EMC ScaleIO Overview

1© Copyright 2014 EMC Corporation. All rights reserved.

ScaleIO

Convergence. Scalability. Performance.Elasticity.


Software-Defined Converged Server SANEMC ScaleIO

Software that creates a server-based SAN from local storage to deliver performance and capacity on demand

Creates virtual pool of storage with varying performance tiers

Media-agnostic: Leverages any HDDs, SSDs, & PCIe flash cards

Server-agnostic: Installed on existing physical & virtual app servers

ScaleIO


ScaleIO Turns DAS Into A Server SAN

Performance—massive I/O parallelism

Elastic—add/remove servers & capacity “on-the-fly”

Convergence of storage and compute

Scale-out to thousands of servers


Simplify Your data center ArchitectureConvergence

Converge storage and compute resources into a single-layer architecture

Aggregate capacity and I/O performance

Make storage as ubiquitous as CPU & RAM

Consolidate storage and applications onto x86 and ARM hardware

Servers

Network

Storage

ConvergedArchitecture


Scalability

Massively scale to thousands of nodes

Add storage devices and servers modularly to increase capacity and performance

Storage growth always automatically aligned with application needs


Elasticity

Dynamically add, move, remove storage and compute resources “on the fly” with no downtime

No capacity planning required

Leverage mixed:– Server brands– Configurations– OS platforms (physical & virtual)– Media types


Massive I/O Parallelism

I/O processing load shared among all servers

Automatic rebuild & rebalance

Autonomous management of performance hot spots and data layout

Storage tiering

Performance

Eliminate I/O Bottlenecks


Before ScaleIOWith ScaleIO

1,000 IOPS 10 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB100 IOPS

1 TB

2,000 IOPS 20 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB

100 IOPS1 TB


ScaleIO EnvironmentEliminate the Need for Array, Switching Fabric, and HBAs

FLEXIBILITY IN EXPANSION


Add Nodes or Disks Dynamically—System Automatically Migrates and Rebalances Storage

Auto-Rebalance


Remove Nodes or Disks Dynamically—System Automatically Migrates and Rebalances Storage

Auto-Rebalance


Simple Management & Monitoring Automated installation scripts and easy

configuration

Any IT admin can manage the entire data center stack

Call-home and incident reporting

Self-regulated data layout, performance optimization, and HW/SW failure recovery—no administrative intervention required


Bandwidth/ IOPS Limiter for

Quality of Service

For Infrastructure as a ServiceEnterprise Features

Writeable Snapshots for

Backups

Protection Domains for

Isolated Servers and Data Sets

Storage Pools for Tiering

Control Performance, Capacity, and Data Location

Light Data Encryption at

Rest for Security

APP 1

APP 2

APP 3


How ScaleIO Lowers Your TCOYour Bottom Line

1. Commodity hardware

2. No dedicated storage components (FC network, HBAs, etc.)

3. Reduced power, cooling, and space

4. No “forklift” upgrades for EOL HW

5. Reduced administrative overhead

6. “Pay as you grow”—linear investments with predictable costs


USE CASES

VSI VDI

DATABASES TEST & DEV


Virtual Server Infrastructure (VSI)

Large storage capacities

Easy growth

Deliver low $/server

Easy manageability

Any hardware

Any size growth

No capacity planning

Easy growth in capacity and performance

Ease of management

Low TCO

Requirements: Why ScaleIO:

VSI



Virtual Desktop Infrastructure (VDI)

Performance at peak times (“boot storms”)

Large storage capacities

Easy growth

Deliver low $/desktop

Parallel processing

Any hardware

Any size growth

No capacity planning

Easy growth in capacity and performance

Ease of management

Low TCO

VDI


Databases

High write performance

High availability

Quick recovery

Low cost of storage

Convergence allows for fast writes

Massive parallelism delivers quick recovery

Stable, predictable performance

Low TCO


DATABASES


Development & Testing

Medium capacity & performanceLow cost Rapidly changing environment with high deduplication rateRapid hardware changes and repurposing

Rebalancing for frequently changing environmentsEasy scale-out and hardware upgrade with no downtimeLow initial cost allows more investment in computeStable, predictable performanceEasy management


TEST & DEV


ARCHITECTURE


ETH/IB

ScaleIO Data Client (SDC)

HOST

SDCCScaleIO

Protocol

DAS

Exposes ScaleIO Volumes to Application


ScaleIO Data Server (SDS)

SDSS

HOST

ETH/IBScaleIO

Protocol

Owns Local Storage Dedicated to ScaleIO

DAS


Configures and monitors the ScaleIO system

Keeps track of storage and data mapping

Monitors capacity, performance, and load balancing

Makes data migration decisions

Metadata Manager (MDM)


Non-VMware Environments

• Identical to physical environments

• SDS and SDC sit inside hypervisor

Hypervisor

VM VM VM………

ETH/IB

C

S

Server


Hypervisor

VMware Environments

SVM: Dedicated ScaleIO VM in each ESX host containing SDC and SDS

SDS: Accesses storage volumes via VMFS

SDC: Exposes volumes as iSCSI targets to VMFS

Hypervisor: iSCSI initiator, exposing VMFS to VMs

VMVM

Server

SVM

ETH/IB

C

S

VMFSVMFS

SCSI

……


ETH/IB

Single Read I/O

C

ETH/IB

S


ETH/IB

Single Write I/O

C

ETH/IB

S S


Fully Converged Configuration

ETH/IB

S CS S S S S

S S S S S S

S S S S S S

C C C C C C

C C C C C C

C C C C C C


Two-Layer Configuration

ETH/IB

S S S S S S

C C C C C C



S S S S S S

S S S S S S

ETH/IB

C C C C C C


Massive Parallelism as SDCs Contact Relevant SDSs Directly


ETH/IB

S S S S S S

C C C C C C


Massive Parallelism as SDCs Contact Relevant SDSs Directly

Fully Converged Configuration

S CS S S S S

S S S S S S

C C C C C C

C C C C C C

ETH/IB


Distributed VolumesSDS 1

SDS 5 SDS 6 SDS 100

SDS 2 SDS 3 SDS 4

Volume 1

Volume 2

…


B

AA

Two-Copy “Mesh” Mirroring

B

C

DD

C

EE


Data Protection

FFBE EE

BB

F


Protection Domains

• Protection Domain: A

logically segregated

grouping of SDSs (nodes

and storage)

• Provides security, data

location, and performance

segregation


Pool 1 Pool 3Pool 2

Storage Pools

• Storage pool: A subset of devices in a protection domain created for performance isolation and data segregation

• Enables tiering between device types

• Can be asymmetric

FLASHDISKS

Protection Domain


Writeable Snapshots

ScaleIO volumes that can be written, read and copied as snapshots

Instantaneous and thinly provisioned to save space

Organized in trees, known as VTrees

Consistency Group: collection of snapshots created together at the same time

C1

V1

S121

S111 S112

VTree1

V2

S211

VTree2

Consistency Group


For Quality of ServiceBandwidth/IOPS Limiter

Current Environment

AVAILABLE

With IOPS Limiter

APP 3APP 2APP 1


Why not rescue that capacity and put it to good use?

EMC ScaleIO Overview

Technology

Transcript of EMC ScaleIO Overview