Optimizing Storage for Mobile ApplicationsWith X3 Technology

JEDEC Mobile Forum

Copyright @ 2014 SanDisk

Karin Werder

September 2014

Director, Product Marketing Management

With X3 Technology

During our meeting today we may make forward-looking statements.

Any statement that refers to expectations, projections or other characterizations of future events or circumstances is a forward-looking statement, including those relating to industry trends, technology transitions and product performance.

Actual results may differ materially from those expressed in these forward-looking statements due to a number of risks and

2

forward-looking statements due to a number of risks and uncertainties, including the factors detailed under the caption “Risk Factors” and elsewhere in the documents we file from time to time with the SEC, including our annual and quarterly reports.

We undertake no obligation to update these forward-looking statements, which speak only as of the date hereof.

Agenda

� Historical Perspective

� The Era of X3 in Mobile Has Arrived

� Performance Requirements

Deep-Dive Into Mobile Storage Use Cases

3

– Deep-Dive Into Mobile Storage Use Cases

� USB Side Loading

� Summary

Historical Perspective (2005)

4

The Era of X3 (TLC) in Mobile Has Arrived

� Successfully addressing OEM concerns

– Reliability

• Mobile workload

• Improving write and read endurance

– Performance

• Mobile use cases

5

• Mobile use cases

Understanding the Real System Behavior

OS Services

Applications

Block Device Drivers

VFS/FS

6

Em

be

dd

ed

Sto

rag

e

NAND Die Stack

Controller

Firmware

Application

Processor

DRAM

Sensors

Network / I/O

Display

Audio

7

� Cell voltage distribution is not fixed:

– Changes during memory lifetime and cycling

– Variations: changes from block to block, word line to word line, . . .

� Using fixed set default thresholds result in high bit error rate and decoding

failure � Reduces endurance limit

Solution:

� Adaptive read thresholds

Adaptive Read Thresholds Improves Endurance

8

� Adaptive read thresholds

Endurance Coding - Data Shaping

� The Challenge: Increasing endurance

� The Means: Data shaping – transform input data sequence into a “shaped” data sequence which induces less wear when programmed to the NAND

� SLC Example: Transform the input data into shaped data having less 0’s

9

Minimize number of programmed cells per P/E cycle

Control Gate

ONO Dielectric

Floating Gate

Tunnel Oxide (ToX)

Host

1

Erase

0

Prog

Data

Shaping

1

Erase

0

ProgNAND

Memory

Extreme Workload Experiment

FeatureMid-End

Smartphone

High-End

SmartphoneImpact on Storage

Phone CPU Dual Core Quad Core (4+4)

RAM 1GB 2GB More concurrent applications

Camera 5MP

720@30fps

13MP

1080P@30fps

Photo sizes multiplied by ~3

Video sizes multiplied by ~1.3

Storage 8GB

eMMC4.41 (52Mhz 8bit)

16GB

eMMC5.0 (HS400)

Wifi 802.11n 802.11ac Faster downloads

10

Wifi 802.11n 802.11ac Faster downloads

Display 4 inch, 480 x 800 5 inch, 1080 x 1920 High-resolution content

SW Jelly Bean Jelly Bean

Usage Total Time 24.5 hours 24.5 hours

User Activity Time 7.5 hours 7.5 hours

Suspend Time 17.0 hours 17.0 hours Hundreds of MB written

during the night

Workload Written Data 3.9 GB 7.0 GB

User Data vs. System Data

Mid-End Smartphone High-End Smartphone

System/Applications Data 2.4GB 3.1GB

User Files (Generated or Downloaded) 1.5GB 3.9GB

3%3%

9%7% 3%

9%7%3%

Apps

Application UsageSum of Minutes

11

9%

16%

6%

5%

5%18%

8%

9%

16%

6%

5%

5%18%

20%

8%

7% Apps

Audio

Connectivity

Email

Gaming

Location-based Services

Multimedia

Social Networking

Talk + Text

Utility

Web Browsing

Mobile Workloads and Endurance

Mobile Segment

DailyWorkload

TBWAverage Capacity

Estimated Required NAND

Endurance*

High-End 8.0 GB 14.0 (5Y) 32GB 1K

Mid-End 3.9 GB 5.7 (4Y) 16GB 700

12

Mid-End 3.9 GB 5.7 (4Y) 16GB 700

Entry-Level 2.5 GB 2.7 (3Y) 8GB 700

*Assuming write amplification of 2. Actual numbers are dependent on use case and product implementation.

