Archive for category Java Development
Exploring Popular Open-source Stream Processing Technologies: Part 2 of 2
Posted by Gary A. Stafford in Analytics, Big Data, Java Development, Python, Software Development, SQL on September 26, 2022
A brief demonstration of Apache Spark Structured Streaming, Apache Kafka Streams, Apache Flink, and Apache Pinot with Apache Superset
Introduction
According to TechTarget, “Stream processing is a data management technique that involves ingesting a continuous data stream to quickly analyze, filter, transform or enhance the data in real-time. Once processed, the data is passed off to an application, data store, or another stream processing engine.” Confluent, a fully-managed Apache Kafka market leader, defines stream processing as “a software paradigm that ingests, processes, and manages continuous streams of data while they’re still in motion.”
This two-part post series and forthcoming video explore four popular open-source software (OSS) stream processing projects: Apache Spark Structured Streaming, Apache Kafka Streams, Apache Flink, and Apache Pinot.
This post uses the open-source projects, making it easier to follow along with the demonstration and keeping costs to a minimum. However, you could easily substitute the open-source projects for your preferred SaaS, CSP, or COSS service offerings.
Part Two
We will continue our exploration in part two of this two-part post, covering Apache Flink and Apache Pinot. In addition, we will incorporate Apache Superset into the demonstration to visualize the real-time results of our stream processing pipelines as a dashboard.
Demonstration #3: Apache Flink
In the third demonstration of four, we will examine Apache Flink. For this part of the post, we will also use the third of the three GitHub repository projects, flink-kafka-demo. The project contains a Flink application written in Java, which performs stream processing, incremental aggregation, and multi-stream joins.
New Streaming Stack
To get started, we need to replace the first streaming Docker Swarm stack, deployed in part one, with the second streaming Docker Swarm stack. The second stack contains Apache Kafka, Apache Zookeeper, Apache Flink, Apache Pinot, Apache Superset, UI for Apache Kafka, and Project Jupyter (JupyterLab).
https://programmaticponderings.wordpress.com/media/601efca17604c3a467a4200e93d7d3ff
The stack will take a few minutes to deploy fully. When complete, there should be ten containers running in the stack.
Flink Application
The Flink application has two entry classes. The first class, RunningTotals, performs an identical aggregation function as the previous KStreams demo.
The second class, JoinStreams, joins the stream of data from the demo.purchases topic and the demo.products topic, processing and combining them, in real-time, into an enriched transaction and publishing the results to a new topic, demo.purchases.enriched.
The resulting enriched purchases messages look similar to the following:
Running the Flink Job
To run the Flink application, we must first compile it into an uber JAR.
We can copy the JAR into the Flink container or upload it through the Apache Flink Dashboard, a browser-based UI. For this demonstration, we will upload it through the Apache Flink Dashboard, accessible on port 8081.
The project’s build.gradle file has preset the Main class (Flink’s Entry class) to org.example.JoinStreams. Optionally, to run the Running Totals demo, we could change the build.gradle file and recompile, or simply change Flink’s Entry class to org.example.RunningTotals.
Before running the Flink job, restart the sales generator in the background (nohup python3 ./producer.py &) to generate a new stream of data. Then start the Flink job.
To confirm the Flink application is running, we can check the contents of the new demo.purchases.enriched topic using the Kafka CLI.
Alternatively, you can use the UI for Apache Kafka, accessible on port 9080.
Demonstration #4: Apache Pinot
In the fourth and final demonstration, we will explore Apache Pinot. First, we will query the unbounded data streams from Apache Kafka, generated by both the sales generator and the Apache Flink application, using SQL. Then, we build a real-time dashboard in Apache Superset, with Apache Pinot as our datasource.
Creating Tables
According to the Apache Pinot documentation, “a table is a logical abstraction that represents a collection of related data. It is composed of columns and rows (known as documents in Pinot).” There are three types of Pinot tables: Offline, Realtime, and Hybrid. For this demonstration, we will create three Realtime tables. Realtime tables ingest data from streams — in our case, Kafka — and build segments from the consumed data. Further, according to the documentation, “each table in Pinot is associated with a Schema. A schema defines what fields are present in the table along with the data types. The schema is stored in Zookeeper, along with the table configuration.”
Below, we see the schema and config for one of the three Realtime tables, purchasesEnriched. Note how the columns are divided into three categories: Dimension, Metric, and DateTime.
To begin, copy the three Pinot Realtime table schemas and configurations from the streaming-sales-generator GitHub project into the Apache Pinot Controller container. Next, use a docker exec command to call the Pinot Command Line Interface’s (CLI) AddTable command to create the three tables: products, purchases, and purchasesEnriched.
To confirm the three tables were created correctly, use the Apache Pinot Data Explorer accessible on port 9000. Use the Tables tab in the Cluster Manager.
We can further inspect and edit the table’s config and schema from the Tables tab in the Cluster Manager.
The three tables are configured to read the unbounded stream of data from the corresponding Kafka topics: demo.products, demo.purchases, and demo.purchases.enriched.
Querying with Pinot
We can use Pinot’s Query Console to query the Realtime tables using SQL. According to the documentation, “Pinot provides a SQL interface for querying. It uses the [Apache] Calcite SQL parser to parse queries and uses MYSQL_ANSI dialect.”
With the generator still running, re-query the purchases table in the Query Console (select count(*) from purchases). You should notice the document count increasing each time you re-run the query since new messages are published to the demo.purchases topic by the sales generator.
If you do not observe the count increasing, ensure the sales generator and Flink enrichment job are running.
Table Joins?
