How hazelcast-jet achieves anything vastly different from what was earlier achievable by submitting EntryProcessors on keys in an IMap?
Curious to know.
Quoting the InfoQ article on Jet:
Sending a runnable to a partition is analogous to the work of a single DAG vertex. The advantage of Jet comes from the ability to have the vertex transform the data it reads, producing items which no longer belong to the same partition, then reshuffle them while sending to the downstream vertex so they are again correctly partitioned. This is essential for any kind of map-reduce operation where the reducing unit must observe all the data items with the same key. To minimize network traffic, Jet can first reduce the data slice produced on the local member, then send only one item per key to the remote member that combines the partial results.
And note that this is just an advantage in the context of the same or similar use cases currently covered by entry processors. Jet can take data from any source and make use of the whole cluster's computational resources to process it.
Related
I've run into a technical challenge around Neo4j usage that has had me stumped for a while. My organization uses Neo4j to model customer interaction patterns. The graph has grown to a size of around 2 million nodes and 7 million edges. All nodes and edges have between 5 and 10 metadata properties. Every day, we export data on all of our customers from Neo4j to a series of python processes that perform business logic.
Our original method of data export was to use paginated cypher queries to pull the data we needed. For each customer node, the cypher queries had to collect many types of surrounding nodes and edges so that the business logic could be performed with the necessary context. Unfortunately, as the size and density of the data grew, these paginated queries began to take too long to be practical.
Our current approach uses a custom Neo4j procedure to iterate over nodes, collect the necessary surrounding nodes and edges, serialize the data, and place it on a Kafka queue for downstream consumption. This method worked for some time but is now taking long enough so that it is also becoming impractical, especially considering that we expect the graph to grow an order of magnitude in size.
I have tried the cypher-for-apache-spark and neo4j-spark-connector projects, neither of which have been able to provide the query and data transfer speeds that we need.
We currently run on a single Neo4j instance with 32GB memory and 8 cores. Would a cluster help mitigate this issue?
Does anyone have any ideas or tips for how to perform this kind of data export? Any insight into the problem would be greatly appreciated!
As far as I remember Neo4j doesn't support horizontal scaling and all data is stored in a single node. To use Spark you could try to store your graph in 2+ nodes and load the parts of the dataset from these separate nodes to "simulate" the parallelization. I don't know if it's supported in both of connectors you quote.
But as told in the comments of your question, maybe you could try an alternative approach. An idea:
Find a data structure representing everything you need to train your model.
Store such "flattened" graph in some key-value store (Redis, Cassandra, DynamoDB...)
Now if something changes in the graph, push the message to your Kafka topic
Add consumers updating the data in the graph and in your key-value store directly after (= make just an update of the graph branch impacted by the change, no need to export the whole graph or change the key-value store at the same moment but it would very probably lead to duplicate the logic)
Make your model querying directly the key-value store.
It depends also on how often your data changes, how deep and breadth is your graph ?
Neo4j Enterprise supports clustering, you could use the Causal Cluster feature and launch as many read replicas as needed, run the queries in parallel on the read replicas, see this link: https://neo4j.com/docs/operations-manual/current/clustering/setup-new-cluster/#causal-clustering-add-read-replica
I use hazelcast to store data in memory, and sometimes one point will lose connection to others, and the value of hazelcast's map may be different.
I want to know how it comes when the connection is rebuilt when the values in map are different.
When network partitioning happens, your cluster might split into two and as you say, the values might differ as they run in parallel. After the clusters join again, split-brain recovery mechanism executes. Depending on the data structure you use, the data in both clusters merge according to the configured merge policy. I suggest reading Split Brain Recovery section of the Hazelcast manual for more detail.
I am learning spark streaming using the book "Learning spark Streaming". In the book i found the following on a section talking about Dstream, RDD, block/partition.
Finally, one important point that is glossed over in this schema is that the Receiver interface also has the option of connecting to a data source that delivers a collection (think Array) of data pieces. This is particularly relevant in some de-serialization uses, for example. In this case, the Receiver does not go through a block interval wait to deal with the segmentation of data into partitions, but instead considers the whole collection reflects the segmentation of the data into blocks, and creates one block for each element of the collection. This operation is demanding on the part of the Producer of data, since it requires it to be producing blocks at the ratio of the block interval to batch interval to function reliably (delivering the correct number of blocks on every batch). But some have found it can provide superior performance, provided an implementation that is able to quickly make many blocks available for serialization.
