Unlock self-serve streaming SQL with Amazon Managed Service for Apache Flink

This submit is co-written with Gal Krispel from Riskified.

Riskified is an ecommerce fraud prevention and threat administration platform that helps companies optimize on-line transactions by distinguishing reputable prospects from fraudulent ones.

Utilizing synthetic intelligence and machine studying (AI/ML), Riskified analyzes real-time transaction knowledge to detect and stop fraud whereas maximizing transaction approval charges. The platform supplies a chargeback assure, defending retailers from losses as a result of fraudulent transactions. Riskified’s options embody account safety, coverage abuse prevention, and chargeback administration software program, making it a complete device for decreasing threat and enhancing buyer expertise. Companies throughout numerous industries, together with retail, journey, and digital items, use Riskified to extend income whereas minimizing fraud-related losses. Riskified’s core enterprise of real-time fraud prevention makes low-latency streaming applied sciences a elementary a part of its resolution.

Companies usually can’t afford to attend for batch processing to make essential choices. With real-time knowledge streaming applied sciences like Apache Flink, Apache Spark, and Apache Kafka Streams, organizations can react immediately to rising tendencies, detect anomalies, and improve buyer experiences. These applied sciences are highly effective processing engines that carry out analytical operations at scale. Nevertheless, unlocking the total potential of streaming knowledge usually requires advanced engineering efforts, limiting accessibility for analysts and enterprise customers.

Streaming pipelines are in excessive demand from Riskified’s Engineering division. Subsequently, a user-friendly interface for creating streaming pipelines is a essential characteristic to extend analytical precision for detecting fraudulent transactions.

On this submit, we current Riskified’s journey towards enabling self-service streaming SQL pipelines. We stroll via the motivations behind the shift from Confluent ksqlDB to Apache Flink, the structure Riskified constructed utilizing Amazon Managed Service for Apache Flink, the technical challenges they confronted, and the options that helped them make streaming accessible, scalable, and production-ready.

Utilizing SQL to create streaming pipelines

Prospects have a variety of open supply knowledge processing applied sciences to select from, akin to Flink, Spark, ksqlDB, and RisingWave. Every platform gives a streaming API for knowledge processing. SQL streaming jobs supply a strong and intuitive method to course of real-time knowledge with minimal complexity. These pipelines use SQL, a extensively identified and declarative language, to carry out real-time transformations, filtering, aggregations, and joins in steady knowledge streams.

As an instance the ability of streaming SQL in ecommerce fraud prevention, take into account the idea of velocity checks, that are a essential fraud detection sample. Velocity checks are a kind of safety measure used to detect uncommon or fast exercise by monitoring the frequency and quantity of particular actions inside a given timeframe. These checks assist establish potential fraud or abuse by analyzing repeated behaviors that deviate from regular person patterns. Widespread examples embody detecting a number of transactions from the identical IP tackle in a short while span, monitoring bursts of account creation makes an attempt, or monitoring the repeated use of a single cost technique throughout totally different accounts.

Use case: Riskified’s velocity checks

Riskified carried out a real-time velocity test utilizing streaming SQL to observe buying conduct primarily based on person identifier.

On this setup, transaction knowledge is repeatedly streamed via a Kafka subject. Every message incorporates person agent info originating from the browser, together with the uncooked transaction knowledge. Streaming SQL queries are used to combination the variety of transactions originating from a single person identifier inside quick time home windows.

For instance, if the variety of transactions from a given person identifier exceeds a sure threshold inside a 10-second interval, this would possibly sign fraudulent exercise. When that threshold is breached, the system can routinely flag or block the transactions earlier than they’re accomplished. The next determine and accompanying code present a simplified instance of the streaming SQL question used to detect this conduct.

SELECT userIdentifier,TUMBLE_START(createdAt, INTERVAL '10' SECONDS) 
  AS windowStart,TUMBLE_END(createdAt, INTERVAL '10' SECONDS) 
  AS windowEnd, COUNT(*) AS paymentAttempts
FROM transactions
  WINDOW TUMBLING (SIZE 10 SECONDS)
GROUP BY userIdentifier;

Though defining SQL queries over static datasets would possibly seem easy, creating and sustaining sturdy streaming purposes introduces distinctive challenges. Conventional SQL operates on bounded datasets, that are finite collections of information saved in tables. In distinction, streaming SQL is designed to course of steady, unbounded knowledge streams resembling the SQL syntax.

