black w 1650 red hat storage server administration deep dive1

Red Hat Storage Server Administration Deep DiveDustin L. Black, RHCASr. Technical Account ManagerRed Hat Global Support Services

** This session will include a live demo from 6-7pm **

Dustin L. Black, RHCA

Sr. Technical Account Manager

Red Hat, Inc.

[email protected]

@dustinlblack

WTH is a TAM?

Premium named-resource supportProactive and early accessRegular calls and on-site engagementsCustomer advocate within Red Hat and upstreamMulti-vendor support coordinatorHigh-touch access to engineering Influence for software enhancementsNOT Hands-on or consulting

Agenda

Technology Overview & Use CasesTechnology StackUnder the HoodVolumes and Layered FunctionalityAsynchronous ReplicationData AccessSWAG IntermissionDemo Time!

Red Hat Storage Server Administration Deep Dive

Technology Overview

What is GlusterFS?

Clustered Scale-out General Purpose Storage PlatformPOSIX-y Distributed File System ...and so much more

Built on Commodity systemsx86_64 Linux ++POSIX filesystems underneath (XFS, EXT4)

No Metadata ServerStandards-Based Clients, Applications, NetworksModular Architecture for Scale and Functionality

Fuse Native

Network Interconnect

NFS

Samba

libgfapi

app

QEMU

What is Red Hat Storage?

Enterprise Implementation of GlusterFS Integrated Software ApplianceRHEL + XFS + GlusterFS

Certified Hardware CompatibilitySubscription Model24x7 Premium Support

release

feedback

iterate

debug

enhance

upstream

release

downstreamglusterfs-3.4

feedback

GlusterFS vs. Traditional Solutions

A basic NAS has limited scalability and redundancyOther distributed filesystems are limited by metadata serviceSAN is costly & complicated, but high performance & scalableGlusterFS is...Linear ScalingMinimal OverheadHigh RedundancySimple and Inexpensive Deployment


Use Cases

Common Solutions

Large Scale File ServerMedia / Content Distribution Network (CDN)Backup / Archive / Disaster Recovery (DR)High Performance Computing (HPC) Infrastructure as a Service (IaaS) storage layerDatabase offload (blobs)Unified Object Store + File Access

Hadoop Map Reduce

Access data within and outside of HadoopNo HDFS name node single point of failure / bottleneckSeamless replacement for HDFSScales with the massive growth of big data

Account Container Object

Volume Directory Subdir/File

Swift Cinder / Glance

libgfapi-QEMU


Technology Stack

Terminology

BrickFundamentally, a filesystem mountpointA unit of storage used as a capacity building block

TranslatorLogic between the file bits and the Global NamespaceLayered to provide GlusterFS functionality

Everything is Modular

Terminology

VolumeBricks combined and passed through translatorsUltimately, what's presented to the end user

Peer / NodeServer hosting the brick filesystemsRuns the gluster daemons and participates in volumes

Disk, LVM, and Filesystems

Direct-Attached Storage (DAS)

-or-Just a Bunch Of Disks (JBOD)Hardware RAIDRHS: RAID 6 required

Logical Volume Management (LVM)POSIX filesystem w/ Extended Attributes (EXT4, XFS, BTRFS, ...)RHS: XFS required

Data Access Overview

GlusterFS Native ClientFilesystem in Userspace (FUSE)

NFSBuilt-in Service

SMB/CIFSSamba server required; NOW libgfapi-integrated!

Data Access Overview

Gluster For OpenStack (G4O; aka UFO)Simultaneous object-based access via OpenStack Swift

NEW! libgfapi flexible abstracted storage Integrated with upstream Samba and Ganesha-NFS

Gluster Components

glusterdManagement daemonOne instance on each GlusterFS server Interfaced through gluster CLI

glusterfsdGlusterFS brick daemonOne process for each brick on each serverManaged by glusterd

Gluster Components

glusterfsVolume service daemonOne process for each volume serviceNFS server, FUSE client, Self-Heal, Quota, ...

mount.glusterfsFUSE native client mount extension

glusterGluster Console Manager (CLI)

services to the public network

Har

d D

isk

0

Har

d D

isk

1

Har

d D

isk

2

Har

d D

isk

3

Har

d D

isk

4

Har

d D

isk

5

Har

d D

isk

6

Har

d D

isk

7

Har

d D

isk

8

Har

d D

isk

9

Har

d D

isk

10

Har

d D

isk

11

Disk StorageLocal JBOD* (SAS, SATA), DAS*

Limited Fibre Channel and iSCSI Support

Hardware RAIDRAID 6*, 1+0

Volume ManagerLVM2*

Local File SystemsXFS*, ext3/4, BTRFS, ...

A 32- or 64-bit* Linux DistributionRHEL*, CentOS, Fedora, Debian, Ubuntu, ...

Storage Network1Gb, 10Gb, Infiniband

Public Network1Gb, 10Gb, Infiniband

GlusterFS Server (glusterd daemon)RPM* and DEB packages, or from source

glusterfsd brick daemons

gfsd gfsd gfsd gfsd gfsd gfsd gfsd gfsd gfsd gfsd

glusterfsclient

HD

FS*

Sw

ift*

Server-sidereplication

Putti

ng it

Tog

ethe

r libgfapi

libvi

rt*

Cin

der

SM

B*

AP

I

glusterfs client*

FTP

*

required

stronglyrecommended

optional

a GlusterFSservice

*Red Hat Storage Supported

NF

S-G

an.

