We are creating a real-time stream processing system with spark streaming which uses large number (millions) of analytic models applied to RDDs in the many different type of incoming metric data streams(more then 100000). This streams are original or transformed streams. Each RDD has to go through an analytical model for processing. Since we do not know which spark cluster node will process which specific RDDs from different streams, we need to make ALL these models available at each Spark compute node. This will create huge overhead at each spark node. We are considering using in-memory data grids to provide these models at spark compute nodes. Is this the right approach?
Or
Should we avoid using Spark streaming all together and just use in-memory data grids like Redis(with pub/sub) to solve this problem. In that case we will stream data to specific Redis nodes which contain the specific models. of course we will have to do all binning/window etc..
Please suggest.
Sounds like to me like you need a combination of stream processing engine and a distributed data store. I would design the system like this.
The distributed datastore (Redis, Cassandra, etc.) can have the data you want to access from all the nodes.
Receive the data streams through a combination data ingestion system (Kafka, Flume, ZeroMQ, etc.) and process it in the stream processing system (Spark Streaming [preferably ;)], Storm, etc.).
In the functions that is used to process the stream records, the necessary data will have to pulled from the data store and maybe cached locally as appropriate.
You may also have to update the data store from spark streaming as application needs it. In which case you will also have to worry about versioning of the data that you want pull in step 3.
Hopefully that made sense. Its hard to give any more specifics of the implementation without the exactly computation model. Hope this helps!
Related
I suggest it's a good idea to process huge JDBC table by reading rows by batches and processing them with Spark Streaming. This approach doesn't require reading all rows into memory. I suppose no monitoring of new rows in the table, but just reading the table once.
I was surprised that there is no JDBC Spark Streaming receiver implementation. Implementing Receiver doesn't look difficult.
Could you describe why such receiver doesn't exist (is this approach a bad idea?) or provide links to implementations.
I've found Stratio/datasource-receiver. But it reads all data in a DataFrame before processing by Spark Streaming.
Thanks!
First of all actual streaming source would require a reliable mechanism for monitoring updates, which is simply not a part of JDBC interface nor it is a standardized (if at all) feature of major RDBMs, not to mention other platforms, which can be accessed through JDBC. It means that streaming from a source like this typically requires using log replication or similar facilities and is highly resource dependent.
At the same what you describe
suggest it's a good idea to process huge JDBC table by reading rows by batches and processing them with Spark Streaming. This approach doesn't require reading all rows into memory. I suppose no monitoring of new rows in the table, but just reading the table once
is really not an use case for streaming. Streaming deals with infinite streams of data, while you ask is simply as scenario for partitioning and such capabilities are already a part of the standard JDBC connector (either by range or by predicate).
Additionally receiver based solutions simply don't scale well and effectively model a sequential process. As a result their applications are fairly limited, and wouldn't be even less appealing if data was bounded (if you're going to read finite data sequentially on a single node, there is no value in adding Spark to the equation).
I don't think it is a bad idea since in some cases you have constraints that are outside your power,e.g. legacy systems to which you cannot apply strategies such as CDC but to which you still have to consume as a source of stream data.
On the other hand, Spark Structure Streaming engine, in micro-batch mode, requires the definition of an offset than can be advanced, as you can see in this class. So, if your table has some column that can be used as an offset, you can definitely stream from it, although RDMDS are not the "streaming-friendly" as far as I know.
I have developed Jdbc2s which is a DataSource V1 streaming source for Spark. It's also deployed to Maven Central, if you need. Coordinates are in the documentation.
We have an exiting batch processing which is working as mentioned below
Hive SQL is using for Daily batch processing.
Data are being either ingested from Files or RDMBS
Data is ingested in Raw --> Staging --> Mart, with staging to mart being all the business transformation and raw to staging is just cleansing and formatting of data.
Now as Part of getting real or near real time data, I am evaluating the Lambda Architecture and this is what plan is?
ALL the source system is going to land on Kafka.
Same batch processing System will consume Kafka topics.
New Spark Application will consume kafka topics for streaming.
Serving layer will create views which will combine both the aggregate data from Streaming and Batch for real (near real) time processing.
The problem is, the Logic will be duplicated in HiveQL (Batch) and Spark (Streaming). is there a way I can avoid this or minimize this?
You can build your processing stages using Spark SQL and Spark Structured Streaming: https://spark.apache.org/docs/2.2.0/structured-streaming-programming-guide.html. Depending on your needs there can be some incompatibilities. But I´d try to build the Spark Aggregations + Transformations using the Dataset[_] api and then try to spawn in both ways, batch and streaming.
The problem of duplicated code base is inherent in lambda architecture. It gets a mention in the 'criticism' section of the wikipedia page
Another issue is that the data between batch and stream are not in sync so can lead to unexpected results when bringing data together. For example, joining across stream and batch when keys do not yet exist in batch.
I believe the lambda architecture comes from an belief that streaming is complex and expensive so keep batch as much as possible and add streaming only for those elements that require near-real time. We already have batch, let's add a few streaming things.