To handle these challenges at scale and make streaming job creation accessible to engineering groups, Riskified carried out a self-serve resolution primarily based on Confluent ksqlDB, utilizing its SQL interface and built-in Kafka integration. Engineers may outline and deploy streaming pipelines utilizing SQL, chaining ksqlDB streams from supply to sink. The system supported each stateless and stateful processing straight on Kafka matters, with Avro schemas used to outline the construction of streaming knowledge.

Though ksqlDB supplied a quick and approachable start line, it will definitely revealed a number of limitations. These included challenges with schema evolution, difficulties in managing compute assets, and the absence of an abstraction for managing pipelines as a cohesive unit. Because of this, Riskified started exploring various applied sciences that would higher assist its increasing streaming use instances. The next sections define these challenges in additional element.

Evolving the stream processing structure

In evaluating options, Riskified targeted on applied sciences that would tackle the particular calls for of fraud detection whereas preserving the simplicity that made the unique method interesting. The workforce encountered the next challenges in sustaining the earlier resolution:

• Schemas are managed in Confluent Schema Registry, and the message format is Avro with FULL compatibility mode enforced. Schemas are consistently evolving in keeping with enterprise necessities. They’re model managed utilizing Git with a strict steady integration and steady supply (CI/CD) pipeline. As schemas grew extra advanced, ksqlDB’s method to schema evolution didn’t routinely incorporate newly added fields. This conduct required dropping streams and recreating them so as to add new fields as a substitute of simply restarting the appliance to include new fields. This method precipitated inconsistencies with offset administration as a result of stream’s tear-down.
• ksqlDB enforces a TopicNameStrategy schema registration technique, which supplies 1:1 schema-to-topic coupling. This implies the precise schema definition needs to be registered a number of instances, one time for every subject it’s used for. Riskified’s schema registry deployment makes use of RecordNameStrategy for schema registration. It’s an environment friendly schema registry technique that enables for sharing schemas throughout a number of matters, storing fewer schemas, and decreasing registry administration overhead. Having combined methods within the schema registry precipitated errors with Kafka client shoppers trying to decode messages, as a result of the consumer implementation anticipated a RecordNameStrategy in keeping with Riskified’s commonplace.
• ksqlDB internally registers schema definitions in particular methods the place fields are interpreted as nullable, and Avro Enum sorts are transformed to Strings. This conduct precipitated deserialization errors when trying emigrate native Kafka client purposes to make use of the ksqlDB output subject. Riskified’s code base makes use of the Scala programming language, the place optionally available fields within the schema are interpreted as Choice. Reworking each discipline as optionally available within the schema definition required heavy refactoring, treating all Enum fields as Strings, and dealing with the Choice knowledge sort for each discipline that requires secure dealing with. This cascading impact made the migration course of extra concerned, requiring further time and assets to attain a clean transition.

Managing useful resource rivalry in ksqlDB streaming workloads

ksqlDB queries are compiled right into a Kafka Streams topology. The question definition defines the topology’s conduct.

Streaming question assets are shared slightly than remoted. This method sometimes results in the overallocation of cluster assets. Its duties are distributed throughout nodes in a ksqlDB cluster. This structure means processing duties with no useful resource isolation, and a particular process can affect different duties operating on the identical node.

Useful resource rivalry between duties on the identical node is widespread in a production-intensive surroundings when utilizing a cluster structure resolution. Operation groups usually fine-tune cluster configurations to take care of acceptable efficiency, incessantly mitigating points by over-provisioning cluster nodes.

Challenges with ksqlDB pipelines

A ksqlDB pipeline is a sequence of particular person streams and lacks flow-level abstraction. Think about a posh pipeline the place a client publishes to a number of matters. In ksqlDB, every subject (each enter and output) have to be managed as a separate stream abstraction. Nevertheless, there isn’t any high-level abstraction to signify a complete pipeline that chains these streams collectively. Because of this, engineering groups should manually assemble particular person streams right into a cohesive knowledge circulate, with out built-in assist for managing them as a single, full pipeline.

