VMworld Europe 2014: Virtual SAN Best Practices and Use Cases

Disclaimer

• This presentation may contain product features that are currently under development.

• This overview of new technology represents no commitment from VMware to deliver these features in any generally available product.

• Features are subject to change, and must not be included in contracts, purchase orders, or sales agreements of any kind.

• Technical feasibility and market demand will affect final delivery.

• Pricing and packaging for any new technologies or features discussed or presented have not been determined.

CONFIDENTIAL 2

3

1 Introduction to VMware Virtual SAN

2 Use Case Overview

3 Use Cases Characteristics & Sizing Considerations

5 Best Practices

6 Q & A – We’d love to hear about your use cases and recommendations

Agenda

Not Covered

Hardware Best Practices:

STO1211 - Virtual SAN Ready Node and Hardware Guidance for

Hypervisor Converged Infrastructure

Architecture Deep Dive:

STO1279 - Virtual SAN Architecture Deep Dive

Troubleshooting:

STO3098 - Virtual SAN Best Practices for Monitoring and

Troubleshooting

The Software-Defined Data Center

Transform storage by aligning it with app demands

Managementtools give wayto automation

Expand virtual compute to all

applications

Virtualize the network for speed

and efficiency

5

The Software-Defined Data Center

Transform storage by aligning it with app demands

6

VMware Software-Defined Storage

Bringing the Efficient Operational Model of Virtualization to Storage

Policy-driven Control Plane

Virtual Data Plane

Virtual Datastores

PerformanceMobilityData Protection

Virtual Data Services

Cloud Object

StorageSAN /

NAS

x86 Servers

7

VMware Virtual SANIntroduction

VMware Virtual SAN Introduction

Hypervisor-Converged Storage Platform

vSphere + Virtual SAN

…

• Software-defined storage software solution.

• Design to aggregates locally attached storage from each ESXi host in a cluster.

• Hybrid disk storage solution

– Magnetic disks (HDD)

– Flash based disks (SSD)

• VM-Centric data operations and policy driven management principles.

• Resilient design based on a Distributed RAID architecture

– No single points of failures

• Dynamic capacity and performance scalability

VMware Virtual SAN Use Case and Best Practices

11

Management Clusters

Use CasesTest / Dev / Staging

Private cloud

ROBO

Virtual SAN

Backup and DR

Target

Site A Site B


…

Production Tier 2 / Tier 3

NewDMZ / Isolated


…

Regulatory Compliance


Dev Test Stage

ProdNew SDLC

Virtual Desktop


…

Horizon

12

Management Clusters


Private cloud

ROBO

Virtual SAN

Backup and DR

Target

Site A Site B


…


DMZ / Isolated


…



Dev Test Stage

ProdSDLC

Virtual Desktop


…

Horizon

Virtual Desktop Infrastructure

Key Requirements

• Handle peak performance requirements:

– boot storms

– login storms

– read/write

• Support High VDI density

• Linear Scalable Capabilities

– Scale Up

– Scale Out

• Cost

• Supportability for all types of desktops

Horizon


Advantages

disk group disk group disk group

VSAN network VSAN networkVSAN network

vsanDatastore

HDD

disk group

HDD HDD HDD

disk group

VSAN network

HDDscale

up

scale out

• Storage for VDI is contained within a single server

• Allows granular scaling – Add one server with SSD & Disk to scale out

• Deployment Simplicity – Eliminates need for extensive storage design and sizing

• Accelerate Deployment - Customers can quickly go from POC to Pilot to Production


Best Practices

1. Desktop Sizing

2. VSAN VM Storage Policy Definition

3. Size the cluster

4. Size the host Horizon

Best Practices

1. Desktop Sizing

• Class of worker and type of pool

– Knowledge worker

• Dedicated full clone

• Dedicated linked clone

– Task worker

• Floating linked clone

• VM instance definition

– OS

– CPU, memory, storage,network

• # of desktops per cluster

• Desired consolidation ratio

2. Virtual SAN Policies

• Failures to Tolerate

– For n failures, n+1 copies and 2n+1 hosts

– FTT has the greatest impact on capacity utilization in the cluster

– Set default FTT value = 1; For linked clone floating pools consider setting FTT=0

