Status

Current state: "Under Discussion"

Discussion thread: To be added

JIRA: To be added

Released: To be decided

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

Table of Contents

Motivation

In order to execute a machine learning algorithm using Flink as the underlying runtime, Flink needs to support the iteration primitive, such that some outputs of a Flink job subgraph can be fed back to the inputs of the iteration body and this loop continuous until some termination criteria is reached.

Flink currently provides DataSet::iterate(...) and DataStream::iterate(...) to support the iteration primitive described above. However, neither API can be used to support iteration on both bounded and unbounded data streams, for the following reasons:

The DataSet::iterate(...) only supports iteration on the bounded data streams. And as described in FLIP-131, we will deprecate the DataSet API in favor of the Table API/SQL in the future.
The DataStream::iterate(...) has a few design issues that prevents it from being reliably used in production jobs. Many of these issues, such as possibility of deadlock, are described in FLIP-15.

This FLIP proposes to add a couple APIs in the flink-ml repository to achieve the following goals:

Provide solution for all the iteration use-cases (see the use-case section below for more detail) supported by the existing APIs, without the issues described above.
Provide solution for a few additional use-cases (e.g. bounded streams + async mode + per-round variable update) not supported by the existing APIs.

Note that we have chosen to put the iteration API (and its implementation) in the flink-ml repository instead of the DataStream class in the Flink core repository, because we believe it is important to keep the Flink core runtime as simple and maintainable as possible.

Besides supporting the basic iteration primitive (i.e. the feedback stream), Flink also needs APIs to support synchronization between parallel execution of the iteration body. This is needed to execute machine learning algorithms that need to be parallelized and still ensure deterministic execution results.

In the following, we describe the example use-cases and the exact semantics that we aim to support via the proposed APIs.

Terminology

We explain a few terminologies in the following to facilitate the reading of this doc.

1) Iteration body

An iteration body is a subgraph of a Flink job graph that implements the iteration logic of e.g. a machine learning algorithm. In particular, the iteration body will create feedback variable streams and emit values into those streams. The feedback variable streams will be joined with the user-provided initial variable streams and the resulting streams will be ingested into the iteration body.

...

The target use-cases (i.e. algorithms) can be described w.r.t. the categories described below. In the following, we first describe the categories, followed by the combinations of those categories supported by the existing APIs and the proposed APIs, respectively.

Categories of Algorithms

Different algorithms might have different requirements for the input datasets (bounded or unbounded), synchronization between parallel subtasks (sync or async), amount of data processed for every variable update (a batch/subset or the entire dataset). We describe each of these requirements below.

1) Algorithms have different needs for whether the input data streams should be bounded or unbounded. We classify those algorithms into online algorithm and offline algorithms as below.

For online training algorithms, the training samples will be unbounded streams of data. The corresponding iteration body should ingest these unbounded streams of data, read each value in each stream once, and update machine learning model repeatedly in near real-time. The iteration will never terminate in this case. The algorithm should be executed as a streaming job.

For offline training algorithms, the training samples will be bounded streams of data. The corresponding iteration body should read these bounded data streams for arbitrary number of rounds and update machine learning model repeatedly until a termination criteria is met (e.g. a given number of rounds is reached or the model has converged). The algorithm should be executed as a batch job.

2) Algorithms (either online or offline) have different needs of how their parallel subtasks.

In the sync mode, parallel subtasks, which execute the iteration body, update the model variables in a coordinated manner. There exists global epoch epochs, such that all subtasks read the shared model variables at the beginning of an epoch, calculate variable updates based on the fetched variable values, and write updates of the variable values at the end of this epoch.

In the async mode, each parallel subtask, which execute the iteration body, could read/update the shared model variables without waiting for variable updates from other subtask. For example, a subtask could have updated model variables 10 times when another subtask has updated model variables only 3 times.

The sync mode is useful when an algorithm should be executed in a deterministic way to achieve best possible accuracy, and the straggler issue (i.e. there is subtask which is considerably slower than others) does not cause slow down the algorithm execution too much. In comparison, the async mode is useful for algorithms which want to be parallelized and executed across many subtasks as fast as possible, without worrying about performance issue caused by stragglers, at the possible cost of reduced accuracy.

3) An algorithm may have additional requirements in how much data should be consumed each time before a subtask can update variables. There are two categories of choices here:

Per-batch variable update: The algorithm wants to update variables every time an arbitrary subset of the user-provided data streams (either bounded or unbounded) is processed.

Per-round variable update: The algorithm wants to update variables every time all data of the user-provided bounded data streams is processed.

In the machine learning domain, some algorithms allow users to configure a batch size and the model will be updated every time each subtask processes a batch of data. Those algorithms fits into the first category. And such an algorithm can be either online or offline.

Other algorithms only update variables every time the entire data is consumed for one round. Those algorithms fit into the second category. And such an algorithm must be offline because, by this definition, the user-provided dataset must be bounded.