It might seem logical to want to replicate the same multi-stream join we performed with Apache Flink in part three of the demonstration on the demo.products and demo.purchases topics. Further, we might presume to join the products and purchases realtime tables by writing a SQL statement in Pinot’s Query Console. However, according to the documentation, at the time of this post, version 0.11.0 of Pinot did not [currently] support joins or nested subqueries.
This current join limitation is why we created the Realtime table, purchasesEnriched, allowing us to query Flink’s real-time results in the demo.purchases.enriched topic. We will use both Flink and Pinot as part of our stream processing pipeline, taking advantage of each tool’s individual strengths and capabilities.
Note, according to the documentation for the latest release of Pinot on the main branch, “the latest Pinot multi-stage supports inner join, left-outer, semi-join, and nested queries out of the box. It is optimized for in-memory process and latency.” For more information on joins as part of Pinot’s new multi-stage query execution engine, read the documentation, Multi-Stage Query Engine.
demo.purchases.enriched topic in real-timeAggregations
We can perform real-time aggregations using Pinot’s rich SQL query interface. For example, like previously with Spark and Flink, we can calculate running totals for the number of items sold and the total sales for each product in real time.
We can do the same with the purchasesEnriched table, which will use the continuous stream of enriched transaction data from our Apache Flink application. With the purchasesEnriched table, we can add the product name and product category for richer results. Each time we run the query, we get real-time results based on the running sales generator and Flink enrichment job.
Query Options and Indexing
Note the reference to the Star-Tree index at the start of the SQL query shown above. Pinot provides several query options, including useStarTree (true by default).
Multiple indexing techniques are available in Pinot, including Forward Index, Inverted Index, Star-tree Index, Bloom Filter, and Range Index, among others. Each has advantages in different query scenarios. According to the documentation, by default, Pinot creates a dictionary-encoded forward index for each column.
SQL Examples
Here are a few examples of SQL queries you can try in Pinot’s Query Console:
Troubleshooting Pinot
If have issues with creating the tables or querying the real-time data, you can start by reviewing the Apache Pinot logs:
Real-time Dashboards with Apache Superset
To display the real-time stream of data produced results of our Apache Flink stream processing job and made queriable by Apache Pinot, we can use Apache Superset. Superset positions itself as “a modern data exploration and visualization platform.” Superset allows users “to explore and visualize their data, from simple line charts to highly detailed geospatial charts.”
According to the documentation, “Superset requires a Python DB-API database driver and a SQLAlchemy dialect to be installed for each datastore you want to connect to.” In the case of Apache Pinot, we can use pinotdb as the Python DB-API and SQLAlchemy dialect for Pinot. Since the existing Superset Docker container does not have pinotdb installed, I have built and published a Docker Image with the driver and deployed it as part of the second streaming stack of containers.
First, we much configure the Superset container instance. These instructions are documented as part of the Superset Docker Image repository.
Once the configuration is complete, we can log into the Superset web-browser-based UI accessible on port 8088.
Pinot Database Connection and Dataset
Next, to connect to Pinot from Superset, we need to create a Database Connection and a Dataset.
The SQLAlchemy URI is shown below. Input the URI, test your connection (‘Test Connection’), make sure it succeeds, then hit ‘Connect’.
Next, create a Dataset that references the purchasesEnriched Pinot table.
purchasesEnriched Pinot tableModify the dataset’s transaction_time column. Check the is_temporal and Default datetime options. Lastly, define the DateTime format as epoch_ms.
transaction_time columnBuilding a Real-time Dashboard
Using the new dataset, which connects Superset to the purchasesEnriched Pinot table, we can construct individual charts to be placed on a dashboard. Build a few charts to include on your dashboard.
Create a new Superset dashboard and add the charts and other elements, such as headlines, dividers, and tabs.
We can apply a refresh interval to the dashboard to continuously query Pinot and visualize the results in near real-time.
Conclusion
In this two-part post series, we were introduced to stream processing. We explored four popular open-source stream processing projects: Apache Spark Structured Streaming, Apache Kafka Streams, Apache Flink, and Apache Pinot. Next, we learned how we could solve similar stream processing and streaming analytics challenges using different streaming technologies. Lastly, we saw how these technologies, such as Kafka, Flink, Pinot, and Superset, could be integrated to create effective stream processing pipelines.
This blog represents my viewpoints and not of my employer, Amazon Web Services (AWS). All product names, logos, and brands are the property of their respective owners. All diagrams and illustrations are the property of the author unless otherwise noted.
Exploring Popular Open-source Stream Processing Technologies: Part 1 of 2
Posted by Gary A. Stafford in Analytics, Big Data, Java Development, Python, Software Development, SQL on September 24, 2022
A brief demonstration of Apache Spark Structured Streaming, Apache Kafka Streams, Apache Flink, and Apache Pinot with Apache Superset
Introduction
According to TechTarget, “Stream processing is a data management technique that involves ingesting a continuous data stream to quickly analyze, filter, transform or enhance the data in real-time. Once processed, the data is passed off to an application, data store, or another stream processing engine.” Confluent, a fully-managed Apache Kafka market leader, defines stream processing as “a software paradigm that ingests, processes, and manages continuous streams of data while they’re still in motion.”
Batch vs. Stream Processing
Again, according to Confluent, “Batch processing is when the processing and analysis happens on a set of data that have already been stored over a period of time.” A batch processing example might include daily retail sales data, which is aggregated and tabulated nightly after the stores close. Conversely, “streaming data processing happens as the data flows through a system. This results in analysis and reporting of events as it happens.” To use a similar example, instead of nightly batch processing, the streams of sales data are processed, aggregated, and analyzed continuously throughout the day — sales volume, buying trends, inventory levels, and marketing program performance are tracked in real time.