This architectural method significantly impacts operational duties. Troubleshooting requires inspecting every stream individually, making it tough to observe and keep pipelines that include dozens of interconnected streams. When points happen, the well being of every stream must be checked individually, with no logical knowledge circulate part to assist perceive the relationships between streams or their position within the total pipeline. The absence of a unified view of the info circulate considerably elevated operational complexity.

Flink instead

Riskified started exploring options for its streaming platform. The necessities had been clear: a robust processing know-how that mixes a wealthy low-level API and a streaming SQL engine, backed by a robust open supply neighborhood, confirmed to carry out in essentially the most demanding manufacturing environments.

In contrast to the earlier resolution, which supported solely Kafka-to-Kafka integration, Flink gives an array of connectors for numerous databases and Streaming platforms. It was rapidly acknowledged that Flink had the potential to deal with advanced streaming use instances.

Flink gives a number of deployment choices, together with standalone clusters, native Kubernetes deployments utilizing operators, and Hadoop YARN clusters. For enterprises in search of a completely managed choice, cloud suppliers like AWS supply managed Flink companies that assist alleviate operational overhead, akin to Managed Service for Apache Flink.

Advantages of utilizing Managed Service for Apache Flink

Riskified determined to implement an answer utilizing Managed Service for Apache Flink. This alternative supplied a number of key benefits:

• It gives a fast and dependable method to run Flink purposes and reduces the operational overhead of independently managing the infrastructure.
• Managed Service for Apache Flink supplies true job isolation by operating every streaming utility in its devoted cluster. This implies you may handle assets individually for every job and scale back the chance of heavy streaming jobs inflicting useful resource hunger for different operating jobs.
• It gives built-in monitoring utilizing Amazon CloudWatch metrics, utility state backup with managed snapshots, and automated scaling.
• AWS gives complete documentation and sensible examples to assist speed up the implementation course of.

With these options, Riskified may deal with what really issues—getting nearer to the enterprise purpose and beginning to write purposes.

Utilizing Flink’s streaming SQL engine

Builders can use Flink to construct advanced and scalable streaming purposes, however Riskified noticed it as greater than only a device for consultants. They wished to democratize the ability of Flink right into a device for the whole group, to unravel advanced enterprise challenges involving real-time analytics necessities without having a devoted knowledge skilled.

To switch their earlier resolution, they envisioned sustaining a “construct as soon as, deploy many” utility, which encapsulates the complexity of the Flink programming and permits the customers to deal with the SQL processing logic.

Kafka was maintained because the enter and output know-how for the preliminary migration use case, which has similarities to the ksqlDB setup. They designed a single, versatile Flink utility the place end-users can modify the enter matters, SQL processing logic, and output locations via runtime properties. Though ksqlDB primarily focuses on Kafka integration, Flink’s in depth connector ecosystem allows it to increase to various knowledge sources and locations in future phases.

Managed Service for Apache Flink supplies a versatile method to configure streaming purposes with out modifying their code. By utilizing runtime parameters, you may change the appliance’s conduct with out modifying its supply code.

Utilizing Managed Service for Apache Flink for this method consists of the next steps:

1. Apply parameters for the enter/output Kafka subject, a SQL question, and the enter/output schema ID (assuming you’re utilizing Confluent Schema Registry).
2. Use AvroSchemaConverter to transform an Avro schema right into a Flink desk.
3. Apply the SQL processing logic and save the output as a view.
4. Sink the view outcomes into Kafka.

The next diagram illustrates this workflow.

Performing Flink SQL question compilation with no Flink runtime surroundings

Offering end-users with important management to outline their pipelines makes it essential to confirm the SQL question outlined by the person earlier than deployment. This validation prevents failed or hanging jobs that would devour pointless assets and incur pointless prices.

A key problem was validating Flink SQL queries with out deploying the total Flink runtime. After investigating Flink’s SQL implementation, Riskified found its dependency on Apache Calcite – a dynamic knowledge administration framework that handles SQL parsing, optimization, and question planning independently of information storage. This perception enabled utilizing Calcite straight for question validation earlier than job deployment.

It’s essential to know the way the info is structured to validate a Flink SQL question on a streaming supply like a Kafka subject. In any other case, surprising errors would possibly happen when trying to question the streaming supply. Though an anticipated schema is used with relational databases, it’s not enforced for streaming sources.

