WS2012 File System Enhancements: ReFS and Storage Spaces

Rick ClausSr. Technical Evangelist@RicksterCDNhttp://RegularITGuy.com

WSV316

The Storage and File Systems (SFS) group builds the platform

storage features shipping across all versions of Windows

Ease of storage administration

Storage Management PowerShell with SMI-S integration

Minimize Op-Ex

Efficient capacity utilization

Thin provisioning integrationMaximize Existing Hardware Dynamic data

movement Offloaded data transfers

Storage virtualization and resiliency

Storage SpacesMinimize Cap-Ex

Maximize Availability

Next-generation scale, availability, and integrity

Resilient File System (ReFS)

Maximize data availability

Modern NTFS CHKDSK

Storage Platform OverviewBenefit Capability Feature

Enabled Deployment ModelsCost-efficient Storage (New) Traditional Storage

Block Access

File Access

JBODSApp Serve

rs

SAS

JBODSCA File

Servers

SAS

App Serve

rs

File

\db

App Serve

rs

FCiSCSI

CA File

Servers

FCiSCSI

App Serve

rs

File

\db

Windows Server 2012 is the most cost-efficient

platform for continuously available and highly scalable data access

Storage Spaces

Storage Virtualization and Resiliency

Storage Spaces delivers a new category of highly capable storage solutions to all Windows

customer segments at dramatically lower price-points and maximized operational simplicity

Cost-Efficient Business Critical Storage

Powerful new platform abstractionsStorage pools: Units of aggregation, administration, isolationStorage spaces: Resiliency, provisioning, and performance

Industry standard commodity storageInterconnects: SAS, USB, and SATAEnclosures: Shared SAS JBOD arrays

Storage Spaces Model

Windows Server 2012 Capabilities

File and Storage Manager

Failover Clustering & CSV v2

ReFS & NTFS

Hyper-V

SMB v3 RDMA

Storage Mgmt. PowerShell

Virtualized Resilient Storage

Mirror Space

Parity Space

Mirror Space

Mirror Space

Mirror Space

Parity Space

Parity Space

Mirror Space

Mirror Space

Mirror Space

Mirror Space

Parity Space

Cost-efficient shared SAS JBOD arrays

Physical or Virtualized Servers

Physical Deployment Topologies

ORSAS

Shared SAS

Capabilities Overview

Optimized and flexible capacity utilization

Resiliency to drive failure

Operational simplicity with support for multi-tenancy

High-availability and scalability

Workload-optimized performance via heterogeneous media

Optimized and Flexible Capacity Utilization

On-demand provisioning with automated capacity reclamation (trim/unmap)

Capacity reclamation for deleted files (NTFS and ReFS)Sub-file capacity reclamation (e.g. through Hyper-V)Scheduled optimization for slab reclamation

Elastic capacity expansionSimply add more drives

60TB Space

60TB Space

60TB Space

60TB Space

Company A Storage Pool 60TB

Space60TB Space

60TB Space

60TB Space

Company A Storage Pool

Thinly provisioned storage spaces

Industry standard commodity storage 36 TB across 72 disks48 TB across 96 disks

Resiliency to Drive Failure

Resilient storageMirror spaces with dirty region trackingParity spaces with integrated journaling

Rapid recovery via per-pool hot spares

Mirror Space

Parity Space

Mirror Space

Mirror Space

Mirror Space

Parity Space

Mirror Space

Mirror Space

Resilient Spaces

Hot Spare

Industry standard storage

Operational Simplicity

Managed via Storage Management PowerShellEnclosure awarenessFlexible and granular administrative control with self-service agility and tenant isolation

Mirror Space

Parity Space

Simple Space

Workload Admin

Tenant Admin

Company A Pool

Mirror Space

Parity Space

Simple Space

Workload Admin

Tenant Admin

Company B Pool

Hoster Admin

Granular management control

Multi-tenant deployment

Effectively Utilize Hardware Investment

Enables mixed deploymentsUtilize both SSD and HDD media typesControl placement at storage space granularity

