I'm doing a POC for running Machine Learning algorithm on stream of data.
My initial idea was to take data, use
Spark Streaming --> Aggregate Data from several tables --> run MLLib on Stream of Data --> Produce Output.
But I cam across KStreams. Now I'm confused !!!
Questions :
1. What is difference between Spark Streaming and Kafka Streaming ?
2. How can I marry KStreams + Spark Streaming + Machine Learning ?
3. My idea is to train the test data continuously rather than have batch training..
First of all, the term "Confluent's Kafka Streaming" is technically not correct.
it's called Kafka's Streams API (aka Kafka Streams)
it's part of Apache Kafka and thus "owned" by the Apache Software Foundation (and not by Confluent)
there is Confluent Open Source and Confluent Enterprise -- two offers from Confluent that both leverage Apache Kafka (and thus, Kafka Streams)
However, Confluent contributes a lot of code to Apache Kafka, including Kafka Streams.
About the differences (I only highlight some main differences and refer to the Internet and documentation for further details: http://docs.confluent.io/current/streams/index.html and http://spark.apache.org/streaming/)
Spark Streaming:
micro-batching (no real record-by-record stream processing)
no sub-second latency
limited window operations
no event-time processing
processing framework (difficult to operate and to deploy)
part of Apache Spark -- a data processing framework
exactly-once processing
Kafka Streams
record-by-record stream processing
ms latency
rich window operations
stream/table duality
event time, ingestion time, and processing time semantics
Java library (easy to run and deploy -- it's just a Java application as any other)
part of Apache Kafka -- a Stream Processing Platform (ie, it offers storage and processing at once)
at-least-once processing (exactly-once processing is WIP; cf KIP-98 and KIP-129)
elastic, ie, dynamically scalable
Thus there is no reasons to "marry" both -- it's a question of choice which one you want to use.
My personal take is, that Spark is not a good solution for stream processing. If you want to use a library like Kafka Streams or a framework like Apache Flink, Apache Storm, or Apache Apex (which are all good option for stream processing) depends on your use case (and maybe personal taste) and cannot be answered on SO.
A main differentiator of Kafka Streams is, that it is a library and does not require a processing cluster. And because it is part of Apache Kafka and if you have Apache Kafka already in place, this might simplify your overall deployment as you do not need to run an extra processing cluster.
I have recently presented at a conference about this topic.
Apache Kafka Streams or Spark Streaming are typically used to apply a machine learning model in real time to new events via stream processing (process data while it is in motion). Matthias answer already discusses their differences.
On the other side, you first use things like Apache Spark MLlib (or H2O.ai or XYZ) to build the analytic models first using historical data sets.
Kafka Streams can be used for online training of models, too. Though, I think online training has various caveats.
All of this is discussed in more details in my slide deck "Apache Kafka Streams and Machine Learning / Deep Learning for Real Time Stream Processing".
Apache Kafka Steams is library and provides embeddable stream processing engine and it is easy to use in Java applications for stream processing and it is not a framework.
I found some Use cases about when to use Kafka Streams and also good comparison with Apache flink from Kafka author.
Spark Streaming and KStreams in one pic from stream processing point of view.
Highlighted the significant advantages of Spark Streaming and KStreams here to make answer short.
Spark Streaming Advantages over KStreams:
Easy to integrate Spark ML models and Graph computing in same application without writing data outside of an application which means you will process the much quicker than writing kafka again and process.
Join non streaming sources like files system and other non kafka sources with other stream sources in same application.
Messages with Schema can be easily processed with most favorite SQL (StructuredStreaming).
Possible to do graph analysis over streaming data with GraphX inbuilt library.
Spark apps can be deployed over (if) existing YARN or Mesos cluster.
KStreams Advantages:
Compact library for ETL processing and ML model serving/training on messages with rich features. So far, both source and target should be Kafka topic only.
Easy to achieve exactly once semantics.
No separate processing cluster required.
Easy to deploy on docker since it's a plain java application to run.
Related
With Spark streaming, I can read Kafka messages and write data to different kind of tables, for example HBase, Hive and Kudu. But this can also be done by using Kafka connectors for these tables. My question is, in which situations I should prefer connectors over the Spark streaming solution.