Schemas assure a deterministic construction for the info saved in a Kafka subject when utilizing a schema registry. A schema will be materialized right into a Calcite desk that defines how knowledge is structured within the Kafka subject. It permits inferring desk constructions straight from schemas (on this case, Avro format was used), enabling thorough field-level validation, together with sort checking and discipline existence, all earlier than job deployment. This desk can later be used to validate the SQL question.

The next code is an instance of supporting primary discipline sorts validation utilizing Calcite’s AbstractTable:

public class FlinkValidator {
    public static void validateSQL(String sqlQuery, Schema avroSchema) throws Exception {
        SqlParser.Config sqlConfig = SqlParser.config()
                .withCaseSensitive(true);
        SqlParser sqlParser = SqlParser.create(sqlQuery, sqlConfig);
        SqlNode parsedQuery = sqlParser.parseQuery();
        RelDataTypeFactory typeFactory = new SqlTypeFactoryImpl(RelDataTypeFactory.DEFAULT);
        CalciteSchema rootSchema = createSchemaWithAvro(avroSchema);
        SqlValidator validator = SqlValidatorUtil.newValidator(
                Frameworks.newConfigBuilder().construct().getOperatorTable(),
                rootSchema.createCatalogReader(Collections.emptyList(), typeFactory),
                typeFactory,
                SqlValidator.Config.DEFAULT
        );
        validator.validate(parsedQuery);
    }
    non-public static CalciteSchema createSchemaWithAvro(Schema avroSchema) {
        CalciteSchema rootSchema = CalciteSchema.createRootSchema(true);
        rootSchema.add("TABLE", new SimpleAvroTable(avroSchema));
        return rootSchema;
    }
    non-public static class SimpleAvroTable extends org.apache.calcite.schema.impl.AbstractTable {
        non-public closing Schema avroSchema;
        public SimpleAvroTable(Schema avroSchema) {
            this.avroSchema = avroSchema;
        }
        @Override
        public RelDataType getRowType(RelDataTypeFactory typeFactory) {
            RelDataTypeFactory.Builder builder = typeFactory.builder();
            for (Schema.Area discipline : avroSchema.getFields()) {
                builder.add(discipline.identify(), convertAvroType(discipline.schema(), typeFactory));
            }
            return builder.construct();
        }
        non-public RelDataType convertAvroType(Schema schema, RelDataTypeFactory typeFactory) {
            swap (schema.getType()) {
                case STRING:
                    return typeFactory.createSqlType(SqlTypeName.VARCHAR);
                case INT:
                    return typeFactory.createSqlType(SqlTypeName.INTEGER);
                default:
                    return typeFactory.createSqlType(SqlTypeName.ANY);
            }
        }
    }
}

You possibly can combine this validation method as an intermediate step earlier than creating the appliance. You possibly can create a streaming job programmatically with the AWS SDK, AWS Command Line Interface (AWS CLI), or Terraform. The validation happens earlier than submitting the streaming job.

Flink SQL and Confluent Avro knowledge sort mapping limitation

Flink supplies a number of APIs designed for various ranges of abstraction and person experience:

• Flink SQL sits on the highest stage, permitting customers to specific knowledge transformations utilizing acquainted SQL syntax, which is right for analysts and groups snug with relational ideas.
• The Desk API gives an analogous method however is embedded in Java or Python, enabling type-safe and extra programmatic expressions.
• For extra management, the DataStream API exposes low-level constructs to handle occasion time, stateful operations, and sophisticated occasion processing.
• On the most granular stage, the ProcessFunction API supplies full entry to Flink’s runtime options. It’s appropriate for superior use instances that demand detailed management over state and processing conduct.

Riskified initially used the Desk API to outline streaming transformations. Nevertheless, when deploying their first Flink job to a staging surroundings, they encountered serialization errors associated to the avro-confluent library and Desk API. Riskified’s schemas rely closely on Avro Enum sorts, which the avro-confluent integration doesn’t absolutely assist. Because of this, Enum fields had been transformed to Strings, resulting in mismatches throughout serialization and errors when trying to sink processed knowledge again to Kafka utilizing Flink’s Desk API.

Riskified developed another method to beat the Enum serialization limitations whereas sustaining schema necessities. They found that Flink’s DataStream API may appropriately deal with Confluent’s Avro data serialization with Enum fields, in contrast to the Desk API. They carried out a hybrid resolution combining each APIs as a result of the pipeline solely required SQL processing on the supply Kafka subject. It may possibly sink to the output with none further processing. The Desk API is used for knowledge processing and transformations, solely changing to the DataStream API on the closing output stage.