Combinations of categories supported by the existing APIs

The existing DataSet::iterate supports algorithms that process bounded streams + sync mode + per-round variable update.

The existing DataStream::iterate has a few bugs that prevent it from being used in production yet (see FLIP-15). Other than this, it expects to support algorithms that process unbounded streams + async mode + per-batch variable update.

Combinations of categories supported by the proposed APIs

The proposed APIs support 6 out of the 8 combinations of the above categories. The following 2 combinations are not supported because, by definition, "per-round variable update" can only be used with bonded data streams.

Public Interfaces

We propose to add the following interfaces and utility classes in the flink-ml repository.

Code Block

language	java
title	The IterationBody API
linenumbers	true

public class DataStreamList {
    public DataStreamList(DataStream<?>... dataStreams) {...}

    // Returns the number of data streams in this list.
    public int size() {...}

    // Returns the data stream at the given index in this list.
    @SuppressWarnings("unchecked")
    public <T> DataStream<T> get(int index) {...}
}

@PublicEvolving
public interface IterationBody {
    /**
     * This method creates the graph for the iteration body.
     *
     * See Utils::iterate, Utils::iterateBoundedStreams and Utils::iterateAndReplayBoundedStreams for how the iteration
     * body can be executed and when execution of the corresponding graph should terminate.
     *
     * Required: the number of feedback variable streams returned by this method must equal the number of variable
     * streams given to this method.
     *
     * @param variableStreams the variable streams.
     * @param dataStreams the data streams.
     * @return a IterationBodyResult.
     */
    IterationBodyResult process(DataStreamList variableStreams, DataStreamList dataStreams);
}


/**
 * A helper class that contains the streams returned by the iteration body.
 */
class IterationBodyResult {
    /**
     * A list of feedback variable streams. These streams will only be used during the iteration execution and will
     * not be returned to the caller of the iteration body. It is assumed that the method which executes the
     * iteration body will feed the records of the feedback variable streams back to the corresponding input variable
     * streams.
     */
    DataStreamList feedbackVariableStreams;

    /**
     * A list of output streams. These streams will be returned to the caller of the methods that execute the
     * iteration body.
     */
    DataStreamList outputStreams;

    /**
     * An optional termination criteria stream. If this stream is not null, it will be used together with the
     * feedback variable streams to determine when the iteration should terminate.
     */
    Optional<DataStream<?>> terminationCriteria;
}


/**
 * The callbacks defined below will be invoked only if the operator instance which implements this interface is used
 * within an iteration body.
 */
@PublicEvolving
public interface IterationListener<T> {
    /**
     * This callback is invoked every time the epoch watermark increments. The initial epoch watermark is -1.
     *
     * The epochWatermark is the maximum integer that meets this requirement: every record that arrives at the operator
     * going forward should have an epoch larger than the epochWatermark. See Java docs in IterationUtils for how epoch
     * is determined for records ingested into the iteration body and for records emitted by operators within the
     * iteration body.
     *
     * If all inputs are bounded, the maximum epoch of all records ingested into this operator is used as the
     * epochWatermark by the last invocation of this callback.
     *
     * @param epochWatermark The incremented epoch watermark.
     * @param context A context that allows emitting side output. The context is only valid during the invocation of
     *                this method.
     * @param collector The collector for returning result values.
     */
    void onEpochWatermarkIncremented(int epochWatermark, Context context, Collector<T> collector);

    /**
     * This callback is invoked after the execution of the iteration body has terminated.
     *
     * See Java doc of methods in IterationUtils for the termination conditions.
     *
     * @param context A context that allows emitting side output. The context is only valid during the invocation of
     *                this method.
     * @param collector The collector for returning result values.
     */
    void onIterationTermination(Context context, Collector<T> collector);

    /**
     * Information available in an invocation of the callbacks defined in the IterationProgressListener.
     */
    interface Context {
        /**
         * Emits a record to the side output identified by the {@link OutputTag}.
         *
         * @param outputTag the {@code OutputTag} that identifies the side output to emit to.
         * @param value The record to emit.
         */
        <X> void output(OutputTag<X> outputTag, X value);
    }
}

