ULLtraDIMM SSD Overview - SNIA ULLtraDIMM_0.pdf · ULLtraDIMM™ SSD Overview . Rob Callaghan ....

Preview:

Citation preview

1 c

ULLtraDIMM™ SSD Overview

Rob Callaghan

June 9th, 2014

2

A Global Leader in Flash Storage Solutions Trailing 4 Qtr Financials*

$6.2B Revenue

$3.6B Net Cash*

$0.7B R&D Investment

All Leading Smartphone & Tablet Manufacturers use SanDisk

Leading Retail Brand°

#1 Global Retail Revenue Share

SanDisk Client & Retail SSDs Approved Supplier to All Leading PC Manufacturers

Qualified at 6 of the Top 7

Server & Storage OEMs

Enterprise SSDs and Storage Software

Global Operations

5,500 Employees†

*Financials as of Q4, ‘13. Net Cash = [Cash + cash equivalents + short-term & long-term marketable securities] less [debt at maturity value] as of the end of Q4, ‘13. †Headcount as of Jan., ‘14. NPD Estimate, Nov., ‘13. Estimates of the memory card & USB markets from NPD (Nov. ‘13) and GfK Retail and Technology, Sep., ‘13.

Rankings

3

Enabling Flash Storage from Wafer to Software

CONTROLLER SCALE ASSY, TEST & PACKAGING

SSD SOFTWARE NAND TECH NAND DIE FLASH MGMT

Close to Half of Industry Bit Output Together with manufacturing partner Toshiba

Fabs: World class NAND capacity

Patents as of Oct., ‘13; NPD Estimate, Nov., ‘13. Gartner: NAND Flash Supply & Demand, WW 1Q ‘12-4Q ‘14, 3Q ’13. Update Dec., ‘13.

World-Leading Innovator 4,900+ Patents

1991 2013

4

The Path to Ultra Low Latency & Scalable Performance La

tenc

y (µ

sec)

1’s

10’s

1000’s

100’s

IOPS 100 100,000 1,000,000

SAS/

SATA

/FC

PCIe

D

DR

Flas

h St

orag

e on

the

high

sp

eed

mem

ory

bus

5

This is a SATA DIMM

This is DRAM with battery backup

Flash Flash

Flash Flash

Flash Flash

Flash Flash

Creating a New Storage Interface

This is DDR3 directly to flash

6

Scalable I/O Performance, Constant Latency

0

10

20

30

40

50

60

70

80

90

100

0

200

400

600

800

1000

1200

1400

1 2 4 8

Late

ncy

(us)

(K) I

OPS

Number of ULLtraDIMMTM Devices

400G

800G

1600G

3200G

Demonstrated At Open Compute Summit in January 2014 IBM 3650 2 socket Server Running RHAT 6.3

7

The World’s First Flash Storage Device on the Memory Channel

Enterprise Class Reliability, Endurance

& Support Cost Effective Memory Channel

Interface

8

ULLtraDIMM Combines MCS with Storage

Enterprise level endurance with MLC 5 year warranty

Guardian Technology ™

7µs write latency 135K Read IOPS 45K Write IOPS

Ultra Low Latency, High Performance

Back up power circuitry Full Data Path Protection 2M hrs MTBF

Enterprise Class Reliability

200, 400 GB Scalable architecture Latest flash technology (19nm MLC)

Scalable, Cost Effective Media

DDR3 protocol Configured as block device (through

device driver)

Memory Channel Interface

9

Guardian TechnologyTM is Key Enabler

Deep Flash Characterization

Understanding Flash Capability

Aggregated Flash Management

Use Each Flash Device to its Full

Capability

Advanced Signal Processing

Optimizing Flash Settings Over Product Life

19nm MLC NAND

Value Based High Performance High Endurance

Enterprise Class SSD

0.5 DWPD Up to 10 DWPD

Base 19nm MLC NAND – 0.5 DWPD

Endurance enhanced MLC NAND – 10 DWPD

10 10

ULLtraDIMM Series

ULLtraDIMM

Usage Model Ultra low latency + scalable IOPS

Capacity 200GB 400GB

Form Factor RDIMM

Endurance (Random) 10 DWPD

MTBF 1.5M Hours

Warranty 5 years

11

Utilizing Existing Server Infrastructure Scalable I/O & Capacity DDR3 Interface & Protocol

Example: IBM x3850 X6 server

Scalable up to 32 ULLtraDIMM’s Up to 12.8TB capacity >4.5M IOPS in a single server