An alternate architecture is to use streaming for everything. This is based on the realization that batch is a special case of streaming, so do your batch and stream processing on a single streaming platform.
use spark structured streaming for batch
lambda architecture issues and how only using streaming solves them
questioning the lambda architecture
Below is the high level usecase which im trying to workon.
we have stream of students data published into a Kafka topic and our module has to read the student ids as stream and fetch associated data from multiple sources for each student and perform some calculation for each student and publish the associated calculation for each student into a kafka topic.
So here the question is it better to write a single big Spark job or use Akka to have separate service for each source so that actors can work parallely take bunch of student ids and get the data from respective source and perform some bunch Transformations and actions and finally a calculation associated with each student .
Or do i really need to use Akka here? Will Spark handles this efficiently internally?
Appreciate any thoughts here.
If your transformations take data from Kafka as input and produce output back into Kafka, it appears the most natural fit is Kafka Streams. I'd look to that first. Kafka Streams take advantage of the partitioning of data on Kafka to process partition groups in parallel to each other, but process messages sequentially within in each group, similarly how akka actors work in parallel to each other but each actor internally processes messages sequentially.
However, if your calculation requires e.g. machine learning or in general some iterative data-processing which does re-partitioning (shuffling in spark lingo) of the data between iterations, then Kafka Streams would no longer be that good a fit, I think. Then I'd consider Spark or Flink.
Akka is really powerful and you can use it in both these cases and more. However, it's a lower level library than Kafka Streams, Spark or Flink. Which means you have more power but also more considerations to think about. If using akka, I'd go for akka-streams. They have a good integration with kafka via the akka-stream-kafka (aka reactive-kafka) library.
What is a better choice for a long-term store (many writes, few reads) of data processed through Spark Streaming: Parquet, HBase or Cassandra? Or something else? What are the trade-offs?
In my experience we have used Hbase as datastore for spark streaming data(we also has same scenario many writes and few reads), since we are using hadoop, hbase has native integration with hadoop and it went well..
Above we have used tostore hight rate of messages coming over from solace.
HBase is well suited for doing Range based scans. Casandra is known for availablity and many other things...
However, I can also observe one general trend in many projects, they are simply storing rawdata in hdfs (parquet + avro) in partitioned structure through spark streaming with spark dataframe(SaveMode.Append) and they are processing rawdata with Spark
Ex of partitioned structure in hdfs :
completion ofbusinessdate/environment/businesssubtype/message type etc....
in this case there is no need for going to Hbase or any other data store.
But one common issue in above approach is when you are getting small and tiny files, through streaming then you would need to repartion(1) or colelese or FileUtils.copymerge to meet block size requirements to single partitioned file. Apart from that above approach also would be fine.
Here is some thing called CAP theorm based on which decision can be taken.
Consistency (all nodes see the same data at the same time).
Availability (every request receives a response about whether it
succeeded or failed).
Partition tolerance (the system continues to
operate despite arbitrary partitioning due to network failures)
Casandra supports AP.
Hbase supports CP.
Look at detailed analysis given here
I'm trying to benchmark a few approaches to putting an image processing algorithm into apache spark. For one step in this algorithm, a computation on a pixel in the image will depend on an unknown amount of surrounding data, so we can't partition the image with guaranteed sufficient overlap a priori.
One solution to that problem I need to benchmark is for a worker node to ask the master node for more data when it encounters a pixel with insufficient surrounding data. I'm not convinced this is the way to do things, but I need to benchmark it anyway because of reasons.
Unfortunately, after a bunch of googling and reading docs I can't find any way for a processingFunc called as part of sc.parallelize(partitions).map(processingFunc) to query the master node for more data from a different partition mid-computation.
Does a way for a worker node to ask the master for more data exist in spark, or will I need to hack something together that kind of goes around spark?
Master Node in Spark is for allocating the resources to a particular job and once the resources are allocated, the Driver ships the complete code with all its dependencies to the various executors.
The first step in every code is to load the data to the Spark cluster. You can read the data from any underlying data repository like Database, filesystem, webservices etc.
Once data is loaded it is wrapped into an RDD which is partitioned across the nodes in the cluster and further stored in the workers/ Executors Memory. Though you can control the number of partitions by leveraging various RDD API's but you should do it only when you have valid reasons to do so.
Now all operations are performed over RDD's using its various methods/ Operations exposed by RDD API. RDD keep tracks of partitions and partitioned data and depending upon the need or request it automatically query the appropriate partition.
In nutshell, you do not have to worry about the way data is partitioned by RDD or which partition stores which data and how they communicate with each other but if you do care, then you can write your own custom partitioner, instructing Spark of how to partition your data.
Secondly if your data cannot be partitioned then I do not think Spark would be an ideal choice because that will result in processing of everything in 1 single machine which itself is contrary to the idea of distributed computing.
Not sure what is exactly your use case but there are people who have been leveraging Spark for Image processing. see here for the comments from Databricks