Managed Service for Apache Flink helps Flink APIs. It may possibly swap between the Desk API and the DataStream API.
A MapFunction can convert the Row sort of the Desk API right into a DataStream of GenericRecord. The MapFunction maps Flink’s Row knowledge sort into GenericRecord sorts by iterating over the Avro schema fields and constructing the GenericRecord from the Flink Row sort, casting the Row fields into the right knowledge sort in keeping with the Avro schema. This conversion is required to beat the avro-confluent library limitation with Flink SQL.

The next diagram and illustrates this workflow.

The next code is an instance question:

// SQL Question for filtering
Desk queryResults = tableEnv.sqlQuery(
       "SELECT * FROM InputTable");
// 1. Convert question outcomes from Desk API to a DataStream and use DataStream API to sink question outcomes to Kafka subject
DataStream rowStream = tableEnv.toDataStream(queryResults);
// Fetch the schema string from the schema registry
String schemaString = fetchSchemaString(schemaRegistryURL, schemaSubjectName);
// 2. Convert Row to GenericRecord with specific TypeInformation, utilizing customized AvroMapper
TypeInformation typeInfo = new GenericRecordAvroTypeInfo(avroSchema);
DataStream genericRecordStream = rowStream
       .map(new AvroMapper(schemaString))
       .returns(typeInfo); // Explicitly set TypeInformation
// 3. Outline Kafka sink utilizing ConfluentRegistryAvroSerializationSchema
KafkaSink kafkaSink = KafkaSink.builder()
       .setBootstrapServers(bootstrapServers)
       .setRecordSerializer(
               KafkaRecordSerializationSchema.builder()
                       .setTopic(sinkTopic)
                       .setValueSerializationSchema(
                               ConfluentRegistryAvroSerializationSchema.forGeneric(
                                       schemaSubjectName,
                                       avroSchema,
                                       schemaRegistryURL
                               )
                       )
                       .construct()
       )
       .construct();
// Sink to Kafka
genericRecordStream.sinkTo(kafkaSink);

CI/CD With Managed Service for Apache Flink

With Managed Service for Apache Flink, you may run a job by deciding on an Amazon Easy Storage Service (Amazon S3) key containing the appliance JAR. Riskified’s Flink code base was structured as a multi-module repository to assist further use instances in addition to supporting self-service SQL. Every Flink job supply code within the repository is an impartial Java module. The CI pipeline carried out a sturdy construct and deployment course of consisting of the next steps:

1. Construct and compile every module.
2. Run assessments.
3. Bundle the modules.
4. Add the artifact to the artifacts bucket twice: one JAR beneath -.jar and the second as -latest.jar, resembling a Docker registry like Amazon Elastic Container Registry (Amazon ECR). Managed Service for Apache Flink jobs makes use of the most recent tag artifact on this case. Nevertheless, a duplicate of outdated artifacts is saved for code rollback causes.

A CD course of follows this course of:

5. When merged, it lists all jobs for every module utilizing the AWS CLI for Managed Service for Apache Flink.
6. The applying JAR location is up to date for every utility, which triggers a deployment.
7. When the appliance is in a operating state with no errors, the next utility might be continued.

To permit secure deployment, this course of is completed steadily for each surroundings, beginning with the staging surroundings.

Self-service interface for submitting SQL jobs

Riskified believes an intuitive UI is essential for system adoption and effectivity. Nevertheless, creating a devoted UI for Flink job submission requires a realistic method, as a result of it may not be price investing in until there’s already an internet interface for inside improvement operations.

Investing in UI improvement ought to align with the group’s current instruments and workflows. Riskified had an inside internet portal for related operations, which made the addition of Flink job submission capabilities a pure extension of the self-service infrastructure.

An AWS SDK was put in on the internet server to permit interplay with AWS parts. The consumer receives person enter from the UI and interprets it into runtime properties to regulate the conduct of the Flink utility. The online server then makes use of the CreateApplication API motion to submit the job to Managed Service for Apache Flink.