Bounded vs. Unbounded Data
According to Packt Publishing’s book, Learning Apache Apex, “bounded data is finite; it has a beginning and an end. Unbounded data is an ever-growing, essentially infinite data set.” Batch processing is typically performed on bounded data, whereas stream processing is most often performed on unbounded data.
Stream Processing Technologies
There are many technologies available to perform stream processing. These include proprietary custom software, commercial off-the-shelf (COTS) software, fully-managed service offerings from Software as a Service (or SaaS) providers, Cloud Solution Providers (CSP), Commercial Open Source Software (COSS) companies, and popular open-source projects from the Apache Software Foundation and Linux Foundation.
The following two-part post and forthcoming video will explore four popular open-source software (OSS) stream processing projects, including Apache Spark Structured Streaming, Apache Kafka Streams, Apache Flink, and Apache Pinot. Each of these projects has some equivalent SaaS, CSP, and COSS offerings.
This post uses the open-source projects, making it easier to follow along with the demonstration and keeping costs to a minimum. However, you could easily substitute the open-source projects for your preferred SaaS, CSP, or COSS service offerings.
Apache Spark Structured Streaming
According to the Apache Spark documentation, “Structured Streaming is a scalable and fault-tolerant stream processing engine built on the Spark SQL engine. You can express your streaming computation the same way you would express a batch computation on static data.” Further, “Structured Streaming queries are processed using a micro-batch processing engine, which processes data streams as a series of small batch jobs thereby achieving end-to-end latencies as low as 100 milliseconds and exactly-once fault-tolerance guarantees.” In the post, we will examine both batch and stream processing using a series of Apache Spark Structured Streaming jobs written in PySpark.
Apache Kafka Streams
According to the Apache Kafka documentation, “Kafka Streams [aka KStreams] is a client library for building applications and microservices, where the input and output data are stored in Kafka clusters. It combines the simplicity of writing and deploying standard Java and Scala applications on the client side with the benefits of Kafka’s server-side cluster technology.” In the post, we will examine a KStreams application written in Java that performs stream processing and incremental aggregation.
Apache Flink
According to the Apache Flink documentation, “Apache Flink is a framework and distributed processing engine for stateful computations over unbounded and bounded data streams. Flink has been designed to run in all common cluster environments, perform computations at in-memory speed and at any scale.” Further, “Apache Flink excels at processing unbounded and bounded data sets. Precise control of time and state enables Flink’s runtime to run any kind of application on unbounded streams. Bounded streams are internally processed by algorithms and data structures that are specifically designed for fixed-sized data sets, yielding excellent performance.” In the post, we will examine a Flink application written in Java, which performs stream processing, incremental aggregation, and multi-stream joins.
Apache Pinot
According to Apache Pinot’s documentation, “Pinot is a real-time distributed OLAP datastore, purpose-built to provide ultra-low-latency analytics, even at extremely high throughput. It can ingest directly from streaming data sources — such as Apache Kafka and Amazon Kinesis — and make the events available for querying instantly. It can also ingest from batch data sources such as Hadoop HDFS, Amazon S3, Azure ADLS, and Google Cloud Storage.” In the post, we will query the unbounded data streams from Apache Kafka, generated by Apache Flink, using SQL.
Streaming Data Source
We must first find a good unbounded data source to explore or demonstrate these streaming technologies. Ideally, the streaming data source should be complex enough to allow multiple types of analyses and visualize different aspects with Business Intelligence (BI) and dashboarding tools. Additionally, the streaming data source should possess a degree of consistency and predictability while displaying a reasonable level of variability and randomness.
To this end, we will use the open-source Streaming Synthetic Sales Data Generator project, which I have developed and made available on GitHub. This project’s highly-configurable, Python-based, synthetic data generator generates an unbounded stream of product listings, sales transactions, and inventory restocking activities to a series of Apache Kafka topics.
Source Code
All the source code demonstrated in this post is open source and available on GitHub. There are three separate GitHub projects:
Docker
To make it easier to follow along with the demonstration, we will use Docker Swarm to provision the streaming tools. Alternatively, you could use Kubernetes (e.g., creating a Helm chart) or your preferred CSP or SaaS managed services. Nothing in this demonstration requires you to use a paid service.
The two Docker Swarm stacks are located in the Streaming Synthetic Sales Data Generator project:
- Streaming Stack — Part 1: Apache Kafka, Apache Zookeeper, Apache Spark, UI for Apache Kafka, and the KStreams application
- Streaming Stack — Part 2: Apache Kafka, Apache Zookeeper, Apache Flink, Apache Pinot, Apache Superset, UI for Apache Kafka, and Project Jupyter (JupyterLab).*
* the Jupyter container can be used as an alternative to the Spark container for running PySpark jobs (follow the same steps as for Spark, below)
Demonstration #1: Apache Spark
In the first of four demonstrations, we will examine two Apache Spark Structured Streaming jobs, written in PySpark, demonstrating both batch processing (spark_batch_kafka.py) and stream processing (spark_streaming_kafka.py). We will read from a single stream of data from a Kafka topic, demo.purchases, and write to the console.
Deploying the Streaming Stack
To get started, deploy the first streaming Docker Swarm stack containing the Apache Kafka, Apache Zookeeper, Apache Spark, UI for Apache Kafka, and the KStreams application containers.
The stack will take a few minutes to deploy fully. When complete, there should be a total of six containers running in the stack.