/**
 * A helper class to apply {@link IterationBody} to data streams.
 */
@PublicEvolving
public class IterationUtils {
    /**
     * This method can use an iteration body to process records in unbounded data streams.
     *
     * This method invokes the iteration body with the following parameters:
     * 1) The 1st parameter is a list of input variable streams, which are created as the union of the initial variable
     * streams and the corresponding feedback variable streams (returned by the iteration body).
     * 2) The 2nd parameter is the data streams given to this method.
     *
     * The epoch values are determined as described below. See IterationListener for how the epoch values are used.
     * 1) All records in the initial variable streams has epoch=0.
     * 2) All records in the data streams has epoch=MAX_LONG. In this case, records in the data stream won't affect
     * any operator's low watermark.
     * 3) For any record emitted by this operator into a non-feedback stream, the epoch of this emitted record = the
     * epoch of the input record that triggers this emission. If this record is emitted by onEpochLWIncremented(), then
     * the epoch of this record = incrementedEpochLW - 1.
     * 4) For any record emitted by this operator into a feedback variable stream, the epoch of the emitted record =
     * min(the epoch of the input record that triggers this emission, MAX_LONG - 1) + 1. If this record is emitted by
     * onEpochLWIncremented(), then the epoch of this record = incrementedEpochLW.
     *
     * The execution of the graph created by the iteration body will not terminate by itself. This is because at least
     * one of its data streams is unbounded.
     *
     * Required:
     * 1) All the init variable streams must be bounded.
     * 2) There is at least one unbounded stream in the data streams list.
     * 3) The parallelism of any stream in the initial variable streams must equal the parallelism of the stream at the
     * same index of the feedback variable streams returned by the IterationBody.
     *
     * @param initVariableStreams The initial variable streams. These streams will be merged with the feedback variable
     *                            streams before being used as the 1st parameter to invoke the iteration body.
     * @param dataStreams The data streams. These streams will be used as the 2nd parameter to invoke the iteration
     *                    body.
     * @param body The computation logic which takes variable/data streams and returns variable/output streams.
     * @return The list of output streams returned by the iteration boy.
     */
    static DataStreamList iterateUnboundedStreams(DataStreamList initVariableStreams, DataStreamList dataStreams, IterationBody body) {...}

    /**
     * This method can use an iteration body to process records in some bounded data streams iteratively until a
     * termination criteria is reached (e.g. the given number of rounds is completed or no further variable update is
     * needed). Because this method does not replay records in the data streams, the iteration body needs to cache those
     * records in order to visit those records repeatedly.
     *
     * This method invokes the iteration body with the following parameters:
     * 1) The 1st parameter is a list of input variable streams, which are created as the union of the initial variable
     * streams and the corresponding feedback variable streams (returned by the iteration body).
     * 2) The 2nd parameter is the data streams given to this method.
     *
     * The epoch values are determined as described below. See IterationListener for how the epoch values are used.
     * 1) All records in the initial variable streams has epoch=0.
     * 2) All records in the data streams has epoch=0.
     * 3) For any record emitted by this operator into a non-feedback stream, the epoch of this emitted record = the
     * epoch of the input record that triggers this emission. If this record is emitted by onEpochLWIncremented(), then
     * the epoch of this record = incrementedEpochLW - 1.
     * 4) For any record emitted by this operator into a feedback variable stream, the epoch of the emitted record = the
     * epoch of the input record that triggers this emission + 1. If this record is emitted by onEpochLWIncremented(),
     * then the epoch of this record = incrementedEpochLW.
     *
     * Suppose there is a coordinator operator which takes all feedback variable streams (emitted by the iteration body)
     * and the termination criteria stream (if not null) as inputs. The execution of the graph created by the
     * iteration body will terminate when any of the following conditions is met:
     * 1) The termination criteria stream is not null. And the coordinator operator has not observed any new value from
     * the termination criteria stream between two consecutive onEpochLWIncremented invocations.
     * 2) The coordinator operator has not observed any new value from any feedback variable stream between two
     * consecutive onEpochLWIncremented invocations.
     *
     * Required:
     * 1) All the init variable streams and the data streams must be bounded.
     * 2) The parallelism of any stream in the initial variable streams must equal the parallelism of the stream at the
     * same index of the feedback variable streams returned by the IterationBody.
     *
     * @param initVariableStreams The initial variable streams. These streams will be merged with the feedback variable
     *                            streams before being used as the 1st parameter to invoke the iteration body.
     * @param dataStreams The data streams. These streams will be used as the 2nd parameter to invoke the iteration
     *                    body.
     * @param body The computation logic which takes variable/data streams and returns variable/output streams.
     * @return The list of output streams returned by the iteration boy.
     */
    static DataStreamList iterateBoundedStreamsUntilTermination(DataStreamList initVariableStreams, DataStreamList dataStreams, IterationBody body) {...}