...

glusterfsNFS*

Tran

slat

ors

Up and Out!


Under the Hood

Elastic Hash Algorithm

No central metadataNo Performance BottleneckEliminates risk scenarios

Location hashed intelligently on filenameUnique identifiers, similar to md5sum

The Elastic PartFiles assigned to virtual volumesVirtual volumes assigned to multiple bricksVolumes easily reassigned on the fly

Translators

Your Storage Servers are Sacred!

Don't touch the brick filesystems directly!They're Linux servers, but treat them like storage appliancesSeparate security protocolsSeparate access standards

Don't let your Jr. Linux admins in!A well-meaning sysadmin can quickly break your system or destroy your data


Basic Volumes

Distributed Volume

The default configurationFiles evenly spread across bricksSimilar to file-level RAID 0Server/Disk failure could be catastrophic

Replicated Volume

Files written synchronously to replica peersFiles read synchronously, but ultimately serviced by the first responderSimilar to file-level RAID 1

Striped Volumes

Individual files split among bricks (sparse files)Similar to block-level RAID 0Limited Use CasesHPC Pre/Post ProcessingFile size exceeds brick size


Layered Functionality

Distributed Replicated Volume

Distributes files across multiple replica sets

Distributed Striped Volume

Distributes files across multiple stripe setsStriping plus scalability

Striped Replicated Volume

Replicated sets of stripe setsSimilar to RAID 10 (1+0)

Distributed Striped Replicated Volume

Limited Use Cases Map Reduce

Don't do it like this -->


Asynchronous Replication

Geo Replication

Asynchronous across LAN, WAN, or InternetMaster-Slave modelCascading possible

Continuous and incrementalOne Way

NEW! Distributed Geo-Replication

Drastic performance improvementsParallel transfersEfficient source scanningPipelined and batchedFile type/layout agnostic

Available now in RHS 2.1Planned for GlusterFS 3.5

Distributed Geo-Replication

Drastic performance improvementsParallel transfersEfficient source scanningPipelined and batchedFile type/layout agnostic

Perhaps it's not just for DR anymore...

http://www.redhat.com/resourcelibrary/case-studies/intuit-leverages-red-hat-storage-for-always-available-massively-scalable-storage


Data Access

GlusterFS Native Client (FUSE)

FUSE kernel module allows the filesystem to be built and operated entirely in userspaceSpecify mount to any GlusterFS serverNative Client fetches volfile from mount server, then communicates directly with all nodes to access data Recommended for high concurrency and high write performanceLoad is inherently balanced across distributed volumes

NFS

Standard NFS v3 clientsStandard automounter is supportedMount to any server, or use a load balancerGlusterFS NFS server includes Network Lock Manager (NLM) to synchronize locks across clientsBetter performance for reading many small files from a single clientLoad balancing must be managed externally

NEW! libgfapi

Introduced with GlusterFS 3.4User-space library for accessing data in GlusterFSFilesystem-like APIRuns in application processno FUSE, no copies, no context switches ...but same volfiles, translators, etc.

SMB/CIFS

NEW! In GlusterFS 3.4 Samba + libgfapiNo need for local native client mount & re-exportSignificant performance improvements with FUSE removed from the equation

Must be setup on each server you wish to connect to via CIFSCTDB is required for Samba clustering

HDFS Compatibility

Gluster 4 OpenStack (G4O)

The feature formerly known as UFO


SWAG Intermission


Demo Time!


Do it!

Do it!

Build a test environment in VMs in just minutes!Get the bits:Fedora 20 has GlusterFS packages natively: fedoraproject.orgRHS 2.1 ISO available on the Red Hat Portal: access.redhat.comGo upstream: gluster.orgAmazon Web Services (AWS)Amazon Linux AMI includes GlusterFS packagesRHS AMI is available

Check Out Other Red Hat Storage Activities at The Summit

Enter the raffle to win tickets for a $500 gift card or trip to LegoLand! Entry cards available in all storage sessions - the more you attend, the more chances you

have to win!

Talk to Storage Experts:Red Hat Booth (# 211)

Infrastructure Infrastructure-as-a-Service

Storage Partner Solutions Booth (# 605)

Upstream Gluster projects Developer Lounge

Follow us on Twitter, Facebook: @RedHatStorage

Thank You!

Red Hat Storage Server Administration Deep DiveSlides Available at: people.redhat.com/dblack

Contact [email protected] [email protected]

Resourceswww.gluster.orgwww.redhat.com/storage/ access.redhat.com/support/offerings/tam/

Twitter @dustinlblack @gluster @RedHatStorage

** Please Leave Your Feedback in the Summit Mobile App Session Survey **

Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28Slide 29Slide 30Slide 31Slide 32Slide 33Slide 34Slide 35Slide 36Slide 37Slide 38Slide 39Slide 40Slide 41Slide 42Slide 43Slide 44Slide 45Slide 46Slide 47Slide 48Slide 49Slide 50Slide 51Slide 52Slide 53Slide 54Slide 55

black w 1650 red hat storage server administration deep dive1

Documents

storage das

servicesmbcifssamba

file bits

clustered scale

metadata servicesan

global namespacelayered

linux posix filesystems

engineering influence