Though an intuitive UI considerably enhances system adoption, it’s not the one path to accessibility. Alternatively, a well-designed CLI device or REST API endpoint can present the identical self-service capabilities.

The next diagram illustrates this workflow.

Manufacturing expertise: Flink’s implementation upsides

The transition to Flink and Managed Service for Apache Flink proved environment friendly in quite a few points:

• Schema evolution and knowledge dealing with – Riskified can both periodically fetch up to date schemas or restart purposes when schemas evolve. They’ll use current schemas with out self-registration.
• Useful resource isolation and administration – Managed Service for Apache Flink runs every Flink job as an remoted cluster, decreasing useful resource rivalry between jobs.
• Useful resource allocation and cost-efficiency – Managed Service for Apache Flink allows minimal useful resource allocation with automated scaling, proving to be extra cost-efficient.
• Job administration and circulate visibility – Flink supplies a cohesive knowledge circulate abstraction via its job and process mannequin. It manages the whole knowledge circulate in a single job and distributes the workload evenly over a number of nodes. This unified method allows higher visibility into the whole knowledge pipeline, simplifying monitoring, troubleshooting, and optimizing advanced streaming workflows.
• Constructed-in restoration mechanism – Managed Service for Apache Flink routinely creates checkpoints and savepoints that allow stateful Flink purposes to get well from failures and resume processing with out knowledge loss. With this characteristic, streaming jobs are sturdy and may get well safely from errors.
• Complete observability – Managed Service for Apache Flink exposes CloudWatch metrics that monitor Flink utility efficiency and statistics. You can even create alarms primarily based on these metrics. Riskfied determined to make use of the Cloudwatch Prometheus Exporter to export these metrics to Prometheus and construct PrometheusRules to align Flink’s monitoring to the Riskified commonplace, which makes use of Prometheus and Grafana for monitoring and alerting.

Subsequent steps

Though the preliminary focus was Kafka-to-Kafka streaming queries, Flink’s big selection of sink connectors gives the potential for pluggable multi-destination pipelines. This versatility is on Riskfied’s roadmap for future enhancements.

Flink’s DataStream API supplies capabilities that reach far past self-serving streaming SQL capabilities, opening new avenues for extra refined fraud detection use instances. Riskified is exploring methods to make use of DataStream APIs to reinforce ecommerce fraud prevention methods.

Conclusions

On this submit, we shared how Riskified efficiently transitioned from ksqlDB to Managed Service for Apache Flink for its self-serve streaming SQL engine. This transfer addressed key challenges like schema evolution, useful resource isolation, and pipeline administration. Managed Service for Apache Flink gives options akin to together with remoted jobs environments, automated scaling, and built-in monitoring, which proved extra environment friendly and cost-effective. Though Flink SQL limitations with Kafka required workarounds, utilizing Flink’s DataStream API and user-defined features resolved these points. The transition has paved the best way for future enlargement with multi-targets and superior fraud detection capabilities, solidifying Flink as a sturdy and scalable resolution for Riskified’s streaming wants.

If Riskified’s journey has sparked your curiosity in constructing a self-service streaming SQL platform, right here’s the way to get began:

• Study extra about Managed Service for Apache Flink:
• Get hands-on expertise:

Concerning the authors

Gal Krispel is a Knowledge Platform Engineer at Riskified, specializing in streaming applied sciences akin to Apache Kafka and Apache Flink. He focuses on constructing scalable, real-time knowledge pipelines that energy Riskified’s core merchandise. Gal is especially concerned with making advanced knowledge architectures accessible and environment friendly throughout the group. His work spans real-time analytics, event-driven design, and the seamless integration of stream processing into large-scale manufacturing methods.

Sofia Zilberman works as a Senior Streaming Options Architect at AWS, serving to prospects design and optimize real-time knowledge pipelines utilizing open-source applied sciences like Apache Flink, Kafka, and Apache Iceberg. With expertise in each streaming and batch knowledge processing, she focuses on making knowledge workflows environment friendly, observable, and high-performing.

Lorenzo Nicora works as Senior Streaming Resolution Architect at AWS, serving to prospects throughout EMEA. He has been constructing cloud-centered, data-intensive methods for over 25 years, working throughout industries each via consultancies and product firms. He has used open-source applied sciences extensively and contributed to a number of tasks, together with Apache Flink, and is the maintainer of the Flink Prometheus connector.