• Object space reservation

– Default is 0

– For full clone, consider setting this to 100%, as it guarantees deterministic placement of desktops

• Flash Read Cache reservation

– Typically default is 0

– Consider setting it to 10% for the replica disk in linked-clone floating pools

• Number of disk stripes per host

– Leverage the default value of 1 for optimal performance

Horizon 6.0 automatically sets the recommended values for policies

Horizon View – Default Storage Policies

• Horizon View automatically creates default storage policies based on Pool type

REPLICA_DISK PERSISTENT_DISK OS_DISK VM_HOME

Stripes 1 1 1 1

Resiliency 1 1 1 1

Space Reservation0%

Thin

100%

Thick

0%

Thin

0%

Thin

Read Cache Reservation 10% 0% 0% 0%

3. Sizing the cluster

• Cluster Size

– (# of Desktops/desired consolidation ratio per host) + 1

– Keep head room for host failure

• Datastore capacity

– Usable: Capacity/Desktop * # of Desktop’s per cluster

– Raw: Usable* (FTT+1) * 130%

• Flash Capacity

– 10% or more usable datastore capacity

• Keep VSAN component count below max limit (3000/host)

4. Sizing the Host

• CPU

– Average # of vCPU’s per desktop

– CPU utilization per desktop

– Additional CPU for View – up to 15%

– Additional CPU for VSAN – up to 10%

• Memory

– Memory allocated per desktop

– Video memory overhead - # of monitors and monitor resolution per desktop

– View Accelerator – up to 2GB

– Additional CPU for VSAN – up to 10%

• Network

– Dedicated redundant 10GbE uplinks

Sizing the host (contd)

• Magnetic Disk

– Linked clone: 10K RPM or better

– Full clone: 7.2K or 10K RPM drives

• Disk groups

– Consider multiple disk groups per host – improved performance, smaller failure domains and quicker rebuilds

– Linked clones: 1 disk group

– Full clones: 2 disk groups

Sizing Example

Desktops

Projected Desktops 1000

VM storage 40 GB

Type of desktop Linked Clone

Virtual SAN VM storage policy

Failures to tolerate 1

Flash Read Cache

Reservation

10% (for replica disk)

0% for rest

Object Space

Reservation

0%

Host Sizing

Capacity/host ~10TB

# of disk groups/host 2

HDD/host 10 x 900GB 10KRPM

SSD/host 2 x 200 GB

Cluster Sizing

# of hosts 10+1

# of VM’s/host 91

Total usable capacity 40TB

Total raw capacity 40TB*2*130%=104TB

Total Flash Capacity 4TB

1 2

34


Ready Nodes – Node and Service Profiles

Linked Clones Full Clones High Medium

# of VM per Node Up to 100 Up to 100 Up to 60 Up to 30

IOPS per Node Up to 10K Up to 10K Up to 20K Up to 12K

Raw Capacity Per

Node

1.2TB 10.8TB 14.4 TB 8TB

CPU 2x10 core 2x10 core 2x10 core 2x10 core

Memory 256 GB 256 GB 384 GB 256 GB

HDD 4x300GB SAS

15K RPM

12x900GB SAS

10K

12x12TB SAS

10K RPM

8x1TB NL-SAS 7.2K

RPM

SSD 1X400GB SSD

(Class E)

2x400GB SSD

(Class E)

2x400GB SSD

(Class E)

2x200GB SSD (Class

D)

IO Controller Queue Depth

>=256

Queue Depth

>=256

Queue Depth

>=512

Queue Depth >=256

NIC 10GbE 10GbE 10GbE 10GbE

Ready Node ProfileService Profile

24

Management Clusters


Private cloud

ROBO

Virtual SAN

Backup and DR

Target

Site A Site B


…


DMZ / Isolated


…



Dev Test Stage

ProdSDLC

Virtual Desktop


…

Horizon

Remote Office / Branch Office

• Small number of virtual machines and capacity

– 10 to 15 virtual machines

– Apps include – file and print server, domain controller, small development environment

– Storage capacity less than 5 Terabyte's

– Centralized management

– Limited or no remote IT support on site

– Smallest footprint possible

– High Availability

– Automated failover

– Data protection to datacenter

Virtual SAN

Key Requirements


Host Design

– Typically 3 1-CPU node cluster – provides availability while maintaining costs!