Also how tolerant is the Kafka connector solution? We know that with Spark streaming, we can use checkpoints and executors running on multiple nodes for fault tolerant execution, but how is fault tolerance (if possibe) achieved with Kafka connectors? By running the connector on multiple nodes?
So, generally, there should be no big difference in functionality when it comes to simply reading records from Kafka and sending them into other services.
Kafka Connect is probably easier when it comes to standard tasks since it offers various connectors out-of-the-box, so it will quite probably reduce the need of writing any code. So, if you just want to copy a bunch of records from Kafka to HDFS or Hive then it will probably be easier and faster to do with Kafka connect.
Having this in mind, Spark Streaming drastically takes over when You need to do things that are not standard i.e. if You want to perform some aggregations or calculations over records and write them to Hive, then You probably should go for Spark Streaming from the beginning.
Genrally, I found doing some substandard things with Kafka connect, like for example splitting one message to multiple ones(assuming it was for example JSON array) to be quite troublesome and often require much more work than it would be in Spark.
As for the Kafka Connect fault tolerance, as it's described in the docs this is achieved by running multiple distributed workers with same group.id, the workers redistribute tasks and connectors if one of them fails.
in which situations I should prefer connectors over the Spark streaming solution.
"It Depends" :-)
Kafka Connect is part of Apache Kafka, and so has tighter integration with Apache Kafka in terms of security, delivery semantics, etc.
If you don't want to write any code, Kafka Connect is easier because it's just JSON to configure and run
If you're not using Spark already, Kafka Connect is arguably more
straightforward to deploy (run the JVM, pass in the configuration)
As a framework, Kafka Connect is more transferable since the concepts are the same, you just plugin the appropriate connector for the technology that you want to integrate with each time
Kafka Connect handles all the tricky stuff for you like schemas, offsets, restarts, scaleout, etc etc etc
Kafka Connect supports Single Message Transform for making changes to data as it passes through the pipeline (masking fields, dropping fields, changing data types, etc etc). For more advanced processing you would use something like Kafka Streams or ksqlDB.
If you are using Spark, and it's working just fine, then it's not necessarily prudent to rip it up to use Kafka Connect instead :)
Also how tolerant is the Kafka connector solution? … how is fault tolerance (if possibe) achieved with Kafka connectors?
Kafka Connect can be run in distributed mode, in which you have one or more worker processes across nodes. If a worker fails, Kafka Connect rebalances the tasks across the remaining ones. If you add a worker in, Kafka Connect will rebalance to ensure workload distribution. This was drastically improved in Apache Kafka 2.3 (KIP-415)
Kafka Connect uses the Kafka consumer API and tracks offsets of records delivered to a target system in Kafka itself. If the task or worker fails you can be sure that it will restart from the correct point. Many connectors support exactly-once delivery too (e.g. HDFS, Elasticsearch, etc)
If you want to learn more about Kafka Connect see the docs here and my talk here. See a list of connectors here, and tutorial videos here.
Disclaimer: I work for Confluent and a big fan of Kafka Connect :-)
Over the past few years we have developed quite some Spark Streaming (Direct API) applications that are reading or writing to/from Kafka, IBM MQ, Hive, HBase, HDFS, and others on our Cloudera Platform. Now that the Direct API of Spark Streaming (we currently have version 2.3.2) is deprecated and we recently added the Confluent platform (comes with Kafka 2.2.0) to our project we plan to migrate these applications.
What is the natural replacement of our Spark Streaming applications? Should we migrate to Spark Structured Streaming or rather to Kafka Streams?
I personally do not have any experience with both frameworks but in my view Spark Structured Streaming seems to be the natural choice. Our code base is mainly written in Scala which could be also used for the Structured API. Kafka Streams has a few limitations with Scala. Although we might loose some flexibility by leaving the low level API of RDDs and moving to a higher level of DataFrames we could build on our knowledge with Spark.
On the other side there is Kafka Streams which is probably the best choice when it comes to processing data between Kafka topics which is our main use case. And looking at all the Kafka Connectors that come with Confluent the other uses cases can be served as well.
You currently have some Spark scheduler, therefore you can use Structured Streaming, which is binary compatible with the old Streaming API.
If you're using Mesos or k8s, then putting Kafka Streams apps in Docker and running those is easier to scale, monitor and configure than Spark, IMO since it acts as any other Docker container in those systems, so you build a pattern around everything
Kafka Streams... is probably the best choice when it comes to processing data between Kafka topics
True.