    /**
     * This method can use an iteration body to process records in some bounded data streams iteratively until a
     * termination criteria is reached (e.g. the given number of rounds is completed or no further variable update is
     * needed). Because this method replays records in the data streams, the iteration body does not need to cache those
     * records to visit those records repeatedly.
     *
     * This method invokes the iteration body with the following parameters:
     * 1) The 1st parameter is a list of input variable streams, which are created as the union of the initial variable
     * streams and the corresponding feedback variable streams (returned by the iteration body).
     * 2) The 2nd parameter is a list of replayed data streams, which are created by replaying the initial data streams
     * round by round until the iteration terminates. The records in the Nth round will be emitted into the iteration
     * body only if the low watermark of the first operator in the iteration body >= N - 1.
     *
     * The epoch values are determined as described below. See IterationListener for how the epoch values are used.
     * 1) All records in the initial variable streams has epoch=0.
     * 2) The records from the initial data streams will be replayed round by round into the iteration body. The records
     * in the first round have epoch=0. And records in the Nth round have epoch = N - 1.
     * 3) For any record emitted by this operator into a non-feedback stream, the epoch of this emitted record = the
     * epoch of the input record that triggers this emission. If this record is emitted by onEpochLWIncremented(), then
     * the epoch of this record = incrementedEpochLW - 1.
     * 4) For any record emitted by this operator into a feedback stream, the epoch of the emitted record = the epoch
     * of the input record that triggers this emission + 1. If this record is emitted by onEpochLWIncremented(), then
     * the epoch of this record = incrementedEpochLW.
     *
     * Suppose there is a coordinator operator which takes all feedback variable streams (emitted by the iteration body)
     * and the termination criteria stream (if not null) as inputs. The execution of the graph created by the
     * iteration body will terminate when any of the following conditions is met:
     * 1) The termination criteria stream is not null. And the coordinator operator has not observed any new value from
     * the termination criteria stream between two consecutive onEpochLWIncremented invocations.
     * 2) The coordinator operator has not observed any new value from any feedback variable stream between two
     * consecutive onEpochLWIncremented invocations.
     *
     * Required:
     * 1) All the init variable streams and the data streams must be bounded.
     * 2) The parallelism of any stream in the initial variable streams must equal the parallelism of the stream at the
     * same index of the feedback variable streams returned by the IterationBody.
     *
     * @param initVariableStreams The initial variable streams. These streams will be merged with the feedback variable
     *                            streams before being used as the 1st parameter to invoke the iteration body.
     * @param initDataStreams The initial data streams. Records from these streams will be repeatedly replayed and used
     *                        as the 2nd parameter to invoke the iteration body.
     * @param body The computation logic which takes variable/data streams and returns variable/output streams.
     * @return The list of output streams returned by the iteration boy.
     */
    static DataStreamList iterateAndReplayBoundedStreamsUntilTermination(DataStreamList initVariableStreams, DataStreamList initDataStreams, IterationBody body) {...}

}

The Goals of the Iteration Library

The Iteration Body and Round Semantics

At the iteration API level, we would need the corresponding concept corresponding to Epoch and Batch. We would call processing one epoch as a round: users would specify a subgraph as the body of the iteration to specify how to calculate the update, after the iteration body process the whole dataset for one time (namely one Epoch). Apparently the round is meaningful only for the bounded cases.

Figure 1. The structure of an iteration body.

To process the inputs for multiple rounds, we would need feedback edges to emit the outputs of the last round to the inputs of the iteration body, and we union the initial inputs and the feedbacks. For bounded dataset when processing the first epoch, the data would be from the initial input edges and when processing the remaining epochs, the data would be from the feedback edges.

There are also inputs to the iteration body that do not have feedbacks. For example, an ML algorithm might have two inputs, one is the initialized model and the other is the training data. The input corresponding to the initial model will have a feedback after each epoch about the update to the model, but the training data would not need to be updated.

The iteration body also have one or multiple output streams. The iteration body might output records at each round, and the records emitted in all the rounds composed the final outputs.

Therefore, an iteration body composed of

Variable inputs with feedbacks.
Constant inputs without feedbacks.
Outputs.

Synchronization

Since for the bounded dataset, all the algorithms, to the best of out extend, are all able to be converted into epoch-based algorithms, thus we could only support the synchronization between epoch, namely between rounds.

How to

Besides, the previous DataStream and DataSet iteration APIs also have some caveats to support algorithm implementation:

Lack of the support for multiple inputs, arbitrary outputs and nested iteration for both iteration APIs, which is required by scenarios like Metapath (multiple-inputs), boost algorithms (nested iteration) or when we want to output both loss and model (multiple-outputs). In the new iteration we would support these functionalities.
Lack of asynchronous iteration support for the DataSet iteration, which is required by algorithms like asynchronous linear regression, in the new iterations we would support both synchronous and asynchronous modes for the bounded iteration.
The current DataSet iteration by default provides a "for each round" semantics, namely users only need to specify the computation logic in each round, and the framework would executes the subgraph multiple times until convergence. To cooperate with the semantics, the DataSet iteration framework would merge the initial input and the feedback (bulk style and delta style), and replay the datasets comes from outside of the iteration. This method is easier to use, but it also limit some possible optimizations.

We also would like to improve these caveats in the new iteration library.

Overall Design

To reduce the development and maintenance overhead, it would be preferred to have a unified implementation for different types of iterations. In fact, the different iteration types shares the same requirements in runtime implementation:

All the iteration types should support multiple inputs and multiple outputs.
All the iteration types require some kind of back edges that transfer the data back to the iteration head. Since Flink does not support cycles in scheduler and network stack, the back edges should not be visible in the StreamGraph and JobGraph.
All the iteration should support checkpoints mechanism in Stream execution mode.

