
- Create a KeyedDataStream by the primary-key
- In the ProcessFunction, save the first record in a State. At the same time, create a Timer for 15 minutes in the future
- When the 2nd record comes, read the 1st record from the State, merge those two, and send out the result, and clear the State and the Timer if it has not fired
- When the Timer fires, read the 1st record from the State and send out as the output record.
- Have a 2nd Timer of 6 hours (or more) if you are not using Windowing to clean up the State
Pro: if you have already having Flink ingesting this stream. Otherwise, I would just go with the 1st solution.
Please refer "Structured Streaming" feature of Spark. Refer "Stream - Stream Join" at https://spark.apache.org/docs/latest/structured-streaming-programming-guide.html#stream-stream-joins . In short you need to specify "Define watermark delays on both inputs" and "Define a constraint on time across the two inputs"
I have to collect different data from multiple sources and store them in a single cloud location. Then perform cleaning and transforming using PySpark, and push the end results to other applications like reporting tools, etc. What would be the best solution? I can only think of Azure Data Factory + Databricks. Are there any alternatives to #AWS services + Databricks?
Hi all,
Currently, we need to ingest the data from Amazon S3 to DB either Amazon Athena or Amazon Redshift. But the problem with the data is, it is in .PSV (pipe separated values) format and the size is also above 200 GB. The query performance of the timeout in Athena/Redshift is not up to the mark, too slow while compared to Google BigQuery. How would I optimize the performance and query result time? Can anyone please help me out?
you can use aws glue service to convert you pipe format data to parquet format , and thus you can achieve data compression . Now you should choose Redshift to copy your data as it is very huge. To manage your data, you should partition your data in S3 bucket and also divide your data across the redshift cluster
First of all you should make your choice upon Redshift or Athena based on your use case since they are two very diferent services - Redshift is an enterprise-grade MPP Data Warehouse while Athena is a SQL layer on top of S3 with limited performance. If performance is a key factor, users are going to execute unpredictable queries and direct and managing costs are not a problem I'd definitely go for Redshift. If performance is not so critical and queries will be predictable somewhat I'd go for Athena.
Once you select the technology you'll need to optimize your data in order to get the queries executed as fast as possible. In both cases you may need to adapt the data model to fit your queries better. In the case you go for Athena you'd also proabably need to change your file format to Parquet or Avro and review your partition strategy depending on your most frequent type of query. If you choose Redshift you'll need to ingest the data from your files into it and maybe carry out some tuning tasks for performance gain.
I'll recommend Redshift for now since it can address a wider range of use cases, but we could give you better advice if you described your use case in depth.
It depend of the nature of your data (structured or not?) and of course your queries (ad-hoc or predictible?). For example you can look at partitioning and columnar format to maximize MPP capabilities for both Athena and Redshift
you can change your PSV fomat data to parquet file format with AWS GLUE and then your query performance will be improved
To provide employees with the critical need of interactive querying, we’ve worked with Presto, an open-source distributed SQL query engine, over the years. Operating Presto at Pinterest’s scale has involved resolving quite a few challenges like, supporting deeply nested and huge thrift schemas, slow/ bad worker detection and remediation, auto-scaling cluster, graceful cluster shutdown and impersonation support for ldap authenticator.
Our infrastructure is built on top of Amazon EC2 and we leverage Amazon S3 for storing our data. This separates compute and storage layers, and allows multiple compute clusters to share the S3 data.
We have hundreds of petabytes of data and tens of thousands of Apache Hive tables. Our Presto clusters are comprised of a fleet of 450 r4.8xl EC2 instances. Presto clusters together have over 100 TBs of memory and 14K vcpu cores. Within Pinterest, we have close to more than 1,000 monthly active users (out of total 1,600+ Pinterest employees) using Presto, who run about 400K queries on these clusters per month.
Each query submitted to Presto cluster is logged to a Kafka topic via Singer. Singer is a logging agent built at Pinterest and we talked about it in a previous post. Each query is logged when it is submitted and when it finishes. When a Presto cluster crashes, we will have query submitted events without corresponding query finished events. These events enable us to capture the effect of cluster crashes over time.
Each Presto cluster at Pinterest has workers on a mix of dedicated AWS EC2 instances and Kubernetes pods. Kubernetes platform provides us with the capability to add and remove workers from a Presto cluster very quickly. The best-case latency on bringing up a new worker on Kubernetes is less than a minute. However, when the Kubernetes cluster itself is out of resources and needs to scale up, it can take up to ten minutes. Some other advantages of deploying on Kubernetes platform is that our Presto deployment becomes agnostic of cloud vendor, instance types, OS, etc.
#BigData #AWS #DataScience #DataEngineering
The platform deals with time series data from sensors aggregated against things( event data that originates at periodic intervals). We use Cassandra as our distributed database to store time series data. Aggregated data insights from Cassandra is delivered as web API for consumption from other applications. Presto as a distributed sql querying engine, can provide a faster execution time provided the queries are tuned for proper distribution across the cluster. Another objective that we had was to combine Cassandra table data with other business data from RDBMS or other big data systems where presto through its connector architecture would have opened up a whole lot of options for us.
Pros of AWS Glue
- Managed Hive Metastore9
Pros of Presto
- Works directly on files in s3 (no ETL)18
- Open-source13
- Join multiple databases12
- Scalable10
- Gets ready in minutes7
- MPP6
Pros of Apache Spark
- Open-source61
- Fast and Flexible48
- One platform for every big data problem8
- Great for distributed SQL like applications8
- Easy to install and to use6
- Works well for most Datascience usecases3
- Interactive Query2
- Machine learning libratimery, Streaming in real2
- In memory Computation2
Sign up to add or upvote prosMake informed product decisions
Cons of AWS Glue
Cons of Presto
Cons of Apache Spark
- Speed4