– 1G network is OK

• vCenter deployed at data center or as a Virtual Appliance on Virtual SAN

• High availability and automated failover

• VDP/VR for asynchronous replication

Advantages



vsanDatastore

HDD

disk group

HDD HDD HDD

disk group

VSAN network

HDD


Best PracticesBest

Practices

Virtual SAN

Avoid possible performance and recoverability issues

– Do not exceed the number of recommended virtual machines on a 1G network

VM Storage Policy Definition

– Create a VM Storage Policy with the following settings:

• Number of failures to tolerate 1 (FTT=1)

• Object Disk Stripe-width 1 (ODSW=1)


Low

# of VM per Node Up to 15

IOPS per Node Up to 2K

Raw Capacity Per

Node

5TB

CPU 1x6 CPU cores

Memory 64 GB

HDD 5x1TB NL-SAS 7.2K

RPM

SSD 1x200GB SSD (Class B

or above)

IO Controller Queue Depth >= 256

NIC 1GbE

Ready Nodes

Ready Node Profiles

29

Management Clusters


Private cloud

ROBO

Virtual SAN

Backup and DR

Target

Site A Site B


…


DMZ / Isolated


…



Dev Test Stage

ProdSDLC

Virtual Desktop


…

Horizon

DMZ | Isolated

Key Requirements

• Complete resource access isolation (Air Gap)

– Logical

– Physical

• Regulatory compliance

– PCI

– HIPPA

– Others

• High Availably with multiple levels of redundancy

• RPO - 30 Minutes


…


DMZ | Isolation

Best Practices



vsanDatastore

HDD

disk group

HDD HDD HDD

disk group

VSAN network

HDDscale

up

scale out

• Set up separate VSAN VLANs for your DMZ network

• vCenter can be shared between other VSAN clusters

• Deployment Simplicity

– Follow sizing principles covered elsewhere in this presentation

DMZ | Isolated

Ready Nodes

High Medium

Up to 60 Up to 30

Up to 20K Up to 12K

14.4 TB 8TB

2x10 core 2x10 core

384 GB 256 GB

12.12TB SAS 10K RPM 8x1TB NL-SAS 7.2K RPM

2x400GB SSD

(Class E)

2x200GB SSD

(Class D)

Queue Depth >=512 Queue Depth >=256

Ready Node Profile

Use Cases

34

Management Clusters

Test / Dev / Staging

Private cloud

ROBO

Virtual SAN

Backup and DR

Target

Site A Site B


…


DMZ / Isolated


…



Dev Test Stage

ProdSDLC

Virtual Desktop


…

Horizon

Management ClusterKey Requirements

• Minimum of 3 ESXi Hosts

• Support all necessary infrastructure application and components

• Separate infrastructure application and components from resources utilized by production workloads

• Improve manageability of infrastructure

Management Cluster

Management ClusterAdvantages

• Storage Managed by vSphere Admin

• Eliminate resource contention between production workloads and infrastructure components

• Support infrastructure application

• Flexible Scalable Capabilities

– Scale Up

– Scale Out

• Optimize performance of resource intensive management application

– vCenter Server

– vCenter Operations

– Etc.



vsanDatastore

HDD

disk group

HDD HDD HDD

disk group

VSAN network

HDDscale

up

scale out

Management ClusterBest Practices

Management Cluster

Best Practices

• Desktop Sizing

– Make sure the # of VM’s/host is below the supported limit of 100

• VM Storage Policy

– Consider the use of object space reservation to sustain performance during data management operations

– Flash Read Cache reservation: Typically set to 0; one can consider reserving capacity for the base image

• Host and Disk Groups

– Consider the use of 4 nodes

– Consider configuring multiple disks groups

Management ClusterReady Nodes

High Medium

Up to 60 Up to 30

Up to 20K Up to 12K

14.4 TB 8TB

2x10 core 2x10 core

384 GB 256 GB

12.12TB SAS 10K RPM 8x1TB NL-SAS 7.2K

RPM

2x400GB SSD

(Class E)

