Transcript
Page 1: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

1

OBJECTIVE ANALYSIS – Semiconductor Market Research

How Many IOPS is

Enough?

Tom Coughlin

Coughlin Associate

&

Jim Handy

Objective Analysis

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Outline

• The Survey

• Application Distribution and Attributes

• More Survey Results: IOPS, Capacity and

Latency

• Developing tiers of storage for enterprise

(and client) applications

• Implications/Projections

• Authors & Sources

2

Page 2: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

2

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Outline

• The Survey

• Application Distribution and Attributes

• More Survey Results: IOPS, Capacity and

Latency

• Developing tiers of storage for enterprise

(and client) applications

• Implications/Projections

• Authors and Sources

3

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Our Survey

• Five-minute survey asked end users

– IOPS needs

– Capacity

– Latency

– System bottleneck IOPS

– Primary application

• Nearly 200 respondents

• Report analyzes and interprets the results

in depth4

Page 3: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

3

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Outline

• The Survey

• Application Distribution and Attributes

• More Survey Results: IOPS, Capacity and

Latency

• Developing tiers of storage for enterprise

(and client) applications

• Implications/Projections

• Authors and Sources

5

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Applications Breakdown

6

Archiving and backup

Video Creation or Distribution

Cloud storage or services

OLTP

Scientific or Engineering

Databases

Mail server and mail storage

Page 4: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

4

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Application Types

• Cloud Storage/Service-Virtualization

• Databases

• On-Line Transaction Processing (OLTP)

• Video Creation and Distribution

• Science & Engineering

• Exchange Servers

7

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Cloud Storage/Services& Virtualization

• The “IO Blender”– Many streams

– Scrambled I/O

– Highly random

• Suits SSDs better than

HDDs for rapid access

• Many VM and VDI

systems using flash cache

to meet demand speed

needs Image courtesy of Waring Corp.8

Page 5: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

5

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Databases

• Large data sets

• Random traffic

• High I/O load

• Early SSD adopter

(and before that

used DRAM-based

SSDs)

• Some users load

their entire DB on

flash memory9

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

OLTP(On-Line Transaction Processing)

• Verified writes– Write/read back

– Doubles I/O load

• No room for errors

• Speed is

imperative– Delays lose customers

10

Page 6: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

6

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Video Creation or Distribution

• Large data sets

• Multiple video

streams– Randomizes access

• High bandwidth

required

• Expensive talent– Don’t want them sitting

around waitingImage courtesy of the US Library of Congress

11

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Recording Media Share inProfessional Video Cameras

2012 Digital Storage for Media and Entertainment Report, Coughlin Associates 12

Page 7: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

7

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Science & Engineering

• Complex problems– Genome sequencing

– CAD/CAM

– Natural Resources

– Nuclear modeling

• Large data sets

• Expensive talent– Don’t want them sitting

around waiting

Image courtesy of Wikimedia Commons

13

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Exchange Server

• Multiple tasks– e-mail

– Scheduling/calendars

– Data storage

• Scads of users

• e-mail chaos– Multiple mailboxes

– Asynchronous sends & receives

– Spam & virus filters

Image courtesy of Dell Computer

14

Page 8: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

8

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Outline

• The Survey

• Application Distribution and Attributes

• More Survey Results: IOPS, Capacity and

Latency

• Developing tiers of storage for enterprise

(and client) applications

• Implications/Projections

• Authors and Sources

15

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Let’s Look At the Overall Results

Report breaks analysis down by application

– That’s 54 charts

– Too much for this presentation!

• We’ll just look at Database

Page 9: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

9

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

IOPS Required

0%

10%

20%

30%

40%R

esp

on

den

ts

10 100 1K 10K 100K 1M 10M

IOPS

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Storage Capacity Needed

0%

5%

10%

15%

20%

25%

Resp

on

den

ts

1GB

10GB

50GB

100G

B

500G

B1T

B5T

B10

TB50

TB

100T

B

Capacity

Page 10: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

10

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

How Do The TwoCompare To Each Other?

