APP-BCA2930

Virtualizing SharePoint Best Practices

Scott Salyer, VMware, Inc.

#vmworldapps

Disclaimer

This session may contain product features that are

currently under development.

This session/overview of the 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.

Agenda
Introduction and Benefits SharePoint on vSphere Performance SharePoint on vSphere Capacity Planning
Workload Modeling and Architectural Design SQL Server Capacity and Performance Deploying to ESX/ESXi

SharePoint on vSphere Availability and Recovery

High Availability Disaster Recovery Backup and Recovery

More Information
3

SharePoint What is it?

Mostly an intranet portal for collaboration It can be customized for all sorts of uses Varies from insignificant to critical

http://www.topsharepoint.com
4

Key Benefits Virtualizing SharePoint

Consolidation

Achieve 2-10x consolidation ratio, especially for larger

deployments

Lower TCO

Significant savings in power, cooling, and datacenter

space

Availability

VM-based protection for SharePoint provides

homogeneous high availability (VMware HA)

Business Continuity

Simplified DR management with vCenter Site Recovery

Manager

Maintenance

vMotion of SharePoint virtual machines Maximized overall performance with balanced HW

utilization across the farm (VMware DRS)

Load Balancing

Rapid Provisioning And Scaling

5

VM templates for fast provisioning and easier scale-out

Virtualizing Server Roles In SharePoint

Server Roles/Priority
1st

What To Consider
CPU User concurrency, Search requests Scaling out is more efficient Network segment vNICs and vSwitches

Web Front End / Query

2nd

Application (Excel, Doc Conv, etc)

CPU Application dependent Scaling out is more efficient

3rd

Redundant (Non redundant in MOSS 2007)

Index/Crawl

CPU Crawling, indexing (depends on content type/size) Scale out (Up only with MOSS 2007) Memory intensive CPU Document updates, Search, Backup

4th

SQL

VMFS/RDM Scale up/out Failover Clustering, Mirroring, VMware HA

Understanding your existing workload is better than any best practice!!!

6

Agenda
Introduction and Benefits SharePoint on vSphere Performance SharePoint on vSphere Capacity Planning
Workload Modeling and Architectural Design SQL Server Capacity and Performance Deploying to ESX/ESXi

SharePoint on vSphere Availability and Recovery

High Availability Disaster Recovery Backup and Recovery

More Information
7

Maximum Scalability and Performance With vSphere 5

Applications Performance Requirements

95% of Apps Require ESX 1 1 VCPUs 2 GB per VM <.5Gb/s <5,000 ESX 2 2 VCPUs 3.6 GB per VM .9 Gb/s 7,000 VMware Inf. 3.0/3.5 4 VCPUs 16/64 GB per VM 9 Gb/s 100,000 VMware vSphere 4 8 VCPUs 256 GB per VM 30 Gb/s 300,000 VMware vSphere 5 32 VCPUs 1,000 GB per VM >36Gb/s 1,000,000

% of Applications

CPU

1 to 2 CPUs

Memory < 4 GB at peak Network IOPS <2.4 Mb/s < 10,000

Maximum Scalability and Performance With vSphere 5

Applications Performance Requirements

95% of Apps Require ESX 1 1 VCPUs 2 GB per VM <.5Gb/s <5,000 VMware ESX 2 vSphere 5 2 32 VCPUs VCPUs VMware Inf. 3.0/3.5 4 VCPUs VMware vSphere 4 8 VCPUs 256 GB per VM 30 Gb/s 300,000

% of Applications

CPU

1 to 2 CPUs

Memory < 4 GB at peak Network IOPS <2.4 Mb/s < 10,000

1,000 3.6 GB GB per per VM VM 16/64 GB per VM .9 >36Gb/s Gb/s 1,000,000 7,000 9 Gb/s 100,000

SharePoint performance - The user experience

Storage
Content/Metadata Search System

Server
CPU Memory HBA/CNA NIC

Network

Client

BLOB Storage (Optional)

10

SharePoint performance - The user experience

Domain Controller
Authentication

SQL Server

Storage
Content/Metadata Search System

Web Front End

Document Request
Acceptable user response time <3 seconds <5 seconds <7 seconds

Type Of operation Common CPU

Server

Examples

Network

Client

BLOB Storage (Optional)