2x200GB SSD

(Class D)


Ready Node Profiles

Use Cases

39

Management Clusters


Private cloud

ROBO

Virtual SAN

Backup and DR

Target

Site A Site B


…


DMZ / Isolated


…



Dev Test Stage

ProdSDLC

Virtual Desktop


…

Horizon

Backup and DR TargetKey Requirements

Self-service development environment with:• Application lifecycle management• Separate and isolated development

execution zones– Compute– Storage– Network

• Accelerated workload deployments• Performing, scalable and secure

dedicated infrastructure• Policy based governance with automated

delivery• Fully automated Policy migration across

execution zones• Interoperable with vSphere and third party

complementary solutions

Backup and DR Target

Site A Site B

Backup and DR TargetAdvantages

Flexible DR capabilities:• Support Virtual SAN as a target destination from any

array• VM Centric based protection• Guaranteed storage requirements characteristics

through VM Storage Policy– Availability– Performance– Capacity

• Integrate with vCenter Site Recovery Manager for extensive capabilities:– Automated DR operation & orchestration– Automated failover – execution of user defined plans– Automated failback – reverse original recovery plan– Planned migrations – ensure zero data loss– Point-in-time-recovery – multiple recovery points– Non-disruptive test – automate test on isolated

infrastructures


Site A Site B

Backup and DR TargetBest Practices

Best Practices

Keep track of component counts when using VR point in time recovery.

– Plan cluster size and component count appropriately based on recovery configuration and requirements

– Maintain consistent Virtual SAN VM storage policies across sites when target and destination are both Virtual SAN


Site A Site B

Backup and DR TargetReady Nodes

High Medium

Up to 60 Up to 30

Up to 20K Up to 12K

14.4 TB 8TB

2x10 core 2x10 core

384 GB 256 GB


RPM

2x400GB SSD

(Class E)

2x200GB SSD

(Class D)


Ready Node Profiles

Use Cases

44

Management Clusters


Private cloud

ROBO

Virtual SAN

Backup and DR

Target

Site A Site B


…


DMZ / Isolated


…



Dev Test Stage

ProdSDLC

Virtual Desktop


…

Horizon

Software Development LifecycleKey Requirements

Dev Test

Prod

Stage

Self-service development environment with:

• Application lifecycle management

• Separate and isolated development execution zones

– Compute

– Storage

– Network

• Accelerated workload deployments

• Performing, scalable and secure dedicated infrastructure

• Policy based governance with automated delivery

• Fully automated Policy migration across execution zones

• Interoperable with vSphere and third party complementary solutions

Software Development LifecycleAdvantages

Dev Test

Prod

Stage

Flexible service offerings:

• High Availability (tolerate failures) and Performance for demanding workloads

• Fully automated Policy migration across execution zones

• Ability to manage service offerings and automated migration capabilities on a per VM basis

• Better linear scalability options

– Scale capacity and IOPS together

– Avoid Storage IO hot spots

• Accelerated workload deployments

• Policy based governance with automated delivery

• Interoperable with vSphere and third party complementary solutions

Software Development LifecycleBest Practices

Dev Test

Prod

Stage

Best Practices

Leverage the vCloud Automation Center Storage Policy-Based Management integration

• Define default policies per Virtual SAN cluster and execution zones

• Size Virtual SAN cluster capacity and acceleration layers suitable to the different requirements per execution zones

Implement approval procedures for:

• Virtual machine allocation

• Migration across execution zones

Software Development LifecycleReady Nodes

High Medium

Up to 60 Up to 30

Up to 20K Up to 12K

14.4 TB 8TB

2x10 core 2x10 core

384 GB 256 GB


RPM

2x400GB SSD

(Class E)

2x200GB SSD

(Class D)


Ready Node Profiles

Summary

Virtual SAN is suitable but not limited to the presented use cases

Always follow the recommended guidance for best results and supportability

We’re here to help you size your environment!

VMworld Europe 2014: Virtual SAN Best Practices and Use Cases

Technology

Transcript of VMworld Europe 2014: Virtual SAN Best Practices and Use Cases