Sales Generator
Before starting the streaming data generator, confirm or modify the configuration/configuration.ini. Three configuration items, in particular, will determine how long the streaming data generator runs and how much data it produces. We will set the timing of transaction events to be generated relatively rapidly for test purposes. We will also set the number of events high enough to give us time to explore the Spark jobs. Using the below settings, the generator should run for an average of approximately 50–60 minutes: (((5 sec + 2 sec)/2)*1000 transactions)/60 sec=~58 min on average. You can run the generator again if necessary or increase the number of transactions.
Start the streaming data generator as a background service:
The streaming data generator will start writing data to three Apache Kafka topics: demo.products, demo.purchases, and demo.inventories. We can view these topics and their messages by logging into the Apache Kafka container and using the Kafka CLI:
Below, we see a few sample messages from the demo.purchases topic:
demo.purchases topicAlternatively, you can use the UI for Apache Kafka, accessible on port 9080.
demo.purchases topic in the UI for Apache Kafka
demo.purchases topic using the UI for Apache KafkaPrepare Spark
Next, prepare the Spark container to run the Spark jobs:
Running the Spark Jobs
Next, copy the jobs from the project to the Spark container, then exec back into the container:
Batch Processing with Spark
The first Spark job, spark_batch_kafka.py, aggregates the number of items sold and the total sales for each product, based on existing messages consumed from the demo.purchases topic. We use the PySpark DataFrame class’s read() and write() methods in the first example, reading from Kafka and writing to the console. We could just as easily write the results back to Kafka.
The batch processing job sorts the results and outputs the top 25 items by total sales to the console. The job should run to completion and exit successfully.
To run the batch Spark job, use the following commands:
Stream Processing with Spark
The stream processing Spark job, spark_streaming_kafka.py, also aggregates the number of items sold and the total sales for each item, based on messages consumed from the demo.purchases topic. However, as shown in the code snippet below, this job continuously aggregates the stream of data from Kafka, displaying the top ten product totals within an arbitrary ten-minute sliding window, with a five-minute overlap, and updates output every minute to the console. We use the PySpark DataFrame class’s readStream() and writeStream() methods as opposed to the batch-oriented read() and write() methods in the first example.
Shorter event-time windows are easier for demonstrations — in Production, hourly, daily, weekly, or monthly windows are more typical for sales analysis.
To run the stream processing Spark job, use the following commands:
We could just as easily calculate running totals for the stream of sales data versus aggregations over a sliding event-time window (example job included in project).
Be sure to kill the stream processing Spark jobs when you are done, or they will continue to run, awaiting more data.
Demonstration #2: Apache Kafka Streams
Next, we will examine Apache Kafka Streams (aka KStreams). For this part of the post, we will also use the second of the three GitHub repository projects, kstreams-kafka-demo. The project contains a KStreams application written in Java that performs stream processing and incremental aggregation.
KStreams Application
The KStreams application continuously consumes the stream of messages from the demo.purchases Kafka topic (source) using an instance of the StreamBuilder() class. It then aggregates the number of items sold and the total sales for each item, maintaining running totals, which are then streamed to a new demo.running.totals topic (sink). All of this using an instance of the KafkaStreams() Kafka client class.
Running the Application
We have at least three choices to run the KStreams application for this demonstration: 1) running locally from our IDE, 2) a compiled JAR run locally from the command line, or 3) a compiled JAR copied into a Docker image, which is deployed as part of the Swarm stack. You can choose any of the options.
Compiling and running the KStreams application locally
We will continue to use the same streaming Docker Swarm stack used for the Apache Spark demonstration. I have already compiled a single uber JAR file using OpenJDK 17 and Gradle from the project’s source code. I then created and published a Docker image, which is already part of the running stack.
Since we ran the sales generator earlier for the Spark demonstration, there is existing data in the demo.purchases topic. Re-run the sales generator (nohup python3 ./producer.py &) to generate a new stream of data. View the results of the KStreams application, which has been running since the stack was deployed using the Kafka CLI or UI for Apache Kafka:
Below, in the top terminal window, we see the output from the KStreams application. Using KStream’s peek() method, the application outputs Purchase and Total instances to the console as they are processed and written to Kafka. In the lower terminal window, we see new messages being published as a continuous stream to output topic, demo.running.totals.
Part Two
In part two of this two-part post, we continue our exploration of the four popular open-source stream processing projects. We will cover Apache Flink and Apache Pinot. In addition, we will incorporate Apache Superset into the demonstration, building a real-time dashboard to visualize the results of our stream processing.
This blog represents my viewpoints and not of my employer, Amazon Web Services (AWS). All product names, logos, and brands are the property of their respective owners. All diagrams and illustrations are the property of the author unless otherwise noted.
Utilizing In-memory Data Caching to Enhance the Performance of Data Lake-based Applications
Posted by Gary A. Stafford in Analytics, AWS, Big Data, Cloud, Java Development, Software Development, SQL on July 12, 2022
Significantly improve the performance and reduce the cost of data lake-based analytics applications using Amazon ElastiCache for Redis
Introduction
The recent post, Developing Spring Boot Applications for Querying Data Lakes on AWS using Amazon Athena, demonstrated how to develop a Cloud-native analytics application using Spring Boot. The application queried data in an Amazon S3-based data lake via an AWS Glue Data Catalog utilizing the Amazon Athena API.
Securely exposing data in a data lake using RESTful APIs can fulfill many data-consumer needs. However, access to that data can be significantly slower than access from a database or data warehouse. For example, in the previous post, we imported the OpenAPI v3 specification from the Spring Boot service into Postman. The API specification contained approximately 17 endpoints.