BLOB Retrieval/Creation

Memory Uncommon HBA/CNA NIC Rare

Browsing to the home page Browsing to a document library Creating a subsite Creating a list Uploading a document to a document library Backing up a site Creating a site collection

http://technet.microsoft.com/en-us/library/cc287790(office.12).aspx

11

SharePoint 2010 Performance Test Logical Architecture

Workload
Root portal configured with collaboration template 260GB content, approximately 600K items in 10 site collections Incremental crawl every four hours and weekly full crawl

VSTS load generator Zero think time (per Microsoft guidelines) Transaction mix 80-10-10 read-write-search Real world settings IIS logging on, SharePoint caching disabled

12

Physical versus Virtual Study

Physical versus virtual Web front end (WFE) comparison shows the
overall request per second (RPS) differs very little, even at higher CPU saturation levels
Physical Versus Virtual WFE CPU Comparison
50 45 40 35 RPS 30 25 20 15 10 5 0 Physical Virtual
Physical Virtual

Physical Virtual

1 CPU (95%+ Saturation)

2 CPU (75-90% Saturation)

13

Scaling Out the SQL Server Back-End

60-20-20 Mix Test compares one SQL instance versus two SQL instances Scaling out SQL Server provides better throughput! Test with your workload for best SQL server scale out throughput

14

Performance Monitoring
vSphere Client: GUI interface, primary tool for observing one or
more ESX/ESXi hosts

Does not require high levels of privilege

Resxtop/Esxtop Gives access to detailed performance data of a

single ESX/ESXi host

Provides fast access to a large number of

performance metrics

Requires root-level access Runs in interactive, batch, or replay mode

In-guest Monitoring tools

SQL Server: Perfmon, Profiler, Dynamic Manage Views Use ESX Counters in PerfMon for more accurate results http://vpivot.com/2009/09/17/using-perfmon-for-accurate-esx-performance-counters/
15

Agenda
Introduction and Benefits SharePoint on vSphere Performance SharePoint on vSphere Capacity Planning
Workload Modeling and Architectural Design SQL Server Capacity and Performance Deploying to ESX/ESXi

SharePoint on vSphere Availability and Recovery

High Availability Disaster Recovery Backup and Recovery

More Information
16

Capacity Planning Process Summary

Estimate User Activity

Select a starting point architecture

Map out resource requirements by server role (virtual machine requirements)

Plan the ESX/ESXi host hardware configuration (ESX/ESXi host requirements) Perform initial placement exercise to verify resource allocations and failover headroom

17

Estimating User Activity

Upgrading from SharePoint 2007

Mine IIS logs and utilize Microsoft or 3rd party testing tools
SharePoint 2010 Load Testing Kit
http://technet.microsoft.com/en-us/library/ff823736.aspx.

Visual Studio 2008 Team System

http://www.microsoft.com/downloads/en/details.aspx?FamilyID=d95598d7-aa6e-4f2482e3-81570c5384cb&DisplayLang=en.

Visual Studio 2008 Service Pack 1

http://www.microsoft.com/download/en/details.aspx?displaylang=en&id=10986

New installation
Requests per second Concurrent users Total daily users Total daily requests

Enterprise Intranet Collaboration Environment Technical Case Study Example http://technet.microsoft.com/en-us/library/ff758650.aspx

Workload Characteristics Average daily RPS Average RPS at peak time Total number of unique users per day Average daily concurrent users Peak concurrent users at peak time Total number of requests per day

Value 157 350 69,702 420 1,433 18,866,527

18

Workload Modifiers

Workload distribution
Understand the distribution of the requests based on the client applications
that are interacting with the server farm

Newer clients, such as Office 2010, offer new capabilities that can increase the
overall load on the system http://technet.microsoft.com/en-us/library/ff758645.aspx

Concurrency/peak usage
Active Users Start by considering the actual number of SharePoint users Concurrent users Out of the active users, consider how many users are
accessing the system simultaneously at any given time

