1.1 Why Spark is Faster Than Hadoop | hadoop Vs spark |Hadoop Interview questions

great video but very poor audio quality.

Why Spark better than MR? 1) Iterative Algorithms: if a single dataset is being used multiple times, Spark does caching and keeps data in RAM without killing executors, while MR always writes to HDFS + kills JVM. 2) Compact syntax and rich library.

How to hear you..

audio is very low

Very bad Audio quality .

Also catalyst optimizer is available in spark and not in hadoop

Sir is knowing spark without knowing Hadoop a wise thing.ive started learning spark but donot know about hadoop is it a great idea

Pathetic audio .. please dont post such videos ..

Audio is very low.. need to use earfone and full volume but that is painful when suddenly an ad plays. Torture for ears.

Hi As you mentioned that executors are not killed on completion of task. i am having a job which is launching many executors as I have given configuration as spark.core.max=20 and spark.executor.cores = 2 so at any instant of time only 10 executors kept on running state but the number of executors that launched throughout the job was more than 10. I am using
spark standalone.

Can you please make a video on hive optimization techniques

How iterative broad cast helps to overcome skew problems. Pls do a video

The explanation is bang on, and even in Spark Summit of a few years back, they explained similar reasons. That hand-offs between multiple MR jobs take a toll due to the involvement of the disk, whereas same is done via memory in Spark. Also, another reason could be that in MR we need to break every problem as {Key, Value} - which is not feasible every time, whereas the same is not required for Spark.

audio is very poor quality

One principle advantage of choosing Spark vs hadoop MR is that Spark is having 100's of inbuilt operators(Transformations + Actions) such as map, filter, reduceByKey, count(), collect() and many more which makes developers to write less code.

Not audible.