Kafka Streams has a few limitations with Scala.
I think you might want to keep reading that section
The Kafka Streams DSL for Scala library is a wrapper over the existing Java APIs for Kafka Streams DSL that addresses the concerns raised
Of course you could always use Kotlin to interop better with the Java API
I have come across three popular streaming techniques that are Spark Streaming, Structured Streaming and Kafka Streaming.
I have gone through various sites but not getting this answer, are these three the same thing or different?
If not same what is the basic difference.
I am not looking for an in depth answer. But an answer to above question (yes or no) and a little intro to each of them so that I can explore more. :)
Thanks in advance
Subrat
I guess you are referring to Kafka Streams when you say "Kafka Streaming".
Kafka Streams is a JVM library, part of Apache Kafka. It is a way of processing data in Kafka topics providing an abstraction layer. Applications running KafkaStreams library can be run anywhere (not just in the Kafka cluster, actually, it is not recommended to). They'll consume, process and produce data to/from the Kafka cluster.
Spark Streaming is a part of Apache Spark distributed data processing library, that provides Stream (as oppposed to batch) processing. Spark initially provided batch computation only, so a specific layer Spark Streaming was provided for stream processing. Spark Streaming can be fed with Kafka data, but it can be connected to other sources as well.
Structured Streaming, within the realm of Apache Spark, is a different approach that came to overcome certain limitations to stream processing of the previous approach that Spark Streaming was using. It was added to Spark from a certain version onwards(2.0 IIRC).
For my project, I need to harvest data from Twitter.
I am currently facing two design choices:
What is the best software architecture? I read that spark has Twitter support but I am not familiar with Scala. On the other hand, Apache Spark seems a good option, but then I'm not sure on how to save data to a common sink
I have some budget constraints. I surely need one server to do the sink and the processing. However, for the data harvesting, I don't know if several VM/containers offer a better performance / cost ratio than a bunch of Raspberry PI running Kafka producers.
Take a look at Confluent platform and especially Kafka Connect [1].
There is a Twitter connector out of the box. All the twitter data will be streamed to Kafka.
[1] https://www.confluent.io/blog/using-ksql-to-analyse-query-and-transform-data-in-kafka
Agree with #leshkin that Kafka Connect is the most natural fit. However, the Twitter connector (available on github here) does not require Confluent Platform, simply Kafka Connect which is a standard part of the Apache Kafka distribution. https://kafka.apache.org/documentation/#connect
If you choose, you can run Kafka connect workers in distributed mode to divide the load across several VMs/containers/boxes and these don't have to be the same boxes you run your kafka brokers (they only need some relevant libs from kafka and the libs for the connector and Java of course)
I have apache access log file and i want to store access counts (total/daily/hourly) of each page in a cassandra table.
I am trying to do it by using kafka connect to stream from log file to a kafka topic. In order to increment metrics counters in Cassandra can I use Kafka Connect again? Otherwise which other tool should be used here e.g. kafka streams, spark, flink, kafka connect etc?
You're talking about doing stream processing, which Kafka can do - either with Kafka's Streams API, or KSQL. KSQL runs on top of Kafka Streams, and gives you a very simple way to build the kind of aggregations that you're talking about.
Here's an example of doing aggregations of streams of data in KSQL
SELECT PAGE_ID,COUNT(*) FROM PAGE_CLICKS WINDOW TUMBLING (SIZE 1 HOUR) GROUP BY PAGE_ID
See more at : https://www.confluent.io/blog/using-ksql-to-analyse-query-and-transform-data-in-kafka
You can take the output of KSQL which is actually just a Kafka topic, and stream that through Kafka Connect e.g. to Elasticsearch, Cassandra, and so on.
You mention other stream processing tools, they're valid too - depends in part on existing skills and language preferences (e.g. Kafka Streams is Java library, KSQL is … KSQL, Spark Streaming has Python as well as Java, etc), but also deployment preferences. Kafka Streams is just a Java library to deploy within your existing application. KSQL is deployable in a cluster, and so on.
This can be easily done with Flink, either as a batch or streaming job, and either with or without Kafka (Flink can read from files and write to Cassandra). This sort of time windowed aggregation is easily done with Flink's SQL api; see the examples here.