Introduction

Swapnil Bawaskar@sbawaskar

• Introduction • What? • Who? • Why? • How?

• Q&A

2

Agenda

3

What?

4

What is it?

A distributed, memory-based data management platform for data oriented apps that need: • high performance, scalability, resiliency and continuous

availability • fast access to critical data set • location aware distributed data processing • event driven data architecture

5

What is it?

6

Who?

• 1000+ systems in production (real customers) • Cutting edge use cases

7

Who are the users?

2004 2008 2014

•  Massive increase in data volumes

•  Falling margins per transaction

•  Increasing cost of IT maintenance

•  Need for elasticity in systems

•  Financial Services Providers (every major Wall Street bank)

•  Department of Defense

•  Real Time response needs •  Time to market constraints •  Need for flexible data

models across enterprise •  Distributed development •  Persistence + In-memory

•  Global data visibility needs •  Fast Ingest needs for data •  Need to allow devices to

hook into enterprise data •  Always on

•  Largest travel Portal •  Airlines •  Trade clearing •  Online gambling

•  Largest Telcos •  Large mfrers •  Largest Payroll processor •  Auto insurance giants •  Largest rail systems on

earth

• 17 billion records in memory • GE Power & Water's Remote Monitoring & Diagnostics Center

• 3 TB operational data in-memory, 400 TB archived • China Railways

• 4.6 Million transactions a day / 40K transactions a second • China Railways

• 120,000 Concurrent Users • Indian Railways

8

Who are the users?

World: ~7,349,000,000

~36% of the world population

Population: 1,251,695,6161,401,586,609

China RailwayCorporation

Indian Railways

10

Why?

Numbers Everyone Should Know

11

L1 cache reference 0.5 ns Branch mispredict 5 ns L2 cache reference 7 ns Mutex lock/unlock 100 ns Main memory reference 100 ns Compress 1K bytes with Zippy 10,000 ns 0.01 ms Send 1K bytes over 1 Gbps network 10,000 ns 0.01 ms Read 1 MB sequentially from memory 250,000 ns 0.25 ms Round trip within same datacenter 500,000 ns 0.5 ms Disk seek 10,000,000 ns 10 ms Read 1 MB sequentially from network 10,000,000 ns 10 ms Read 1 MB sequentially from disk 30,000,000 ns 30 ms Send packet CA->Netherlands->CA 150,000,000 ns 150 ms

http://static.googleusercontent.com/media/research.google.com/en/us/people/jeff/stanford-295-talk.pdf

What makes it fast?• No ORM

• Minimize copying

• Minimize contention points

• Run user code in-process

• Partitioning and parallelism

• Avoid disk seeks

• Automated benchmarks

YCSB

13

oper

atio

ns p

er s

econ

d

0

200000

400000

600000

800000

YCSB Workloads

A Re

ads

A U

pdat

es

B Re

ads

B U

pdat

es

C R

eads

D In

serts

D R

eads

F Re

ads

F U

pdat

es

CassandraGeode

Horizontal scaling for reads, consistent latency and CPU

0

4.5

9

13.5

18

Speedu

p

0

1.25

2.5

3.75

5

ServerHosts2 4 6 8 10

speeduplatency(ms)CPU%

• Scaled from 256 clients and 2 servers to 1280 clients and 10 servers• Partitioned region with redundancy and 1K data size

15

How?

• Clone & Build

16

Hands-on: Build & run

gitclonehttps://github.com/apache/geodecdincubator-geode./gradlewbuild

• Start a servercdgemfire-assembly/build/install/apache-geode./bin/gfshgfsh>startlocator--name=locatorgfsh>startserver--name=servergfsh>createregion--name=myRegion--type=REPLICATE

$dockerrun-itapachegeode/geode

• Docker

• Downloadhttp://geode.apache.org/releases/

• Locator

• Discovery service

• JMX manager

• Cluster config manager

• Servers

• Stores data

• Embeddable within your application

• Clients

• Your Application

17

Member Types

Client

Locator

Server

• Region

• Distributed java.util.ConcurrentHashMap on steroids (Key/Value)

• Consistent API regardless of where or how data is stored

• Observable (reactive)

• Highly available, redundant on cache Member (s).

Concepts - Region

18

Region

Cache

java.util.Map

JVM

Key Value

K01 May

K02 Tim

• Region

• Local, Replicated or Partitioned

• In-memory or persistent

• Redundant

• LRU, TTL

• Overflow

Region Options

19

Region

Cache

java.util.Map

JVM

Key Value

K01 May

K02 Tim

Region

Cache

java.util.Map

JVM

Key Value

K01 May

K02 Tim

LOCALLOCAL_HEAP_LRULOCAL_OVERFLOWLOCAL_PERSISTENTLOCAL_PERSISTENT_OVERFLOWPARTITIONPARTITION_HEAP_LRUPARTITION_OVERFLOWPARTITION_PERSISTENTPARTITION_PERSISTENT_OVERFLOWPARTITION_PROXYPARTITION_PROXY_REDUNDANTPARTITION_REDUNDANTPARTITION_REDUNDANT_HEAP_LRUPARTITION_REDUNDANT_OVERFLOWPARTITION_REDUNDANT_PERSISTENTPARTITION_REDUNDANT_PERSISTENT_OVERFLOWREPLICATEREPLICATE_HEAP_LRUREPLICATE_OVERFLOWREPLICATE_PERSISTENTREPLICATE_PERSISTENT_OVERFLOWREPLICATE_PROXY

• Object Query Language (OQL) • SQL like • Query Complex Objects, attributes, methods • Not as performant at get()

Concepts - OQL

20

class Portfolio { int ID; String type; String status; Map positions; }

class Position { String secId; double mktValue; double qty; }

• SELECT * FROM /portfolio WHERE status = ‘active’ • SELECT p, pos FROM /portfolio p, p.positions.values pos WHERE pos.secId

= ‘VMW' • SELECT DISTINCT * FROM /portfolio p WHERE p.positions.size >= 2

Adapters

21

• write-through as opposed to cache-aside

• Stale Cache • Inconsistent Cache • Thundering Herds

memcached

22

• Scalable Data-Structures • Use All Cores • WAN Replication

Redis

23

Hands on

• Teeny repo at

URL shortener

25

https://github.com/sbawaska/teeny

cdapache-geode./bin/gfshgfsh>startlocator--name=locatorgfsh>startserver--name=server1--server-port=40404gfsh>startserver--name=server2--server-port=40405gfsh>createregion--name=myRegion--type=REPLICATE• From the Application:

• Create regions • Save Teeny • Lookup Teeny/update stats • Remove Teeny • Query Teeny

• Download Geodehttp://geode.apache.org/releases/