From my local development laptop, the Postman API test run times for all service endpoints took an average of 32.4 seconds. The Spring Boot service was running three Kubernetes pod replicas on Amazon EKS in the AWS US East (N. Virginia) Region.
Compare the data lake query result times to equivalent queries made against a minimally-sized Amazon RDS for PostgreSQL database instance containing the same data. The average run times for all PostgreSQL queries averaged 10.8 seconds from a similar Spring Boot service. Although not a precise benchmark, we can clearly see that access to the data in the Amazon S3-based data lake is substantially slower, approximately 3x slower, than that of the PostgreSQL database. Tuning the database would easily create an even greater disparity.
Caching for Data Lakes
According to AWS, the speed and throughput of a database can be the most impactful factor for overall application performance. Consequently, in-memory data caching can be one of the most effective strategies to improve overall application performance and reduce database costs. This same caching strategy can be applied to analytics applications built atop data lakes, as this post will demonstrate.
In-memory Caching
According to Hazelcast, memory caching (aka in-memory caching), often simply referred to as caching, is a technique in which computer applications temporarily store data in a computer’s main memory (e.g., RAM) to enable fast retrievals of that data. The RAM used for the temporary storage is known as the cache. As an application tries to read data, typically from a data storage system like a database, it checks to see if the desired record already exists in the cache. If it does, the application will read the data from the cache, thus eliminating the slower access to the database. If the desired record is not in the cache, then the application reads the record from the source. When it retrieves that data, it writes it to the cache so that when the application needs that same data in the future, it can quickly retrieve it from the cache.
Further, according to Hazelcast, as an application tries to read data, typically from a data storage system like a database, it checks to see if the desired record already exists in the cache. If it does, the application will read the data from the cache, thus eliminating the slower access to the database. If the desired record is not in the cache, then the application reads the record from the source. When it retrieves that data, it writes it to the cache so that when the application needs that same data in the future, it can quickly retrieve it from the cache.
Redis In-memory Data Store
According to their website, Redis is the open-source, in-memory data store used by millions of developers as a database, cache, streaming engine, and message broker. Redis provides data structures such as strings, hashes, lists, sets, sorted sets with range queries, bitmaps, hyperloglogs, geospatial indexes, and streams. In addition, Redis has built-in replication, Lua scripting, LRU eviction, transactions, and different levels of on-disk persistence and provides high availability via Redis Sentinel and automatic partitioning with Redis Cluster.
Amazon ElastiCache for Redis
According to AWS, Amazon ElastiCache for Redis, the fully-managed version of Redis, according to AWS, is a blazing fast in-memory data store that provides sub-millisecond latency to power internet-scale real-time applications. Redis applications can work seamlessly with ElastiCache for Redis without any code changes. ElastiCache for Redis combines the speed, simplicity, and versatility of open-source Redis with manageability, security, and scalability from AWS. As a result, Redis is an excellent choice for implementing a highly available in-memory cache to decrease data access latency, increase throughput, and ease the load on relational and NoSQL databases.
ElastiCache Performance Results
In the following post, we will add in-memory caching to the Spring Boot service introduced in the previous post. In preliminary tests with Amazon ElastiCache for Redis, the Spring Boot service delivered a 34x improvement in average response times. For example, test runs with the best-case scenario of a Redis cache hit ratio of 100% averaged 0.95 seconds compared to 32.4 seconds without Redis.
Source Code
All the source code and Docker and Kubernetes resources are open-source and available on GitHub.
git clone --depth 1 -b redis \
https://github.com/garystafford/athena-spring-app.git
In addition, a Docker image for the Redis-base Spring Boot service is available on Docker Hub. For this post, use the latest tag with the .redis suffix.
Code Changes
The following code changes are required to the Spring Boot service to implement Spring Boot Cache with Redis.
Gradle Build
The gradle.build file now implements two additional dependencies, Spring Boot’s spring-boot-starter-cache and spring-boot-starter-data-redis (lines 45–46).
Application Properties
The application properties file, application.yml, has been modified for both the dev and prod Spring Profiles. The dev Spring Profile expects Redis to be running on localhost. Correspondingly, the project’s docker-compose.yml file now includes a Redis container for local development. The time-to-live (TTL) for all Redis caches is arbitrarily set to one minute for dev and five minutes for prod. To increase application performance and reduce the cost of querying the data lake using Athena, increase Redis’s TTL. Note that increasing the TTL will reduce data freshness.
Athena Application Class
The AthenaApplication class declaration is now decorated with Spring Framework’s EnableCaching annotation (line 22). Additionally, two new Beans have been added (lines 58–68). Spring Redis provides an implementation for the Spring cache abstraction through the org.springframework.data.redis.cache package. The RedisCacheManager cache manager creates caches by default upon the first write. The RedisCacheConfiguration cache configuration helps to customize RedisCache behavior such as caching null values, cache key prefixes, and binary serialization.
POJO Data Model Classes
Spring Boot Redis caching uses Java serialization and deserialization. Therefore, all the POJO data model classes must implement Serializable (line 14).
Service Classes
Each public method in the Service classes is now decorated with Spring Framework’s Cachable annotation (lines 42 and 66). For example, the findById(int id) method in the CategoryServiceImp class is annotated with @Cacheable(value = "categories", key = "#id"). The method’s key parameter uses Spring Expression Language (SpEL) expression for computing the key dynamically. Default is null, meaning all method parameters are considered as a key unless a custom keyGenerator has been configured. If no value is found in the Redis cache for the computed key, the target method will be invoked, and the returned value will be stored in the associated cache.