I have been banging my head around and can't simply understand what the Author is talking about, although i feel like i should understand it. Can someone give me some pointers on that ?
Disclosure: I'm co-author of the book.
What we want to express there is that the custom receiver API has 2 working modes: one where the producing side delivers one-message-at-time and the other where the receiver may deliver many messages at once (bulk).
In the one-message-at-time mode, Spark is responsible of buffering and collecting the data into blocks for further processing.
In the bulk mode, the burden of buffering and grouping is on the producing side, but it might be more efficient in some scenarios.
This is reflected in the API:
def store(dataBuffer: ArrayBuffer[T]): Unit
Store an ArrayBuffer of received data as a data block into Spark's memory.
def store(dataItem: T): Unit
Store a single item of received data to Spark's memory.
I agree with you that the paragraph is convoluted and might not convey the message as clear as we would like. I'll take care of improving it.
Thanks for your feedback!
I am trying to get some change event from Kafka that I would like to propagate downstream in another system. However the Change order matters. Hence I wonder what is the appropriate way to do that with some Spark transformation in the middle.
The only thing I see is to loose the parallelism and make the DStream on one partition. Maybe there is a way to do operation in parallel and bring everything back in one partition and then send it to the external system or back in Kafka and then use a Kafka Sink for the matter.
What approach can I try?
In a distributed environment, with some form of cashing/buffering at most layer, message generated from same machine may reach back-end in different order. Also the definition of order is subjective. Implementing a global definition of order will be restrictive (may not be correct) for the data as a whole.
So, Kafka is meant for keeping the data in order in the order of put but partition comes as a catch!!! Partition defines the level of parallelism per topic.
Typically, the level of abstraction at which kafka is kept, it should not bother much about order. It should be optimised for maximum throughput, where partitioning will come handy!!! Consider ordering just a side effect of supporting streaming!!!
Now, what ever logic ensures, that data is put in to kafka in order, that makes more sense in your application (spark job).
I'm coming from a Hadoop background and have limited knowledge about Spark. BAsed on what I learn so far, Spark doesn't have mapper/reducer nodes and instead it has driver/worker nodes. The worker are similar to the mapper and driver is (somehow) similar to reducer. As there is only one driver program, there will be one reducer. If so, how simple programs like word count for very big data sets can get done in spark? Because driver can simply run out of memory.
The driver is more of a controller of the work, only pulling data back if the operator calls for it. If the operator you're working on returns an RDD/DataFrame/Unit, then the data remains distributed. If it returns a native type then it will indeed pull all of the data back.
Otherwise, the concept of map and reduce are a bit obsolete here (from a type of work persopective). The only thing that really matters is whether the operation requires a data shuffle or not. You can see the points of shuffle by the stage splits either in the UI or via a toDebugString (where each indentation level is a shuffle).
All that being said, for a vague understanding, you can equate anything that requires a shuffle to a reducer. Otherwise it's a mapper.
Last, to equate to your word count example:
sc.textFile(path)
.flatMap(_.split(" "))
.map((_, 1))
.reduceByKey(_+_)
In the above, this will be done in one stage as the data loading (textFile), splitting(flatMap), and mapping can all be done independent of the rest of the data. No shuffle is needed until the reduceByKey is called as it will need to combine all of the data to perform the operation...HOWEVER, this operation has to be associative for a reason. Each node will perform the operation defined in reduceByKey locally, only merging the final data set after. This reduces both memory and network overhead.
NOTE that reduceByKey returns an RDD and is thus a transformation, so the data is shuffled via a HashPartitioner. All of the data does NOT pull back to the driver, it merely moves to nodes that have the same keys so that it can have its final value merged.
Now, if you use an action such as reduce or worse yet, collect, then you will NOT get an RDD back which means the data pulls back to the driver and you will need room for it.
Here is my fuller explanation of reduceByKey if you want more. Or how this breaks down in something like combineByKey