Introduction
Spark Cache and Persist are optimization techniques in DataFrame / Dataset for iterative and interactive Spark applications to improve the performance of Jobs.
Though Spark provides computation 100 x times faster than traditional Map Reduce jobs, If you have not designed the jobs to reuse the repeating computations you will see a degrade in performance when you are dealing with billions or trillions of data. Hence, we may need to look at the stages and use optimization techniques as one of the ways to improve performance.
Using persist() methods, Spark provides an optimization mechanism to store the intermediate computation of a Spark DataFrame so they can be reused in subsequent actions.
When you persist a dataset, each node stores its partitioned data in memory and reuses them in other actions on that dataset. And Spark’s persisted data on nodes are fault-tolerant meaning if any partition of a Dataset is lost, it will automatically be recomputed using the original transformations that created it.
Advantages of Caching and Persistence of DataFrame
Below are the advantages of using Spark Cache and Persist methods.
- Cost-efficient – Spark computations are very expensive hence reusing the computations is used to save cost.
- Time-efficient – Reusing repeated computations saves lots of time.
- Execution time – Saves execution time of the job and we can perform more jobs on the same cluster.
DataFrame Cache Syntax and Example
Spark DataFrame or Dataset cache() method by default saves it to storage level MEMORY_AND_DISK because recomputing the in-memory columnar representation of the underlying table is expensive. Note that this is different from the default cache level of RDD.cache() which is MEMORY_ONLY.
Syntax
cache() : Dataset.this.type
Spark cache() method in Dataset class internally calls persist() method which in turn uses sparkSession.sharedState.cacheManager.cacheQuery to cache the result set of DataFrame or Dataset. Let’s look at an example.
Example
val spark: SparkSession = SparkSession.builder()
.master("local[1]")
.appName("AIStudyLabs.com")
.getOrCreate()
// Read CSV with options
val df = spark.read.options(Map("inferSchema" -> "true", "delimiter" -> ",", "header" -> "true"))
.csv("src/main/resources/zipcodes.csv")
val df2 = df.where(col("State") === "PR").cache()
df2.show(false)
println(df2.count())
val df3 = df2.where(col("Zipcode") === 704)
println(df2.count())
Output
+-------+-----+-----+-----+
|Zipcode|City |State|Other|
+-------+-----+-----+-----+
| ...| ... | PR | ... |
+-------+-----+-----+-----+
DataFrame Persist Syntax and Example
Spark persist() method is used to store the DataFrame or Dataset to one of the storage levels MEMORY_ONLY, MEMORY_AND_DISK, MEMORY_ONLY_SER, MEMORY_AND_DISK_SER, DISK_ONLY, MEMORY_ONLY_2, MEMORY_AND_DISK_2 and more.
Caching or persisting of Spark DataFrame or Dataset is a lazy operation, meaning a DataFrame will not be cached until you trigger an action.
Syntax
- persist() : Dataset.this.type
- persist(newLevel : org.apache.spark.storage.StorageLevel) : Dataset.this.type
Spark persist has two signatures. The first signature doesn’t take any argument which by default saves it to MEMORY_AND_DISK storage level and the second signature which takes StorageLevel as an argument to store it to different storage levels.
Example
val dfPersist = df.persist()
dfPersist.show(false)
Using the second signature, you can save a DataFrame/Dataset to any storage level.
val dfPersist = df.persist(StorageLevel.MEMORY_ONLY)
dfPersist.show(false)
This stores DataFrame/Dataset into Memory.
Note that Dataset cache() is an alias for persist(StorageLevel.MEMORY_AND_DISK)
Unpersist Syntax and Example
Spark automatically monitors every persist() and cache() calls you make and it checks usage on each node and drops persisted data if not used or by using the least-recently-used (LRU) algorithm. You can also manually remove using the unpersist() method. unpersist() marks the Dataset as non-persistent, and removes all blocks for it from memory and disk.
Syntax
unpersist() : Dataset.this.type unpersist(blocking : scala.Boolean) : Dataset.this.type
Example
val dfPersist = dfPersist.unpersist()
unpersist(Boolean) with a boolean as an argument blocks until all blocks are deleted.
Spark Persist Storage Levels
All different storage levels PySpark supports are available at org.apache.spark.storage.StorageLevel class. The storage level specifies how and where to persist or cache a Spark DataFrame and Dataset.
MEMORY_ONLY – This is the default behavior of the RDD cache() method and stores the RDD or DataFrame as deserialized objects to JVM memory. When there is not enough memory available, it will not save the DataFrame of some partitions, and these will be recomputed as and when required. This takes more memory, but unlike RDD, this would be slower than MEMORY_AND_DISK level as it recomputes the unsaved partitions, and recomputing the in-memory columnar representation of the underlying table is expensive.
MEMORY_ONLY_SER – This is the same as MEMORY_ONLY but stores RDD as serialized objects to JVM memory. It takes lesser memory (space-efficient) than MEMORY_ONLY as it saves objects as serialized and takes an additional few more CPU cycles to deserialize.
MEMORY_ONLY_2 – Same as MEMORY_ONLY storage level, but replicates each partition to two cluster nodes.
MEMORY_ONLY_SER_2 – Same as MEMORY_ONLY_SER storage level, but replicates each partition to two cluster nodes.
MEMORY_AND_DISK – This is the default behavior of the DataFrame or Dataset. In this Storage Level, The DataFrame will be stored in JVM memory as a deserialized object. When required storage is greater than available memory, it stores some of the excess partitions into a disk and reads the data from the disk when required. It is slower as there is I/O involved.
MEMORY_AND_DISK_SER – This is the same as MEMORY_AND_DISK storage level, with the difference being it serializes the DataFrame objects in memory and on disk when space is not available.
MEMORY_AND_DISK_2 – Same as MEMORY_AND_DISK storage level, but replicates each partition to two cluster nodes.
MEMORY_AND_DISK_SER_2 – Same as MEMORY_AND_DISK_SER storage level, but replicates each partition to two cluster nodes.
DISK_ONLY – In this storage level, the DataFrame is stored only on disk, and the CPU computation time is high as I/O is involved.
DISK_ONLY_2 – Same as DISK_ONLY storage level, but replicates each partition to two cluster nodes.