Industry standard commodity storage

Diverse spaces for diverse workloads

15K RPM

Home Space

Silver

7200 RPM

Archive Space

Bronze

Solid state

Finance Space

ERP Space

Gold

Standalone Server

Low number of individual servers, each with directly-attached storage

Acceptable periods of planned downtimes for traditional maintenance

Space Space Space

Storage spaces with a diverse set of workloads

Standalone Deployment

Standalone Configuration

Standalone Dell R710 server with dual LSI 9205-8e SAS adapters

Dell MD1200 12-bay JBOD array with 300GB 15K RPM disks

Performance of a Well-tuned Standalone System

1 2 4 8 12 16 20 24 28 320

50010001500200025003000350040004500

Simple Spaces vs. Native DisksRead Throughput

Simple Spaces

Native Disks

Thro

ughput

MB

/s

1MB Sequential Read, 7.2K RPM SAS

1 2 4 8 12 16 20 24 28 320

50010001500200025003000350040004500

Simple Spaces vs. Native DisksWrite Throughput

Simple Spaces

Native Disks

Thro

ughput

MB

/s

1MB Sequential Write, 7.2K RPM SAS

1 2 4 8 12 16 20 24 28 320

1000

2000

3000

4000

5000

6000

7000

Simple Spaces vs. Native DisksRead IOPS

Simple Spaces

Native Disks

IOP

S

4K Random Read, 7.2K RPM SAS

1 2 4 8 12 16 20 24 28 320

1000

2000

3000

4000

5000

6000

Simple Spaces vs. Native DisksWrite IOPS

Simple Spaces

Native Disks

IOP

S

4K Random Write, 7.2K RPM SAS

Clustered Storage Spaces Deployment Element

Building block for scalable, highly available deployments

Easily expandable through additional DeploymentElements

Can move HA workload off a cluster node for maintenance without disruption

Live Migration for running VMsSimple “Move” Operation for others

Space Space SpaceClustered Storage Spaces

\CRMUnified Cluster Shared Volume Namespace

Enterprise Workloads

2-4 servers

Shared SAS storage

Deployment Element

Scalable and Continuous Availability

Physical View(64 servers max)

Space Space Space Space Space Space Space Space Space

Clustered Pools and Storage Spaces

\Finance \CRM \File-SharesUnified Cluster Shared Volume Namespace

Logical View(Highly available, scalable, cost-efficient)

Enterprise Workloads.\db .\VHD

VMs

.\Home

CA File Server

Deployment ElementDeployment Element

2-4 servers

Shared SAS storage

Deployment Element

Representative Deployment Element

60-baySAS array

60-baySAS array

60-baySAS array

60-baySAS array

3TB Near-line SAS HDDs

720TBHigh-Cap

600GB 15K SAS HDDs

144TBMid-Perf

400GB SAS SSDs 96TBHigh-Perf

64 Node ClusterItem Count

Deployment Elements 16

Servers per Element 4

Total number of servers 64

JBODs per Element 1

Total number of disks 242 HDDs

and SSDs

40Gbps networks per server

4

Virtual Machines 4000

Storage Spaces delivers a new category of highly capable storage solutions to all Windows

customer segments at dramatically lower price-points and maximized operational simplicity

Resilient File System (ReFS)

Next-generation scale, availability, and integrity

ReFS is a new file system delivering on demanding customer needs while providing a foundation for significant future innovations.

Windows Server 2012 customers can deploy the most cost-efficient platform for available and scalable data access using commodity storage,

by utilizing an integrated storage stack comprising ReFS and Storage Spaces.

ReFS Overview (1/2)

• All metadata corruption is always detected. Optionally, user data corruption is also detected.

• ReFS utilizes Storage Spaces’ resiliency to auto-correct.

Integrity

• ReFS efficiently scales to PB datasets comprising very large files.

Scalability

ReFS Overview (2/2)

• ReFS maximizes data availability.• If corruption occurs and a redundant copy is

unavailable, the non-corrupt portions remain unaffected and online.

