Tachyon Presentation at AMPCamp 6 (November, 2015)

Haoyuan Li, Tachyon Nexus & UC Berkeley

November 19, 2015 @ AMPCamp 6

An Open Source Memory-Centric Distributed Storage System

Outline

•  Open Source

•  Introduction to Tachyon (Before 2015)

•  Deployments and New Features

•  Getting Involved

2

Background •  Started at UC Berkeley AMPLab –  From summer 2012

•  Open sourced –  April 2013 (two and half years ago) –  Apache License 2.0 –  Latest Release: Version 0.8.2 (November 2015)

•  Deployed at > 100 companies

3

4

Tachyon: one of the Fastest GrowingBig Data Open Source Project

Contributors Growth

5

v0.4!Feb ‘14

v0.3!Oct ‘13

v0.2 Apr ‘13

v0.1 Dec ‘12

v0.6!Mar ‘15

v0.5!Jul ‘14

v0.7!Jul ‘15

1 3 15

30

46

70

111

Contributors Growth

6

> 170 Contributors (V0.8) (3x increment over the last AMPCamp)

> 50 Organizations

Thanks to Contributors and Users!

7

h"p://tachyon-project.org/community/

One Tachyon ProductionDeployment Example

•  Baidu (Dominant Search Engine in China, ~ 50 Billion USD Market Cap)

•  Framework: SparkSQL •  Under Storage: Baidu’s File System •  Storage Media: MEM + HDD •  100+ nodes deployment •  1PB+ managed space •  30x Performance Improvement

8

Outline

•  Open Source


•  New Features


9

Tachyon is an Open Source

Memory-centricDistributed

Storage System 10

11

Why Tachyon?

Performance Trend: Memory is Fast

•  RAM throughput increasing exponentially

•  Disk throughput increasing slowly

12

Memory-locality key to interactive response times

Price Trend: Memory is Cheaper

source:jcmit.com13

Realized by many…

14

15

Is the Problem Solved?

16

Missing a Solution for the Storage Layer

A Use Case Example with -

•  Fast, in-memory data processing framework – Keep one in-memory copy inside JVM – Track lineage of operations used to derive data – Upon failure, use lineage to recompute data

map

filter map

join reduce

Lineage Tracking

17

Issue 1

18

Data Sharing is the bottleneck in analytics pipeline:Slow writes to disk

Spark Job1

Spark mem block manager

block 1

block 3

Spark Job2


block 3

block 1

HDFS / Amazon S3 block 1

block 3

block 2

block 4

storage engine & execution engine same process (slow writes)

Issue 1

19

Spark Job


block 1

block 3

Hadoop MR Job

YARN


block 3

block 2

block 4

Data Sharing is the bottleneck in analytics pipeline:Slow writes to disk

storage engine & execution engine same process (slow writes)

Issue 1 resolved with Tachyon

20

Memory-speed data sharingamong jobs in different

frameworks execution engine & storage engine same process (fast writes)

Spark Job

Spark mem

Hadoop MR Job

YARN


block 3

block 2

block 4

HDFSdisk

block1

block3

block2

block4Tachyon!in-memory

block 1

block 3 block 4

Issue 2

21

Spark Task

Spark memory block manager

block 1

block 3


block 3

block 2

block 4

execution engine & storage engine same process

Cache loss when process crashes

Issue 2

22

crash


block 1

block 3


block 3

block 2

block 4



HDFS / Amazon S3

Issue 2

23

block 1

block 3

block 2

block 4


crash



block 3

block 2

block 4 Tachyon!in-memory

block 1

block 3 block 4


24

Spark Task



Keep in-memory data safe,even when a job crashes.


25

HDFSdisk

block1

block3

block2

block4


Tachyon!in-memory

block 1

block 3 block 4

crash


block 3

block 2

block 4

Keep in-memory data safe,even when a job crashes.

HDFS / Amazon S3

Issue 3

26

In-memory Data Duplication & Java Garbage Collection

Spark Job1


block 1

block 3

Spark Job2


block 3

block 1

block 1

block 3

block 2

block 4

execution engine & storage engine same process (duplication & GC)


27

No in-memory data duplication,much less GC

Spark Job1

Spark mem

Spark Job2

Spark mem


block 3

block 2

block 4

execution engine & storage engine same process (no duplication & GC)

HDFSdisk

block1

block3

block2

block4Tachyon!in-memory

block 1

block 3 block 4

Previously Mentioned

•  A memory-centric storage architecture

•  Push lineage down to storage layer

28

Tachyon Memory-Centric Architecture

29

Tachyon Memory-Centric Architecture

30

Lineage in Tachyon

31

Outline

•  Open Source




32

1) Eco-system: Enable new workload in any storage;

Work with the framework of your choice;

33

2) Tachyon running in production environments,

both in the Cloud and on Premise.

34

Use Case: Baidu

•  Framework: SparkSQL •  Under Storage: Baidu’s File System •  Storage Media: MEM + HDD •  100+ nodes deployment •  1PB+ managed space •  30x Performance Improvement

35

Use Case: an Oil Company

•  Framework: Spark

•  Under Storage: GlusterFS

•  Storage Media: MEM only

•  Analyzing data in traditional storage

36

Use Case: a SAAS Company

•  Framework: Impala

•  Under Storage: S3

•  Storage Media: MEM + SSD

•  15x Performance Improvement

37

Use Case: a Biotechnology Company

•  Framework: Spark & MapReduce

•  Under Storage: GlusterFS

•  Storage Media: MEM and SSD

38

Use Case: a SAAS Company

•  Framework: Spark

•  Under Storage: S3

•  Storage Media: SSD only

•  Elastic Tachyon deployment

39

Use Case: a Retail Company

•  Framework: Spark & MapReduce

•  Under Storage: HDFS

•  Storage Media: MEM

40

Run Everywhere

41

Enable Faster Innovation in Storage Layer

42

What if data size exceeds memory capacity?

43

3) Tiered Storage:Tachyon Manages More Than DRAM

MEM SSD

HDD

Faster

Higher Capacity

44

Configurable Storage Tiers

MEM only

MEM + HHD

SSD only

45

4) Pluggable Data Management Policy

Evict stale data to lower tier

Promote hot data to upper tier

46

Pin Data in Memory

5) Transparent Naming

47

6) Unified Namespace

48

More Features

•  7) Remote Write Support •  8) Easy deployment with Mesos and Yarn •  9) Initial Security Support •  10) One Command Cluster Deployment •  11) Metrics Reporting for Clients, Workers,

and Master

49

12) More Under Storage Supports

50

Reported Tachyon Usage

51

•  Team consists of Tachyon creators, top contributors

•  Series A ($7.5 million) from Andreessen Horowitz

•  Committed to Tachyon Open Source

•  http://www.tachyonnexus.com

52

Outline

•  Open Source




53

Memory-Centric Distributed Storage

Welcome to try, contact, and collaborate!

54

JIRA New Contributor Tasks

•  Try Tachyon: http://tachyon-project.org

•  Develop Tachyon: https://github.com/amplab/tachyon

•  Meet Friends: http://www.meetup.com/Tachyon

•  Get News: http://goo.gl/mwB2sX

•  Tachyon Nexus: http://www.tachyonnexus.com •  Contact us: [email protected]

55

Tachyon Presentation at AMPCamp 6 (November, 2015)

Technology

Transcript of Tachyon Presentation at AMPCamp 6 (November, 2015)