Different types of iterations differ in their requirements for Progress tracking. Progress tracking is analogous to the watermark outside the iteration and it tracks the “progress” inside the iteration:

For bounded iteration, we could track if we have processed all the records for a specific round. This is necessary for operators like aggregation inside the iteration: if it is notified all the records of the current round is processed, it could output the result of this round. We could also track if the whole iteration is end, namely all the inputs are finished and no pending records inside the iteration.
For unbounded iteration, there is no concept of global rounds, and the only progress tracking is at the end of iteration.

The difference of the progress tracking would also affect the API. For example, for bounded iteration, we could allow users to specify the termination condition based on number of rounds, but it is meaningless for the unbounded iteration.

To make the API easy to use, we propose to have dedicated API for different types of iteration, and underlying we will translate them onto the same framework. would implements the basic functionality like iteration StreamGraph building, runtime structure and checkpoint, and it allows to implement different iterations to implement different types of progress tracking support.

Public Interfaces

As shown in Figure 1, an iteration is composed of

The inputs from outside of the iteration.
An iteration body specify the structure inside the iteration.
1. The subgraph inside the iteration.
2. Some input have corresponding feedbacks to update the underlying data stream. The feedbacks are union with the corresponding inputs: the original inputs are emitted into the iteration body for only once, and the feedbacks are also emitted to the same set of operators.
3. The outputs going out of the iteration. The outputs could be emitted from arbitrary data stream.

Unbounded Iteration

Similar to FLIP-15, we would more tend to provide a structural iteration API to make it easier to be understand. With this method, users are required to specify an IterationBody that generates the part of JobGraph inside the iteration. The iteration body should specify the DAG inside the iteration, and also the list of feedback streams and the output streams. The feedback streams would be union with the corresponding inputs and the output streams would be provided to the caller routine.

However, since we do not know the accurate number and type of input streams, it is not easy to define a unified interface for the iteration body without type casting. Thus we would propose to use the annotation to allows for arbitrary number of inputs:

The interface for the unbounded iteration is straightforward:

To avoid more data is read from the inputs while too much data accumulate inside the iteration, the iteration would first process the feedback data if both side of data is available.

For termination detection, the iteration would continue until

All the inputs are terminated.
And there is no records inside the iteration subgraph.

Then the iteration terminates.

Bounded Iteration

As mentioned in the motivation, the existing dataset iteration API uses the "per-round" semantics: it views the iteration as a repeat execution of the same DAG, thus underlying it would automatically merge the inputs and feedbacks and replay the inputs without feedbacks, and the operators inside the iteration live only for one-round. This might cause bad performance for some algorithms who could cache these data in a more efficient way.

To avoid this issue, similar to the unbounded iteration, by default we use the "per-iteration" semantics:

Operators inside the iteration would live till the whole iteration is finished.
We do not automatically merge the inputs and feedbacks. Instead, we union the original inputs and the feedbacks so that users could decide how the merge them.
We do not replay the inputs without feedbacks. Users could decide to how to cache them more efficiently.

Besides, to cooperate with the "per-round" semantics, previously the iteration is by default synchronous: before the current round fully finished, the feedback data is cached and would not be emitted. Thus it could not support some algorithms like asynchronous regression. To cope with this issue, we view synchronous iteration as a special case of asynchronous iteration with additional synchronization. Thus by default the iteration is asynchronous.

Based on the above assumption, the API to add iteration to a job is nearly the same compared to the unbounded iteration. The only difference is that bounded iteration supports more sophisticated termination conditions: a function is evaluated when each round ends based on the round or the records of a specified data stream. If it returns true, the iteration would deserts all the following feedback records, ends all the ongoing rounds and finish.

Since now the operators would live across multiple rounds and multiple rounds might be concurrent, the operators inside the iteration needs to know the rounds of the current record and when one round is fully finished, namely the progress tracking. For example, an operator computes the sum of the records in each rounds would like to add the record to the corresponding partial sum, and when one round is finished, it would emit the sum for this round. To support the progress track, UDFs / operators inside the iteration could implementation `BoundedIterationProgressListener` to acquire the additional information about the progress.

Based on the progress tracking interface, if users want to implement a synchronous method, some operators inside the subgraph needs to be synchronous: they only emits the records in `onRoundEnd`, namely after all the data of the current round is received. If for the subgraph of iteration body, every path from input to the feedbacks has at least such an operator, then the iteration would be synchronous.

For users still want to use the iteration with the "per-round" semantics, a utility `forEachRound()` is provided. With the utility users could add a subgraph inside the iteration body that

The operators inside the subgraph would live only for one round.
If an input stream without feedback is referenced, the input stream would be replayed for each round.

For input stream with feedbacks, we also provide two utility processFunction that automatically merge the original inputs and feedbacks. Both the existing bulk and delta method is supported. Then users would be able to implement a per-round iteration with input.process(bulkCache()).forEachRound(() → {...}).

The API for the bounded iteration is as follows:

Key Implementation

The runtime physical structure of the iteration is shown in Figure 1, which is similar to the current implementation. The head & tail is added by the framework. They would be colocated so that we could implement the feedback edge with the local queue. The head could coordinator with an operator coordinator bind to a virtual operator ID for synchronization, including progress tracking and termination condition calculating.

