Cloudify Your Business

Preview:

DESCRIPTION

 

Citation preview

CLOUDIFY YOUR BUSINESS

Wayne wallscloud evangelist

Wednesday, August 21, 13

Wednesday, August 21, 13

3

Self-Service

Wednesday, August 21, 13

4

On-Demand

Wednesday, August 21, 13

5

Metered

Wednesday, August 21, 13

6

Resource Pooling

Wednesday, August 21, 13

7

Broad Network Access

Wednesday, August 21, 13

NIST SAys...

8

Wednesday, August 21, 13

NIST says...

9

Wednesday, August 21, 13

NIST says...

10

Wednesday, August 21, 13

11

Wednesday, August 21, 13

12

Financial Services Data Analysis & ForecastingMapReduce Data AnalyticsDistributed, Web 2.0 Apps

REVOLUTIONARY

Wednesday, August 21, 13

13

SAPPeopleSoftSharePointSQLHome GrownClient-Server Based

EVOLUTIONARY

Wednesday, August 21, 13

14

Big Promise of Cloud

Wednesday, August 21, 13

15

Scale!

Wednesday, August 21, 13

16

High-Availability vs Service Resiliency

Wednesday, August 21, 13

17

What is HA?

Active/PassiveActive/Active

Wednesday, August 21, 13

18

:Problem:Single Point of Failure

Wednesday, August 21, 13

19

:Solution:MAKE TWO OF ‘EM!

Wednesday, August 21, 13

20

Hammer & Nail Solution

Wednesday, August 21, 13

21

High-Availability has Problems

Wednesday, August 21, 13

22

Fails In Terrible Ways &Doesn’t Really Scale

Wednesday, August 21, 13

23

Two Options to “HA”

Wednesday, August 21, 13

24

Take a non-distributed system and drop HA on top of it

Option #1

Wednesday, August 21, 13

25

NFS

Wednesday, August 21, 13

26

Mean Time Between Failures (MTBF)

Wednesday, August 21, 13

27

Wednesday, August 21, 13

28

Mean Time To Recovery (MTTR)

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

System not designed to be distributed

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

System not designed to be distributed

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

System not designed to be distributed

Failure forces it to be distributed

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

System not designed to be distributed

Failure forces it to be distributed

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

System not designed to be distributed

Failure forces it to be distributed

Cannot take distributed failure conditions into account

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

System not designed to be distributed

Failure forces it to be distributed

Cannot take distributed failure conditions into account

Wednesday, August 21, 13

29

Expected failures of a traditional HA system are catastrophic

System not designed to be distributed

Failure forces it to be distributed

Cannot take distributed failure conditions into account

Best case scenario: complete failure

Wednesday, August 21, 13

30

Take a distributed system and make the right tradeoffs

Option #2

Wednesday, August 21, 13

31

Service Resiliency

Wednesday, August 21, 13

32

Examples

Wednesday, August 21, 13

33

Percona XtraDBCassandra

Riak

Wednesday, August 21, 13

34

CephGlusterFS

Wednesday, August 21, 13

35

Constant Health Checks

Wednesday, August 21, 13

36

What is in a Running Application?

Wednesday, August 21, 13

37

Availability vs Reliability

Wednesday, August 21, 13

38

Wednesday, August 21, 13

38

Reliability is the likelihood that a given component or system will be functioning when needed as measured over a given period of time.

Wednesday, August 21, 13

38

Reliability is the likelihood that a given component or system will be functioning when needed as measured over a given period of time.

Availability is the percentage of times that a given system will be functioning as required.

Wednesday, August 21, 13

39

Server Service

Wednesday, August 21, 13

40

Hardware does not get a pass...

Wednesday, August 21, 13

41

Think availability from the service layer(s) perspective

Wednesday, August 21, 13

42

Wednesday, August 21, 13

43

Apps in a Cloudy World

Wednesday, August 21, 13

44

Greenfield vs Legacy

Wednesday, August 21, 13

45

Greenfield

Wednesday, August 21, 13

45

Greenfield

1. Focus on the service, not the server

Wednesday, August 21, 13

45

Greenfield

1. Focus on the service, not the server2. Identify & tear apart stateless and stateful parts of your application

Wednesday, August 21, 13

45

Greenfield

1. Focus on the service, not the server2. Identify & tear apart stateless and stateful parts of your application3. Make stateful parts redundant using distributed data stores

Wednesday, August 21, 13

45

Greenfield

1. Focus on the service, not the server2. Identify & tear apart stateless and stateful parts of your application3. Make stateful parts redundant using distributed data stores4. Know the dependencies of your system and the impact of failure

Wednesday, August 21, 13

45

Greenfield

1. Focus on the service, not the server2. Identify & tear apart stateless and stateful parts of your application3. Make stateful parts redundant using distributed data stores4. Know the dependencies of your system and the impact of failure5. Use micro-services to make dependencies explicit

Wednesday, August 21, 13

46

Legacy

Wednesday, August 21, 13

46

Legacy

1. Cloud Instances != server

Wednesday, August 21, 13

46

Legacy

1. Cloud Instances != server2. Plan to reduce mean time to recovery (MTTR)

Wednesday, August 21, 13

46

Legacy

1. Cloud Instances != server2. Plan to reduce mean time to recovery (MTTR)3. "We're HA, we're all good." -> Wrong.

Wednesday, August 21, 13

46

Legacy

1. Cloud Instances != server2. Plan to reduce mean time to recovery (MTTR)3. "We're HA, we're all good." -> Wrong.4. Think about stateful vs stateless parts of your application and work piece by piece

Wednesday, August 21, 13

46

Legacy

1. Cloud Instances != server2. Plan to reduce mean time to recovery (MTTR)3. "We're HA, we're all good." -> Wrong.4. Think about stateful vs stateless parts of your application and work piece by piece5. Be creative about trade-offs:  many apps that run on more than one server have some type of common backend (NFS)

Wednesday, August 21, 13

Thank You!

Wednesday, August 21, 13

Recommended