12

IBM Redpaper

Lower is better

Near Linear Scaling

13

Software Architecture

Applications Management

Software

OS Stack Block Layer

MCS Kernel Driver

MCS Firmware

BIOS/UEFI

Hardware

Kernel space

User space

Diablo OEM 3rd Party

Flash Controller Firmware SanDisk

14

ULLtraDIMM Population Schema

CPU0

Core0

Core2

Core4

Core6

Core1

Core3

Core5

Core7

Mem Chan

Mem Chan

Mem Chan

Mem Chan

CPU1

Core0

Core2

Core4

Core6

Core1

Core3

Core5

Core7

Mem Chan

Mem Chan

Mem Chan

Mem Chan

CPU3 Core0

Core2

Core4

Core6

Core1

Core3

Core5

Core7

Mem Chan

Mem Chan

Mem Chan

Mem Chan

CPU2 Core0

Core2

Core4

Core6

Core1

Core3

Core5

Core7

Mem Chan

Mem Chan

Mem Chan

Mem Chan

Operates from 800 – 1,600 MT/s – Supports 1600MT/s – Interoperable with standard

RDIMM in the same channel Can be populated to support many

different performance points – Minimum Requirement =

1 RDIMM in the system Recommended population:

– 1 ULLtraDIMM per channel – Matched with 1 RDIMM per

channel

DIMM slots

# of UD’s

# of RDIMMs

Flash Capacity

DRAM Capacity

Read IOPS

Write IOPS

24 2 22 800GB 352GB 300K 130K

24 4 20 1.6TB 320GB 600K 260K

24 8 16 3.2TB 256GB 1.2M 520K

Examples

15

Bus Utilization Q&A Does ULLtraDIMM utilize the memory bus for FTL or other flash management?

– No, Using its embedded processors, ULLtraDIMM performs all flash management on the device itself and only transfers on DDR3 bus after its internal data processing is complete houg.

• Analogous to a host bus adapter

Does the ULLtraDIMM slow down the memory bus – No, but ULLtraDIMM is limited by memory population guidelines (no different than an RDIMM)

– Memory controller will map out a slow device if it does not meet expected performance

– It is possible but can be mitigated by adding more parallel data processing by adding additional ULLtraDIMMs

Can the ULLtraDIMM run slower than optimal under situations where its internal buffer gets saturated?

– It is possible but can be mitigated by adding more devices for additional parallel processing

16

SanDisk BIOS Modification for UlltraDIMMTM Why BIOS modifications are required for UlltraDIMMTM (UD)

– Need to manage UD differently when detected (i.e. not as a standard DRAM) – Memory reservation (E820) and type (MTRR) tables need UD awareness – Initialization of UD hardware – ACPI table updates

BIOS modification guide contains following 4 high-level steps – MRC changes: disable memory test and interleaving of UD – E820 table: UD address space marked as reserved – MTRR table: UD address space marked as write-back – Training Sequence : SanDisk UD Training to be uploaded into UEFI – ACPI table: entries needed to expose UD to the OS device driver

17

CPU Utilization SanDisk is in the process of finalizing the CPU usage based on production level hardware

and software

General guideline is to assign one Hyper thread for Two ULLtraDIMMs (or ~200K to 270K iops max)

H1 H2

Core

H1 H2

Core

H1 H2

Core

H1 H2

Core

H1 H2

Core

H1 H2

Core

H1 H2

Core

H1 H2

Core

18

ULLtraDIMM Usage Model Ultra low latency + scalable IOPS

Capacity 200GB, 400GB

Form Factor RDIMM

Endurance (Random) 10 DWPD

MTBF 1.5M Hours

Warranty 5 years

Software Drivers Linux RedHat Linux SuSe Microsoft Windows Vmware ESX

6.3 SLES 11 SP1 2008R2 5.1U1

6.4 SLES 11 SP2 2012 5.1U2

6.5 SLES 11 SP3 2012 R2 5.5

ULLtraDIMM Series

19

Benefits to Applications

Memory extension Reduce response

times for analytics queries * Future roadmap

In Memory Compute*

Block device Utilizes empty

DIMM slots Enables high

density storage blades

Blade Server

Block device Increased VMs per

Node Faster response

times per VM

Virtualization

Block device/ memory extension* Increase

Transactions per Second Memcached

consolidation

Database/Cloud

Block device Low, predictable

latency Fast Interactive

Data Analysis

Financial Services

20

Lowest latency SSD solution in the market Near linear Scalability in IOPS and bandwidth with parallel processing architecture

Benefits to Applications

Latency sensitive

Write intensive

Read intensive

Bandwidth intensive

IOPS intensive

Good for Ulltradimm

OLTP database No SQL database Virtual Desktop High Frequency Trading Web 2.0

21

E-trading Performance Using ULLtraDIMM

Use ULLtraDIMM as primary storage for messages and events

Architecture Enables: – Reduced Transaction Latency

– Increased Transaction Throughput

– Lower Transaction Latency Jitter

ULLtraDIMM latency approaches RAMDisk

performance using low cost 19nm MLC NAND

22

Reduced Latency Enables Real-Time Analytics

The Application Has Become the Bottleneck in E-Trading

23

Example: Blade Servers

Enables high density, high performance storage

Block and memory expansion storage Improves blade server efficiency

Requires dedicated blade server slot Decreases server efficiency Custom Form Factor

24

Summary First enterprise-class SSD that utilizes

the memory bus

ULLtraDIMM takes advantage of the 12GB/s memory bus lanes

Lowest latency SSD in the market

Parallel architecture allows high IOPS and bandwidth performance

25 c

Thank You!

SanDisk is a trademark of SanDisk Corporation, registered in the United States and other countries. ULLtraDIMM is a trademark of SanDisk Enterprise IP LLC. Other brand names mentioned herein are for identification purposes only and may be the trademarks of their respective holder(s). *1 GB = 1,000,000,000 bytes. Actual user capacity less.

Recommended