Figure 1. The physical runtime structure for the iteration.

To support the progress tracking, we would introduce new events inside the iteration body, like how watermark is implemented. However, since the normal operators could not identify these event, we would wrap the operators inside the iteration to parse these events.

To wrap the operators for the part of DAG inside the iteration, when building the stream graph we would introduce a mock execution environment and build the iteration DAG inside this environment first, then when apply() method is called, we would translate the DAG into the real execution environment with the suitable wrapper. Besides, all the edges inside the iteration should be PIPELINE, we would also set the edge property when translating.

The operator wrapper needs to simulates the context that an operator executes. Specially, for operators with single-round lifecycle in bounded iteration, we would need to isolate the states used for each round and cleanup the corresponding state after the round end.

Examples

This sections shows how general used ML algorithms could be implemented with the iteration API.

Offline Training with Bounded Iteration

We would like to first show the usage of the bounded iteration with the linear regression case: the model is Y = XA, and we would like to acquire the best estimation of A with the SGD algorithm. To simplify we assume the parameters could be held in the memory of one task.

The job graph of the algorithm could be shown in the Figure 3: in each round, we use the latest parameters to calculate the update to the parameters: ΔA = ∑(Y - XA)X. To achieve this, the Parameters vertex would broadcast the latest parameters to the Train vertex. Each subtask of the Train vertex holds a part of dataset. Follow the sprite of SGD, it would sample a small batch of training records, and calculate the update with the above equation. Then the Train vertex emit ΔA to the Parameters node to update the parameters.

Figure 3. The JobGraph for the offline training of the linear regression case.

We will start with the synchronous training. The synchronous training requires the updates from all the Train vertex subtask is merged before the next round of training. It could be done by only emit the next round of parameters on the end of round. The code is shown as follows:

Code Block

language	java
linenumbers	true

public class SynchronousBoundedLinearRegression {
    private static final N_DIM = 50;
    private static final OutputTag<double[]> FINAL_MODEL_OUTPUT_TAG = new OutputTag<double[]>{};

    public static void main(String[] args) {
        DataStream<double[]> initParameters = loadParameters().setParallelism(1);
        DataStream<Tuple2<double[], Double>> dataset = loadDataSet().setParallelism(1);

        int batch = 5;
        int epochEachBatch = 10;

        ResultStreams resultStreams = new BoundedIteration()
            .withBody(new IterationBody(
                @IterationInput("model") DataStream<double[]> model,
                @IterationInput("dataset") DataStream<Tuple2<double[], Double>> dataset
            ) {
                SingleOutputStreamOperator<double[]> parameters = model.process(new ParametersCacheFunction());
                DataStream<double[]> modelUpdate = parameters.setParallelism(1)
                    .broadcast()
                    .connect(dataset)
                    .coProcess(new TrainFunction())
                    .setParallelism(10)

                return new BoundedIterationDeclarationBuilder()
                    .withFeedback("model", modelUpdate)
                    .withOutput("final_model", parameters.getSideOut(FINAL_MODEL_OUTPUT_TAG))
                    .until(new TerminationCondition(null, context -> context.getRound() >= batch * epochEachBatch))
                    .build();
            })
            .build();
        
        DataStream<double[]> finalModel = resultStreams.get("final_model");
        finalModel.print();
    }

    public static class ParametersCacheFunction extends ProcessFunction<double[], double[]>
        implements BoundedIterationProgressListener<double[]> {  
        
        private final double[] parameters = new double[N_DIM];

        public void processElement(double[] update, Context ctx, Collector<O> output) {
            // Suppose we have a util to add the second array to the first.
            ArrayUtils.addWith(parameters, update);
        }

        public void onRoundEnd(int[] round, Context context, Collector<T> collector) {
            collector.collect(parameters);
        }

        public void onIterationEnd(int[] round, Context context) {
            context.output(FINAL_MODEL_OUTPUT_TAG, parameters);
        }
    }

    public static class TrainFunction extends CoProcessFunction<double[], Tuple2<double[], Double>, double[]> implements BoundedIterationProgressListener<double[]> {

        private final List<Tuple2<double[], Double>> dataset = new ArrayList<>();
        private double[] firstRoundCachedParameter;

        private Supplier<int[]> recordRoundQuerier;

        public void setCurrentRecordRoundsQuerier(Supplier<int[]> querier) {
            this.recordRoundQuerier = querier;
        } 

        public void processElement1(double[] parameter, Context context, Collector<O> output) {
            int[] round = recordRoundQuerier.get();
            if (round[0] == 0) {
                firstRoundCachedParameter = parameter;
                return;
            }

            calculateModelUpdate(parameter, output);
        }

        public void processElement2(Tuple2<double[], Double> trainSample, Context context, Collector<O> output) {
            dataset.add(trainSample)
        }