Read Refresh Advantagee.MMC vs. Raw NAND Read Endurance Specification

NAND Read

Endurance

Product useful

life read

endurance

Early life

read

endurance

13

P/E Cycles

End of life read

endurance

Performance RequirementsDeep-Dive Into Mobile Storage Use Cases

14

Deep-Dive Into Mobile Storage Use Cases

Write Performance Over TimeRecorded on Leading Android-Based Smartphone

Megabytes per Second — Writes

/sec

15

MB

ytes

/sec

Social Media Use Case2.5 hours

Parameter Total

Total Data Written 670.23MB

Using Facebook, uploads pictures/videos, browse feeds, watch movies, write

messages, add comments, search people /groups, idle

Chunk Size IOs Distribution – Writes

(Facebook)

4KB:

16

Total Data Read 5.80GB4KB:

8KB:

12KB:

16KB:

32KB:

Other:

Social Media BW Over TimeMeasured on Leading Android-Based Smartphone

Megabytes per Second – Writes (FB_total_Write)

12MB/sec (Video recording –

media server)

7MB/sec (Taking picture –camera) 6.7MB/sec

(Browse groups –browser)

6.5MB/sec (Taking picture –

camera)

11.2MB/sec (Video recording –

media server)

17

browser) camera)

3.3MB/sec (browse feeds –

Facebook)

MB

ytes

/sec

Facebook Process – Storage Activity Profiling

Max Write: 250 IO/Sec

Operations per Second – Writes (FB_Write)

Max Read: 522 IO/Sec

Operations per Second – Reads (FB_Read)

18

Camera Burst ShotLeading Android-Based Smartphone

� 2x30 photos

� 4MB per photo (on average)

Megabytes per Second – Writes, Reads

19

Burst #1

– 40MB/S Max

Burst #2 –

40MB/S Max

MB

ytes

/sec

Gaming - Real Racing 2014Leading Android-Based Smartphone

Task Write Read

Installation 2.08GB 146MB

Launch 143.61MB 183.99MB

App Size (Google Play): 41MB

20

Play (~5 min) 55.59MB 93.11MB

Gaming - Real Racing 2014Leading Android-Based Smartphone

Installation PlayLaunch Game

21

Gaming - Real Racing 2014Leading Android-Based Smartphone

Write –

33MB/sec (Install)

Read –

Read –

26MB/sec (Play)

Megabytes per Second – Writes, Reads, Comments

/sec

22

Read –

17MB/sec (Launch)

MB

ytes

/sec

Productivity Use Case: PowerPoint (Office Suite Pro) Leading Android-Based 12.2” Tablet

� Generate 100 slides in PowerPoint

– 34MB file

� Animation, insert pictures, usage of S-pen

to write parts of text and save file in-between

23

Parameter Total

Total Data Written 311.0MB

Total Data Read 14.2MB

*Note: data written to flash is almost 10 times bigger than the file created.

Productivity Use Case: PowerPoint (Office Suite Pro) Leading Android-Based 12.2” Tablet

Megabytes per Second – Writes12.2MB/sec

24

MB

ytes

/sec

USB Side Loading

25

USB Side Loading

USB 2.0 to Internal Storage

Zoom

~285ms

~19.75MB

USB 2.0 is the bottleneck . . .

26

Regular PerformanceTotal Data – 1024MB

Total Time – 32.424s

Rate – 31.58MB/s

Write chunk size – 94.8% 512K

USB 3.0 to Internal StorageLeading Android-Based Smartphone

USB 3.0 – major improvements but still some bottleneck observed

Zoom

~15.46MB

~163ms

27

Zoom X2

Continuous Continuous

Regular PerformanceTotal Data – 1024MB

Total Time – 21.482s

Rate – 47.67MB/s

Summary

� Advance system solutions (HW, FW) enable highly reliable X3

solutions for mobile applications

� Performance requirements in mobile are burst intensive

– A lot of idle time is an integral part of usage

– High performance needed in short spikes

� Optimizing X3 products to the mobile use case will enable

28

� Optimizing X3 products to the mobile use case will enable

– Best cost structure

– Excellent user experience