Controller Classes
There are no changes required to the Controller classes.
Amazon ElastiCache for Redis
Multiple options are available for creating an Amazon ElastiCache for Redis cluster, including cluster mode, multi-AZ option, auto-failover option, node type, number of replicas, number of shards, replicas per shard, Availability Zone placements, and encryption at rest and encryption in transit options. The results in this post are based on a minimally-configured Redis version 6.2.6 cluster, with one shard, two cache.r6g.large nodes, and cluster mode, multi-AZ option, and auto-failover all disabled. In addition, encryption at rest and encryption in transit were also disabled. This cluster configuration is sufficient for development and testing, but not Production.
Testing the Cache
To test Amazon ElastiCache for Redis, we will use Postman again with the imported OpenAPI v3 specification. With all data evicted from existing Redis caches, the first time the Postman tests run, they cause the service’s target methods to be invoked and the returned data stored in the associated caches.
To confirm this caching behavior, use the Redis CLI’s --scan option. To access the redis-cli, I deployed a single Redis pod to Amazon EKS. The first time the --scan command runs, we should get back an empty list of keys. After the first Postman test run, the same --scan command should return a list of cached keys.
Use the Redis CLI’s MONITOR option to further confirm data is being cached, as indicated by the set command.
After the initial caching of data, use the Redis CLI’s MONITOR option, again, to confirm the cache is being hit instead of calling the target methods, which would then invoke the Athena API. Rerunning the Postman tests, we should see get commands as opposed to set commands.
Lastly, to confirm the Spring Boot service is effectively using Redis to cache data, we can also check Amazon Athena’s Recent queries tab in the AWS Management Console. After repeated sequential test runs within the TTL window, we should only see one Athena query per endpoint.
Conclusion
In this brief follow-up to the recent post, Developing Spring Boot Applications for Querying Data Lakes on AWS using Amazon Athena, we saw how to substantially increase data lake application performance using Amazon ElastiCache for Redis. Although this caching technique is often associated with databases, it can also be effectively applied to data lake-based applications, as demonstrated in the post.
This blog represents my viewpoints and not of my employer, Amazon Web Services (AWS). All product names, logos, and brands are the property of their respective owners. All diagrams and illustrations are property of the author unless otherwise noted.
Developing Spring Boot Applications for Querying Data Lakes on AWS using Amazon Athena
Posted by Gary A. Stafford in Analytics, AWS, Big Data, Cloud, Java Development, Software Development on June 26, 2022
Learn how to develop Cloud-native, RESTful Java services that query data in an AWS-based data lake using Amazon Athena’s API
Introduction
AWS provides a collection of fully-managed services that makes building and managing secure data lakes faster and easier, including AWS Lake Formation, AWS Glue, and Amazon S3. Additional analytics services such as Amazon EMR, AWS Glue Studio, and Amazon Redshift allow Data Scientists and Analysts to run high-performance queries on large volumes of semi-structured and structured data quickly and economically.
What is not always as obvious is how teams develop internal and external customer-facing analytics applications built on top of data lakes. For example, imagine sellers on an eCommerce platform, the scenario used in this post, want to make better marketing decisions regarding their products by analyzing sales trends and buyer preferences. Further, suppose the data required for the analysis must be aggregated from multiple systems and data sources; the ideal use case for a data lake.
salesbyseller endpointIn this post, we will explore an example Java Spring Boot RESTful Web Service that allows end-users to query data stored in a data lake on AWS. The RESTful Web Service will access data stored as Apache Parquet in Amazon S3 through an AWS Glue Data Catalog using Amazon Athena. The service will use Spring Boot and the AWS SDK for Java to expose a secure, RESTful Application Programming Interface (API).
Amazon Athena is a serverless, interactive query service based on Presto, used to query data and analyze big data in Amazon S3 using standard SQL. Using Athena functionality exposed by the AWS SDK for Java and Athena API, the Spring Boot service will demonstrate how to access tables, views, prepared statements, and saved queries (aka named queries).
TL;DR
Do you want to explore the source code for this post’s Spring Boot service or deploy it to Kubernetes before reading the full article? All the source code, Docker, and Kubernetes resources are open-source and available on GitHub.
git clone --depth 1 -b main \
https://github.com/garystafford/athena-spring-app.git
A Docker image for the Spring Boot service is also available on Docker Hub.
Data Lake Data Source
There are endless data sources to build a demonstration data lake on AWS. This post uses the TICKIT sample database provided by AWS and designed for Amazon Redshift, AWS’s cloud data warehousing service. The database consists of seven tables. Two previous posts and associated videos, Building a Data Lake on AWS with Apache Airflow and Building a Data Lake on AWS, detail the setup of the data lake used in this post using AWS Glue and optionally Apache Airflow with Amazon MWAA.
Those two posts use the data lake pattern of segmenting data as bronze (aka raw), silver (aka refined), and gold (aka aggregated), popularized by Databricks. The data lake simulates a typical scenario where data originates from multiple sources, including an e-commerce platform, a CRM system, and a SaaS provider must be aggregated and analyzed.
Spring Projects with IntelliJ IDE
Although not a requirement, I used JetBrains IntelliJ IDEA 2022 (Ultimate Edition) to develop and test the post’s Spring Boot service. Bootstrapping Spring projects with IntelliJ is easy. Developers can quickly create a Spring project using the Spring Initializr plugin bundled with the IntelliJ.
The Spring Initializr plugin’s new project creation wizard is based on start.spring.io. The plugin allows you to quickly select the Spring dependencies you want to incorporate into your project.