        public void onRoundEnd(int[] round, Context context, Collector<T> output) {
            if (round[0] == 0) {
                calculateModelUpdate(firstRoundCachedParameter, output);
                firstRoundCachedParameter = null;                
            }
        }

        private void calculateModelUpdate(double[] parameters, Collector<O> output) {
            List<Tuple2<double[], Double>> samples = sample(dataset);

            double[] modelUpdate = new double[N_DIM];
            for (Tuple2<double[], Double> record : samples) {
                double diff = (ArrayUtils.muladd(record.f0, parameters) - record.f1);
                ArrayUtils.addWith(modelUpdate, ArrayUtils.multiply(record.f0, diff));
            }

            output.collect(modelUpdate);
        }
    }
}

If instead we want to do asynchronous training, we would need to do the following change:

The Parameters vertex would not wait till round end to ensure received all the updates from the iteration. Instead, it would immediately output the current parameters values once it received the model update from one train subtask.
To label the source of the update, we would like to change the input type to be Tuple2<Integer, double[]>. The Parameters would only output the new parameters values to the Train task that send the update.

We omit the change to the graph building code since the change is trivial (change the output type and the partitioner to be customized one). The change to the Parameters vertex is the follows:

Code Block

language	java

public static class ParametersCacheFunction extends ProcessFunction<Tuple2<Integer, double[]>, Tuple2<Integer, double[]>>
    implements BoundedIterationProgressListener<double[]> {  
    
    private final double[] parameters = new double[N_DIM];

    public void processElement(Tuple2<Integer, double[]> update, Context ctx, Collector<Tuple2<Integer, double[]>> output) {
        // Suppose we have a util to add the second array to the first.
        ArrayUtils.addWith(parameters, update);
        output.collect(new Tuple2<>(update.f0, parameters))
    }

    public void onIterationEnd(int[] round, Context context) {
        context.output(FINAL_MODEL_OUTPUT_TAG, parameters);
    }
}

Online Training with Unbounded Iteration

Suppose now we would change the algorithm to unbounded iteration, compared to the offline, the differences is that

The dataset is unbounded. The Train operator could not cache all the data in the first round.
The training algorithm might be changed to others like FTRL. But we keep using SGD in this example since it does not affect showing the usage of the iteration.

We also start with the synchronous case. for online training, the Train vertex usually do one update after accumulating one mini-batch. This is to ensure the distribution of the samples is similar to the global statistics. In this example we omit the complex data re-sample process and just fetch the next several records as one mini-batch.

The JobGraph for online training is still shown in Figure 1, with the training dataset become unbounded. Similar to the bounded cases, for the synchronous training, the process would be expected like

The Parameters broadcast the initialized values on received the input values.
All the Train task read the next mini-batch of records, Calculating an update and emit to the Parameters vertex. Then it would wait till received update parameters from the Parameters Vertex before it head to process the next mini-batch.
The Parameter vertex would wait received the updates from all the Train tasks before it broadcast the updated parameters.

Since in the unbounded case there is not the concept of round, and we do update per-mini-batch, thus we could instead use the InputSelectable functionality to implement the algorithm:

Code Block

language	java

public class SynchronousUnboundedLinearRegression {
    private static final N_DIM = 50;
    private static final OutputTag<double[]> MODEL_UPDATE_OUTPUT_TAG = new OutputTag<double[]>{};

    public static void main(String[] args) {
        DataStream<double[]> initParameters = loadParameters().setParallelism(1);
        DataStream<Tuple2<double[], Double>> dataset = loadDataSet().setParallelism(1);

        ResultStreams resultStreams = new UnboundedIteration()
            .withBody(new IterationBody(
                @IterationInput("model") DataStream<double[]> model,
                @IterationInput("dataset") DataStream<Tuple2<double[], Double>> dataset
            ) {
                SingleOutputStreamOperator<double[]> parameters = model.process(new ParametersCacheFunction(10));
                DataStream<double[]> modelUpdate = parameters.setParallelism(1)
                    .broadcast()
                    .connect(dataset)
                    .transform(
                                "operator",
                                BasicTypeInfo.INT_TYPE_INFO,
                                new TrainOperators(50));
                    .setParallelism(10);

                return new UnBoundedIterationDeclarationBuilder()
                    .withFeedback("model", modelUpdate)
                    .withOutput("model_update", parameters.getSideOut(FINAL_MODEL_OUTPUT_TAG))
                    .build();
            })
            .build();
        
