You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Ufuk Celebi (JIRA)" <ji...@apache.org> on 2016/04/18 18:49:25 UTC
[jira] [Updated] (FLINK-3779) Add support for queryable state
[ https://issues.apache.org/jira/browse/FLINK-3779?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ufuk Celebi updated FLINK-3779:
-------------------------------
Description:
Flink offers state abstractions for user functions in order to guarantee fault-tolerant processing of streams. Users can work with both non-partitioned (Checkpointed interface) and partitioned state (getRuntimeContext().getState(ValueStateDescriptor) and other variants).
The partitioned state interface provides access to different types of state that are all scoped to the key of the current input element. This type of state can only be used on a KeyedStream, which is created via stream.keyBy().
Currently, all of this state is internal to Flink and used in order to provide processing guarantees in failure cases (e.g. exactly-once processing).
The goal of Queryable State is to expose this state outside of Flink by supporting queries against the partitioned key value state.
This will help to eliminate the need for distributed operations/transactions with external systems such as key-value stores which are often the bottleneck in practice. Exposing the local state to the outside moves a good part of the database work into the stream processor, allowing both high throughput queries and immediate access to the computed state.
This is the initial design doc for the feature: https://docs.google.com/document/d/1NkQuhIKYmcprIU5Vjp04db1HgmYSsZtCMxgDi_iTN-g. Feel free to comment.
was:
Flink offers state abstractions for user functions in order to guarantee fault-tolerant processing of streams. Users can work with both non-partitioned (Checkpointed interface) and partitioned state (getRuntimeContext().getState(ValueStateDescriptor) and other variants).
The partitioned state interface provides access to different types of state that are all scoped to the key of the current input element. This type of state can only be used on a KeyedStream, which is created via stream.keyBy().
Currently, all of this state is internal to Flink and used in order to provide processing guarantees in failure cases (e.g. exactly-once processing).
The goal of Queryable State is to expose this state outside of Flink by supporting queries against the partitioned key value state.
This will help to eliminate the need for distributed operations/transactions with external systems such as key-value stores which are often the bottleneck in practice. Exposing the local state to the outside moves a good part of the database work into the stream processor, allowing both high throughput queries and immediate access to the computed state.
> Add support for queryable state
> -------------------------------
>
> Key: FLINK-3779
> URL: https://issues.apache.org/jira/browse/FLINK-3779
> Project: Flink
> Issue Type: Improvement
> Components: Distributed Runtime
> Reporter: Ufuk Celebi
> Assignee: Ufuk Celebi
>
> Flink offers state abstractions for user functions in order to guarantee fault-tolerant processing of streams. Users can work with both non-partitioned (Checkpointed interface) and partitioned state (getRuntimeContext().getState(ValueStateDescriptor) and other variants).
> The partitioned state interface provides access to different types of state that are all scoped to the key of the current input element. This type of state can only be used on a KeyedStream, which is created via stream.keyBy().
> Currently, all of this state is internal to Flink and used in order to provide processing guarantees in failure cases (e.g. exactly-once processing).
> The goal of Queryable State is to expose this state outside of Flink by supporting queries against the partitioned key value state.
> This will help to eliminate the need for distributed operations/transactions with external systems such as key-value stores which are often the bottleneck in practice. Exposing the local state to the outside moves a good part of the database work into the stream processor, allowing both high throughput queries and immediate access to the computed state.
> This is the initial design doc for the feature: https://docs.google.com/document/d/1NkQuhIKYmcprIU5Vjp04db1HgmYSsZtCMxgDi_iTN-g. Feel free to comment.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)