[ 
https://issues.apache.org/jira/browse/FLINK-5544?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15930166#comment-15930166
 ] 

ASF GitHub Bot commented on FLINK-5544:
---------------------------------------

Github user StefanRRichter commented on a diff in the pull request:

    https://github.com/apache/flink/pull/3359#discussion_r106669974
  
    --- Diff: 
flink-streaming-java/src/main/java/org/apache/flink/streaming/api/operators/InternalTimerService.java
 ---
    @@ -18,43 +18,306 @@
     package org.apache.flink.streaming.api.operators;
     
     import org.apache.flink.annotation.Internal;
    +import org.apache.flink.annotation.VisibleForTesting;
    +import org.apache.flink.api.common.typeutils.TypeSerializer;
    +import org.apache.flink.core.memory.DataInputViewStreamWrapper;
    +import org.apache.flink.core.memory.DataOutputViewStreamWrapper;
    +import org.apache.flink.runtime.state.KeyGroupRange;
    +import org.apache.flink.streaming.runtime.tasks.EventTimeCallback;
    +import org.apache.flink.streaming.runtime.tasks.ProcessingTimeCallback;
    +import org.apache.flink.streaming.runtime.tasks.ProcessingTimeService;
    +import org.apache.flink.util.InstantiationUtil;
    +import org.apache.flink.util.Preconditions;
    +
    +import java.io.IOException;
    +import java.util.ArrayList;
    +import java.util.Collection;
    +import java.util.List;
    +import java.util.Set;
    +import java.util.concurrent.ScheduledFuture;
    +
    +import static org.apache.flink.util.Preconditions.checkArgument;
    +import static org.apache.flink.util.Preconditions.checkNotNull;
     
     /**
      * Interface for working with time and timers.
      *
      * <p>This is the internal version of {@link 
org.apache.flink.streaming.api.TimerService}
      * that allows to specify a key and a namespace to which timers should be 
scoped.
      *
    + * All d
    + * 
    + * @param <K> Type of the keys in the stream
      * @param <N> Type of the namespace to which timers are scoped.
      */
     @Internal
    -public interface InternalTimerService<N> {
    +public abstract class InternalTimerService<K, N> implements 
ProcessingTimeCallback, EventTimeCallback {
    --- End diff --
    
    I would suggest to still keep the old interface and rename this to 
`AbstractInternalTimerService<K, N> implements InternalTimerService<N>`. Like 
that, we don't need to introduce the generic parameter K in all places, which 
is actually giving away some implementation detail (K is used only for a 
member, not for the interface methods). I also like to keep the interface slim, 
and probably not every code that deals with `InternalTimerService` has to see 
all the methods, e.g. for snapshots.


> Implement Internal Timer Service in RocksDB
> -------------------------------------------
>
>                 Key: FLINK-5544
>                 URL: https://issues.apache.org/jira/browse/FLINK-5544
>             Project: Flink
>          Issue Type: New Feature
>          Components: State Backends, Checkpointing
>            Reporter: Xiaogang Shi
>            Assignee: Xiaogang Shi
>
> Now the only implementation of internal timer service is 
> HeapInternalTimerService which stores all timers in memory. In the cases 
> where the number of keys is very large, the timer service will cost too much 
> memory. A implementation which stores timers in RocksDB seems good to deal 
> with these cases.
> It might be a little challenging to implement a RocksDB timer service because 
> the timers are accessed in different ways. When timers are triggered, we need 
> to access timers in the order of timestamp. But when performing checkpoints, 
> we must have a method to obtain all timers of a given key group.
> A good implementation, as suggested by [~StephanEwen], follows the idea of 
> merge sorting. We can store timers in RocksDB with the format 
> {{KEY_GROUP#TIMER#KEY}}. In this way, the timers under a key group are put 
> together and are sorted. 
> Then we can deploy an in-memory heap which keeps the first timer of each key 
> group to get the next timer to trigger. When a key group's first timer is 
> updated, we can efficiently update the heap.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Reply via email to