Availability

• Architecture enables efficient evolution for new storage devices and new access patterns.

Architectural Evolution

Integrity

Allocate-on-write eliminates corruption from errors associated with in-place update

Checksums detect whether the data on media has changed since it was last written

File system metadata is always protected, maximizing online operation

Flexible user data protection is configurable on a per-volume, per-directory, and per-file basis

ReFS and Storage Spaces

Data Volume - NTFS

Scrubber

Data Volume - ReFS

Storage Pool

Data

Copy0

Data

Copy1

Data

Copy2

On media failure, NTFS will get an alternate copy of the data

With NTFS, Scrubber verifies all copies can be read

Checksum is inspected on all copies of the dataIf corruption detected, trigger auto-correct and return a healthy copy of the data

With ReFS, Scrubber verifies checksums on all copies and auto-corrects as needed

Copy1

DataCHK

Copy2

DataCHK

Copy0

DataCHK

Mirror Space

Mirror Space

ReFS Auto Correction

Application accesses data

ReFS volume detects corruption and automatically replaces the corrupt data with correct data

Mirror space offers all data copies to the file system

ReFS

Scalability

ReFS uses a highly scalable, flexible, B+ tree data structure for all metadata• Ensures maximum scalability and volume sizes

beyond NTFS

File system abstractions are implemented using key-value tables built from the B+ trees

Capacity allocation is handled by a hierarchy of allocators• Ability to allocate at different sized granularities

Application Compatibility

While delivering on new capabilities, ReFS also builds on NTFS’ 20 years of application compatibility

ReFS largely reuses the upper-layer to provide API compatibility to applications while providing a new on-media engine that enables new capabilities

NTFS.SYS

Upper API Layer

On-media Store

ReFS.SYS

Upper API Layer

MinStore

Application Compatibility

ReFS supports the majority of the Win32 APIs

However, there are certain features that ReFS does not support

Legacy Short Names TxF

NTFS-specific Reparse Index Named Streams MFT

Superseded NTFS

Encryption NTFS Quotas NTFS

Compression

Other Hard Links

          Extended

Attributes Object IDs

ReFS DeploymentsDeployable on cost-efficient commodity hardware

Optional integrity streams guarantee user data integrityStorage Spaces provides resiliency against media failure

Failover Clustering provides high availability for File Server

ConsiderationsOnly available with Windows Server 2012Can only be deployed as a data volumeDoes not guarantee the same level of application compatibility as NTFS

Deployment Element

2-4 servers

Shared SAS storage

Space Space SpaceClustered Storage Spaces

Home Dirs

VHD Librar

y

Mark’t

Dept.

ReFS volumes with integrity streams

\Docs \Home \VHDs Highly Available File Server

VHDs

ReFS is a new file system delivering on demanding customer needs while providing a foundation for significant future innovations.

Windows Server 2012 customers can deploy the most cost-efficient platform for available and scalable data access using commodity storage,

by utilizing an integrated storage stack comprising ReFS and Storage Spaces.

Ease of storage administration

Storage Management PowerShell with SMI-S integration

Minimize Op-Ex

Efficient capacity utilization

Thin provisioning integrationMaximize Existing Hardware Dynamic data

movement Offloaded data transfers

Storage virtualization and resiliency

Storage SpacesMinimize Cap-Ex

Maximize Availability

Next-generation scale, availability, and integrity

Resilient File System (ReFS)

Maximize data availability

Modern NTFS CHKDSK

Storage Platform OverviewBenefit Capability Feature

Windows Server 2012 is the most cost-efficient

platform for continuously available and highly scalable data access

Some Windows Storage Resources

Virtualizing Storage for Scale, Resiliency, and Efficiencyhttp://go.microsoft.com/fwlink/?LinkID=254536

How to Configure Clustered Storage Spaces in Windows Server 2012http://go.microsoft.com/fwlink/?LinkID=254538

Storage Spaces FAQhttp://go.microsoft.com/fwlink/?LinkID=254539

© 2012 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the

part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.