Visual Studio Code
There are also several Spring extensions for the popular Visual Studio Code IDE, including Microsoft’s Spring Initializr Java Support extension.
Gradle
This post uses Gradle instead of Maven to develop, test, build, package, and deploy the Spring service. Based on the packages selected in the new project setup shown above, the Spring Initializr plugin’s new project creation wizard creates a build.gradle file. Additional packages, such as Lombak, Micrometer, and Rest Assured, were added separately.
Amazon Corretto
The Spring boot service is developed for and compiled with the most recent version of Amazon Corretto 17. According to AWS, Amazon Corretto is a no-cost, multiplatform, production-ready distribution of the Open Java Development Kit (OpenJDK). Corretto comes with long-term support that includes performance enhancements and security fixes. Corretto is certified as compatible with the Java SE standard and is used internally at Amazon for many production services.
Source Code
Each API endpoint in the Spring Boot RESTful Web Service has a corresponding POJO data model class, service interface and service implementation class, and controller class. In addition, there are also common classes such as configuration, a client factory, and Athena-specific request/response methods. Lastly, there are additional class dependencies for views and prepared statements.
refined_tickit_public_category tableThe project’s source code is arranged in a logical hierarchy by package and class type.
Amazon Athena Access
There are three standard methods for accessing Amazon Athena with the AWS SDK for Java: 1) the AthenaClient service client, 2) the AthenaAsyncClient service client for accessing Athena asynchronously, and 3) using the JDBC driver with the AWS SDK. The AthenaClient and AthenaAsyncClient service clients are both parts of the software.amazon.awssdk.services.athena package. For simplicity, this post’s Spring Boot service uses the AthenaClient service client instead of Java’s asynchronously programming model. AWS supplies basic code samples as part of their documentation as a starting point for writing Athena applications using the SDK. The code samples also use the AthenaClient service client.
POJO-based Data Model Class
For each API endpoint in the Spring Boot RESTful Web Service, there is a corresponding Plain Old Java Object (POJO). According to Wikipedia, a POGO is an ordinary Java object, not bound by any particular restriction. The POJO class is similar to a JPA Entity, representing persistent data stored in a relational database. In this case, the POJO uses Lombok’s @Data annotation. According to the documentation, this annotation generates getters for all fields, a useful toString method, and hashCode and equals implementations that check all non-transient fields. It also generates setters for all non-final fields and a constructor.
Each POJO corresponds directly to a ‘silver’ table in the AWS Glue Data Catalog. For example, the Event POJO corresponds to the refined_tickit_public_event table in the tickit_demo Data Catalog database. The POJO defines the Spring Boot service’s data model for data read from the corresponding AWS Glue Data Catalog table.
The Glue Data Catalog table is the interface between the Athena query and the underlying data stored in Amazon S3 object storage. The Athena query targets the table, which returns the underlying data from S3.
Service Class
Retrieving data from the data lake via AWS Glue, using Athena, is handled by a service class. For each API endpoint in the Spring Boot RESTful Web Service, there is a corresponding Service Interface and implementation class. The service implementation class uses Spring Framework’s @Service annotation. According to the documentation, it indicates that an annotated class is a “Service,” initially defined by Domain-Driven Design (Evans, 2003) as “an operation offered as an interface that stands alone in the model, with no encapsulated state.” Most importantly for the Spring Boot service, this annotation serves as a specialization of @Component, allowing for implementation classes to be autodetected through classpath scanning.
Using Spring’s common constructor-based Dependency Injection (DI) method (aka constructor injection), the service auto-wires an instance of the AthenaClientFactory interface. Note that we are auto-wiring the service interface, not the service implementation, allowing us to wire in a different implementation if desired, such as for testing.
The service calls the AthenaClientFactoryclass’s createClient() method, which returns a connection to Amazon Athena using one of several available authentication methods. The authentication scheme will depend on where the service is deployed and how you want to securely connect to AWS. Some options include environment variables, local AWS profile, EC2 instance profile, or token from the web identity provider.
The service class transforms the payload returned by an instance of GetQueryResultsResponse into an ordered collection (also known as a sequence), List<E>, where E represents a POJO. For example, with the data lake’srefined_tickit_public_event table, the service returns a List<Event>. This pattern repeats itself for tables, views, prepared statements, and named queries. Column data types can be transformed and formatted on the fly, new columns added, and existing columns skipped.
For each endpoint defined in the Controller class, for example, get(), findAll(), and FindById(), there is a corresponding method in the Service class. Below, we see an example of the findAll() method in the SalesByCategoryServiceImp service class. This method corresponds to the identically named method in the SalesByCategoryController controller class. Each of these service methods follows a similar pattern of constructing a dynamic Athena SQL query based on input parameters, which is passed to Athena through the AthenaClient service client using an instance of GetQueryResultsRequest.
Controller Class
Lastly, there is a corresponding Controller class for each API endpoint in the Spring Boot RESTful Web Service. The controller class uses Spring Framework’s @RestController annotation. According to the documentation, this annotation is a convenience annotation that is itself annotated with @Controller and @ResponseBody. Types that carry this annotation are treated as controllers where @RequestMapping methods assume @ResponseBody semantics by default.
The controller class takes a dependency on the corresponding service class application component using constructor-based Dependency Injection (DI). Like the service example above, we are auto-wiring the service interface, not the service implementation.
The controller is responsible for serializing the ordered collection of POJOs into JSON and returning that JSON payload in the body of the HTTP response to the initial HTTP request.
