From DataFrames to Tungsten: A Peek into Spark's Future @ Spark Summit San Francisco 2015
50 likes9,801 views
The document discusses Spark's DataFrame API and the Tungsten project. DataFrames make Spark accessible to different users by providing a common API across languages like Python, R and Scala. Tungsten aims to improve Spark's performance for the next five years through techniques like runtime code generation and off-heap memory management. Initial results show Tungsten doubling performance. Together, DataFrames and Tungsten will help Spark scale to larger data and queries across different languages and environments.
![Data frame: lingua franca for “small data”
head(flights)
#>
Source:
local
data
frame
[6
x
16]
#>
#>
year
month
day
dep_time
dep_delay
arr_time
arr_delay
carrier
tailnum
#>
1
2013
1
1
517
2
830
11
UA
N14228
#>
2
2013
1
1
533
4
850
20
UA
N24211
#>
3
2013
1
1
542
2
923
33
AA
N619AA
#>
4
2013
1
1
544
-‐1
1004
-‐18
B6
N804JB
#>
..
...
...
...
...
...
...
...
...
...](https://image.slidesharecdn.com/2015-06-16sparksummit-150617075504-lva1-app6892/85/From-DataFrames-to-Tungsten-A-Peek-into-Spark-s-Future-Spark-Summit-San-Francisco-2015-6-320.jpg)
From DataFrames to Tungsten: A Peek into Spark's Future @ Spark Summit San Francisco 2015
- 1. From DataFrames to Tungsten: A Peek into Spark’s Future Reynold Xin @rxin Spark Summit, San Francisco June 16th, 2015
- 2. DataFrame noun Making Spark accessible to everyone (data scientists, engineers, statisticians, …)
- 3. Tungsten noun Making Spark faster & prepare for the next five years.
- 4. How do DataFrames and Tungsten relate to each other?
- 5. Google Trends for “dataframe” Single-node tabular data structure, with API for relational algebra (filter, join, …) math and stats input/output (CSV, JSON, …) ad infinitum
- 6. Data frame: lingua franca for “small data” head(flights) #> Source: local data frame [6 x 16] #> #> year month day dep_time dep_delay arr_time arr_delay carrier tailnum #> 1 2013 1 1 517 2 830 11 UA N14228 #> 2 2013 1 1 533 4 850 20 UA N24211 #> 3 2013 1 1 542 2 923 33 AA N619AA #> 4 2013 1 1 544 -‐1 1004 -‐18 B6 N804JB #> .. ... ... ... ... ... ... ... ... ...
- 7. Spark DataFrame > head(filter(df, df$waiting < 50)) # an example in R ## eruptions waiting ##1 1.750 47 ##2 1.750 47 ##3 1.867 48 Distributed data frame for Java, Python, R, Scala Similar APIs as single-node tools (Pandas, dplyr), i.e. easy to learn
- 8. data size KB MB GB TB PB Existing Single-node Data Frames Spark DataFrame
- 9. It is not Spark vs Python/R, but Spark and Python/R.
- 11. Spark RDD Execution Java/Scala API JVM Execution Python API Python Execution opaque closures (user-defined functions)
- 12. Spark DataFrame Execution DataFrame Logical Plan Physical Execution Catalyst optimizer Intermediate representation for computation
- 13. Spark DataFrame Execution Python DF Logical Plan Physical Execution Catalyst optimizer Java/Scala DF R DF Intermediate representation for computation Simple wrappers to create logical plan
- 14. Benefit of Logical Plan: Simpler Frontend Python : ~2000 line of code (built over a weekend) R : ~1000 line of code i.e. much easier to add new language bindings (Julia, Clojure, …)
- 15. Performance 0 2 4 6 8 10 Java/Scala Python Runtime for an example aggregation workload RDD
- 16. Benefit of Logical Plan: Performance Parity Across Languages 0 2 4 6 8 10 Java/Scala Python Java/Scala Python R SQL Runtime for an example aggregation workload (secs) DataFrame RDD
- 20. Hardware Trends 2010 2015 Storage 50+MB/s (HDD) 500+MB/s (SSD) Network 1Gbps 10Gbps CPU ~3GHz ~3GHz
- 21. Hardware Trends 2010 2015 Storage 50+MB/s (HDD) 500+MB/s (SSD) 10X Network 1Gbps 10Gbps 10X CPU ~3GHz ~3GHz L
- 22. Tungsten: Preparing Spark for Next 5 Years Substantially speed up execution by optimizing CPU efficiency, via: (1) Runtime code generation (2) Exploiting cache locality (3) Off-heap memory management
- 23. From DataFrame to Tungsten Python DF Logical Plan Java/Scala DF R DF Tungsten Execution 5PM Deep Dive into Project Tungsten Developer Track by Josh Rosen
- 24. Initial Performance Results 0 200 400 600 800 1000 1200 1x 2x 4x 8x Runtime(seconds) Data set size (relative) Tungsten-off Tungsten-on
- 25. Python Java/Scala RSQL … DataFrame Logical Plan LLVMJVM GPU NVRAM Unified API, One Engine, Automatically Optimized Tungsten backend language frontend …
- 26. Tungsten Execution PythonSQL R Streaming DataFrame Advanced Analytics
- 27. Spark Office Hours Today Databricks booth A1 Topic Area 1:00-1:45 Core, YARN, Ops 1:45-2:30 Core/SQL/Data Science 3:00-3:40 Streaming 3:40-4:15 Core, Python, R 4:30-5:15 Machine Learning 5:15-6:00 Matei Zaharia