        DataStream<double[]> finalModel = resultStreams.get("model_update");
        finalModel.addSink(...)
    }

    public static class ParametersCacheFunction extends ProcessFunction<double[], double[]> {  
        
        private final int numOfTrainTasks;

        private final int numOfUpdatesReceived = 0;
        private final double[] parameters = new double[N_DIM];

        public ParametersCacheFunction(int numOfTrainTasks) {
            this.numOfTrainTasks = numOfTrainTasks;
        }

        public void processElement(double[] update, Context ctx, Collector<O> output) {
            // Suppose we have a util to add the second array to the first.
            ArrayUtils.addWith(parameters, update);
            numOfUpdatesReceived++;

            if (numOfUpdatesReceived == numOfTrainTasks) {
                output.collect(parameters);
                numOfUpdatesReceived = 0;
            }
        }
    }

    public static class TrainOperators extends AbstractStreamOperator<double[]> implements TwoInputStreamOperator<double[], Tuple2<double[], Double>, double[]>, InputSelectable {

        private final int miniBatchSize;

        private final List<Tuple2<double[], Double>> miniBatch = new ArrayList<>();
        private double[] firstRoundCachedParameter;

        public TrainOperators(int miniBatchSize) {
            this.miniBatchSize = miniBatchSize;
        }

        public void processElement1(double[] parameter, Context context, Collector<O> output) {
            calculateModelUpdate(parameter, output);
			miniBatchSize.clear();
        }

        public void processElement2(Tuple2<double[], Double> trainSample, Context context, Collector<O> output) {
            dataset.add(trainSample);
        }

        public InputSelection nextSelection() {
            if (miniBatch.size() < miniBatchSize) {
                return InputSelection.SECOND;
            } else {
                return InputSelection.FIRST;
            }
        }

        private void calculateModelUpdate(double[] parameters, Collector<O> output) {
            double[] modelUpdate = new double[N_DIM];
            for (Tuple2<double[], Double> record : miniBatchSize) {
                double diff = (ArrayUtils.muladd(record.f0, parameters) - record.f1);
                ArrayUtils.addWith(modelUpdate, ArrayUtils.multiply(record.f0, diff));
            }

            output.collect(modelUpdate);
        }
    }
}

Also similar to the bounded case, for the asynchronous training the Parameters vertex would not wait for received updates from all the Train tasks. Instead, it would directly response to the task sending update:

Code Block

language	java

public static class ParametersCacheFunction extends ProcessFunction<Tuple2<Integer, double[]>, Tuple2<Integer, double[]>> {  
    
    private final double[] parameters = new double[N_DIM];

    public void processElement(Tuple2<Integer, double[]> update, Context ctx, Collector<Tuple2<Integer, double[]>> output) {
        ArrayUtils.addWith(parameters, update);
                
        if (update.f0 < 0) {
            // Received the initialized parameter values, broadcast to all the downstream tasks
            for (int i = 0; i < 10; ++i) {
                output.collect(new Tuple2<>(i, parameters))        
            }
        } else {
            output.collect(new Tuple2<>(update.f0, parameters))
        }
    }
}

Implementation Plan

Logically all the iteration types would support both BATCH and STREAM execution mode. However, according to the algorithms' requirements, we would implement

Unbounded iteration + STREAM mode.
Bounded iteration + BATCH mode.

Currently we do not see requirements on Bounded iteration + STREAM mode, if there are additional requirement in the future we would implement this mode, and it could also be supported with the current framework.

Compatibility, Deprecation, and Migration Plan

The API is added as a library inside flink-ml repository, thus it does not have compatibility problem. However, it has some difference with the existing iteration API and the algorithms would need some re-implementation.

For the long run, the new iteration implementation might provide an alternative for the iteration functionality, and we may consider deprecating and removing the existing API to reduce the complexity of core flink code.

Rejected Alternatives

Naiad has proposed a unified model for watermark mechanism (namely progress tracking outside of the iteration) and the progress tracking inside the iteration. It extends the event time and watermark to be a vector (long timestamp, int[] rounds) and implements a vectorized alignment algorithm. Although Naiad provides an elegant model, the direct implementation on Flink would requires a large amount of modification to the flink runtime, which would cause a lot of complexity and maintenance overhead. Thus we would choose to implement a simplified version on top of FLINK, as a part of the flink-ml library.

For the iteration DAG build graph, it would be more simpler if we could directly refer to the data stream variables outside of the closure of iteration body. However, since we need to make the iteration DAG creation first happen in the mock execution environment, we could not use these variables directly, otherwise we would directly modify the real environment and won't have chance to add wrappers to the operators.

Page tree

Versions Compared

Old Version 5

New Version Current

Key

Status

Motivation

Terminology

Public Interfaces

The Goals of the Iteration Library

Overall Design

Public Interfaces

Unbounded Iteration

Bounded Iteration

Key Implementation

Examples

Offline Training with Bounded Iteration

Online Training with Unbounded Iteration

Implementation Plan

Compatibility, Deprecation, and Migration Plan

Rejected Alternatives

Page tree

Page History

Versions Compared

Old Version 5

New Version Current

Key

Status

Motivation

Terminology

Public Interfaces

The Goals of the Iteration Library

Overall Design

Public Interfaces

Unbounded Iteration

Bounded Iteration

Key Implementation

Examples

Offline Training with Bounded Iteration

Online Training with Unbounded Iteration

Implementation Plan

Compatibility, Deprecation, and Migration Plan

Rejected Alternatives