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

IOPS vs. Capacity

1

10

100

1,000

10,000

100,000

1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07

IOPS

Ca

pa

cit

y (

GB

)

20

Page 11: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

11

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

System Bottleneck IOPSWhat is the Fastest Storage Your System Can Use?

0%

10%

20%

30%

40%R

esp

on

den

ts

10 100 1K 10K 100K 1M 10M

IOPS

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Minimum Latency Requirement

0%

10%

20%

30%

40%

Resp

on

den

ts

10ns 1µs 100µs 10ms 1 sec

Latency

Page 12: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

12

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Outline

• The Survey

• Application Distribution and Attributes

• More Survey Results: IOPS, Capacity and

Latency

• Developing tiers of storage for enterprise

(and client) applications

• Implications/Projections

• Authors and Sources

23

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

DRAMs Now 6,000x HDD Speed!

From : HDDs and Flash Memory: A Marriage of Convenience

24

Page 13: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

13

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Memory and StoragePrice vs. Bandwidth

1.E+00

1.E+01

1.E+02

1.E+03

1.E+04

1.E+05

1.E+06

1.E+07

1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06

Price per Gigabyte

Ban

dw

idth

(M

B/s

)

Tape

HDD

NAND

DRAM

L3

L2

L1

101

100

102

103

104

105

106

107

101

102

103

104

105

106

100

10-1

From Objective Analysis: Are Hybrid Drives Finally Coming of Age?

25

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Price/GB Roughly Follows IOPS

$0.01

$0.10

$1.00

$10.00

$100.00

$1,000.00

0 200 400 600 800 1,000 1,200

IOPS

$/G

B

26

Page 14: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

14

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Device IOPS by Form Factor

102 103 104 105 106

HDD SATA SAS/FC 2-Hop 1-Hop

27

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Outline

• The Survey

• Application Distribution and Attributes

• More Survey Results: IOPS, Capacity and

Latency

• Developing tiers of storage for enterprise

(and client) applications

• Implications/Projections

• Authors and Sources

28

Page 15: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

15

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Implications/Projections (1)

• SSDs adoption will increase

– Usually more IOPS is better

• Fast storage is changing

– From short-stroked HDDs to SSDs

• HDDs becoming a tier behind SSDs

• Other system elements will become the

bottleneck

– Network, software, servers…

29

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Implications/Projections (2)

• Users will focus more attention on IOPS

– Understanding will be greater than it is today

• Higher IOPS will support data/content

growth

– This means more storage

• SSD

• HDD

• Even tape

30

Page 16: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

16

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Report Compiles Survey Results

• An analysis of complete survey results by application

• Report, published July 2014, can be purchased for immediate download at www.Objective-Analysis.com.

• Orders can also be processed through Coughlin

Assocaites at:

http://www.tomcoughlin.com/techpapers.htm.

• You can contact Coughlin Associates by calling Tom at 408-871-8808, or e-mailing: [email protected].

31

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Analysts

32

Tom Coughlin, President, Coughlin Associates is a

highly-respected storage analyst and consultant with

over 30 years in the data storage industry in

engineering and management at high profile

companies.

Jim Handy is a widely recognized semiconductor

analyst, has over 30 years in the electronics industry.

His background includes marketing and design

positions at market-leading suppliers.Jim Handy

Objective Analysis

Thomas Coughlin

Coughlin Associates

Page 17: Data Center Survey: How Many IOPS Is Enough. Compares to site capacity and system latency

17

OBJECTIVE ANALYSIS – www.OBJECTIVE-ANALYSIS.com

Sources

• How Many IOPS do You Really Need?

• 2012 Digital Storage for Media and Entertainment, Coughlin Associates: (www.tomcoughlin.com/techpapers)

• HDDs and Flash Memory: A Marriage of Convenience,

Coughlin Associates and Objective Analysis, 2011

(www.tomcoughlin.com/techpapers)

• Two may be Better than One: Why HDD and Flash Belong

• SNIA SSSI White Paper, Coughlin/Handy 2010

• Are Hybrid Drives Finally coming of Age?, Objective

Analysis, 2010

33


Recommended