Querying Views
In addition to querying AWS Glue Data Catalog tables (aka Athena tables), we also query views. According to the documentation, a view in Amazon Athena is a logical table, not a physical table. Therefore, the query that defines a view runs each time the view is referenced in a query.
For convenience, each time the Spring Boot service starts, the main AthenaApplication class calls the View.java class’s CreateView() method to check for the existence of the view, view_tickit_sales_by_day_and_category. If the view does not exist, it is created and becomes accessible to all application end-users. The view is queried through the service’s /salesbycategory endpoint.
This confirm-or-create pattern is repeated for the prepared statement in the main AthenaApplication class (detailed in the next section).
Below, we see the View class called by the service at startup.
Aside from the fact the /salesbycategory endpoint queries a view, everything else is identical to querying a table. This endpoint uses the same model-service-controller pattern.
Executing Prepared Statements
According to the documentation, you can use the Athena parameterized query feature to prepare statements for repeated execution of the same query with different query parameters. The prepared statement used by the service, tickit_sales_by_seller, accepts a single parameter, the ID of the seller (sellerid). The prepared statement is executed using the /salesbyseller endpoint. This scenario simulates an end-user of the analytics application who wants to retrieve enriched sales information about their sales.
The pattern of querying data is similar to tables and views, except instead of using the common SELECT...FROM...WHERE SQL query pattern, we use the EXECUTE...USING pattern.
For example, to execute the prepared statement for a seller with an ID of 3, we would use EXECUTE tickit_sales_by_seller USING 3;. We pass the seller’s ID of 3 as a path parameter similar to other endpoints exposed by the service: /v1/salesbyseller/3.
Again, aside from the fact the /salesbyseller endpoint executes a prepared statement and passes a parameter; everything else is identical to querying a table or a view, using the same model-service-controller pattern.
Working with Named Queries
In addition to tables, views, and prepared statements, Athena has the concept of saved queries, referred to as named queries in the Athena API and when using AWS CloudFormation. You can use the Athena console or API to save, edit, run, rename, and delete queries. The queries are persisted using a NamedQueryId, a unique identifier (UUID) of the query. You must reference the NamedQueryId when working with existing named queries.
There are multiple ways to use and reuse existing named queries programmatically. For this demonstration, I created the named query, buyer_likes_by_category, in advance and then stored the resulting NamedQueryId as an application property, injected at runtime or kubernetes deployment time through a local environment variable.
Alternately, you might iterate through a list of named queries to find one that matches the name at startup. However, this method would undoubtedly impact service performance, startup time, and cost. Lastly, you could use a method like NamedQuery() included in the unused NamedQuery class at startup, similar to the view and prepared statement. That named query’s unique NamedQueryId would be persisted as a system property, referencable by the service class. The downside is that you would create a duplicate of the named query each time you start the service. Therefore, this method is also not recommended.
Configuration
Two components responsible for persisting configuration for the Spring Boot service are the application.yml properties file and ConfigProperties class. The class uses Spring Framework’s @ConfigurationProperties annotation. According to the documentation, this annotation is used for externalized configuration. Add this to a class definition or a @Bean method in a @Configuration class if you want to bind and validate some external Properties (e.g., from a .properties or .yml file). Binding is performed by calling setters on the annotated class or, if @ConstructorBinding in use, by binding to the constructor parameters.
The @ConfigurationProperties annotation includes the prefix of athena. This value corresponds to the athena prefix in the the application.yml properties file. The fields in the ConfigProperties class are bound to the properties in the the application.yml. For example, the property, namedQueryId, is bound to the property, athena.named.query.id. Further, that property is bound to an external environment variable, NAMED_QUERY_ID. These values could be supplied from an external configuration system, a Kubernetes secret, or external secrets management system.
AWS IAM: Authentication and Authorization
For the Spring Boot service to interact with Amazon Athena, AWS Glue, and Amazon S3, you need to establish an AWS IAM Role, which the service assumes once authenticated. The Role must be associated with an attached IAM Policy containing the requisite Athena, Glue, and S3 permissions. For development, the service uses a policy similar to the one shown below. Please note this policy is broader than recommended for Production; it does not represent the security best practice of least privilege. In particular, the use of the overly-broad * for Resources should be strictly avoided when creating policies.
In addition to the authorization granted by the IAM Policy, AWS Lake Formation can be used with Amazon S3, AWS Glue, and Amazon Athena to grant fine-grained database-, table-, column-, and row-level access to datasets.
Swagger UI and the OpenAPI Specification
The easiest way to view and experiment with all the endpoints available through the controller classes is using the Swagger UI, included in the example Spring Boot service, by way of the springdoc-openapi Java library. The Swagger UI is accessed at /v1/swagger-ui/index.html.
The OpenAPI Specification (formerly Swagger Specification) is an API description format for REST APIs. The /v1/v3/api-docs endpoint allows you to generate an OpenAPI v3 specification file. The OpenAPI file describes the entire API.
The OpenAPI v3 specification can be saved as a file and imported into applications like Postman, the API platform for building and using APIs.
/users API endpoint using Postman
Integration Tests
Included in the Spring Boot service’s source code is a limited number of example integration tests, not to be confused with unit tests. Each test class uses Spring Framework’s @SpringBootTest annotation. According to the documentation, this annotation can be specified on a test class that runs Spring Boot-based tests. It provides several features over and above the regular Spring TestContext Framework.
The integration tests use Rest Assured’s given-when-then pattern of testing, made popular as part of Behavior-Driven Development (BDD). In addition, each test uses the JUnit’s @Test annotation. According to the documentation, this annotation signals that the annotated method is a test method. Therefore, methods using this annotation must not be private or static and must not return a value.