Peak Usage Period SharePoint concurrent usage might vary significantly

over the course of any given 24-hour period. The peak usage period is the point in time where the maximum number of concurrent users are accessing the SharePoint environment

19

Selecting a Starting Point Architecture

SharePoint 2010 topologies

Published Microsoft topologies from
small to large enterprise farms
SharePoint 2010 Medium Topology Example

Use recommended role requirements

to plan resource allocation for VMs

http://technet.microsoft.com
/en-us/library/cc263044.aspx

SharePoint Server 2010 technical

case studies Published Microsoft technical case
studies illustrating existing production environments

Select the case study that is most applicable to your organizations expected
usage patterns http://technet.microsoft.com/en-us/library/cc261716.aspx.

20

SharePoint Farm Topologies

Small
Web
H/W RAM CPU MOSS 2007 >2GB >3.0GHz Dual SP 2010 8GB >2.5GHz Quad

Medium

Large
Web Servers Groups

Web/Query

Web

User requests

Crawling/Admin

Application
H/W RAM CPU MOSS 2007 4GB >2.5GHz Dual SP 2010 8GB >2.5GHz Quad

App Servers Groups

App

Query/Crawl

App

Query

Crawl

Central Admin /Office/Other

Database
H/W RAM CPU MOSS 2007 >2GB >2.0GHz SP 2010 8 - 64GB >2.5GHz Quad

All DBs

Search DBs

SharePoint DBs

Search DBs

SharePoint DBs

Content DBs

21

SharePoint Farm Topologies

Scale out approach = More servers Small Medium Large ?
Web
H/W RAM CPU MOSS 2007 >2GB >3.0GHz Dual SP 2010 8GB >2.5GHz Quad

Web Servers Groups

Features that impact SQL Server Sizing The size of content databases

Web/Query

Web

User requests

Crawling/Admin

Application
H/W RAM CPU MOSS 2007 4GB >2.5GHz Dual

App Servers Groups

SP The addition of service applications or features into the 2010 environment The use of SQL Server mirroring >2.5GHz
Quad 8GB

App

Query/Crawl

App

Query

Crawl

Central Admin /Office/Other

Database
H/W RAM CPU MOSS 2007 >2GB >2.0GHz

The frequent use of files larger than 15MB (think about using BLOB)

SP 2010 8 - 64GB >2.5GHz Quad

All DBs

Search DBs

SharePoint DBs

Search DBs

SharePoint DBs

Content DBs

22

SQL Server Capacity and Performance

23

A Day in the life of SharePointSQL Server CPU

The majority of load comes from systematic operations

24

A Day in the life of SharePointSQL Server Storage I/O

Plan for user load peaks, not systematic peaks

25

Database Sizing
Central Administration 2GB capacity; minimal disk throughput required Configuration Database 2GB capacity; minimal disk throughput required Can slowly grow beyond 1GB; approx. 40MB for every 50K site
collections

Transaction logs can be large; change recovery model from full

to simple unless mirroring

Content Databases
Database size = ((D V) S) + (10KB (L + (V D)))

D = the number of documents you expect to host S = the average size of each document L = the number of list items in the environment; start with 3 X D
and adjust

V = the approximate number of document versions

26

Content Database Sizing and Performance

SharePoint features that impact content database size

Recycle bin contents http://technet.microsoft.com/enus/library/cc263125.aspx

Auditing data http://technet.microsoft.com/en-us/library/cc298801.aspx Office Web Apps http://technet.microsoft.com/en-us/library/ee837422.aspx

Content database size guidelines

Microsoft recommends that you limit the size of content databases to 200GB A site collection should not exceed 100GB unless it is the only site collection in
the database

Content database sizes up to 1TB are supported only for large, single-site
repositories and archives in which data remains reasonably static

Disk throughput requirements for content databases

Disk throughput requirements can vary significantly between implementations.
Microsoft recommends that you match your expected workload to one of their tested solutions at http://technet.microsoft.com/en-us/library/ff608068.aspx
27

SQL Server Storage Best Practices

Plan for performance in addition to capacity Search database and temp database are the most demanding for
disk I/O. Search database is write intensive when crawling.

When possible, place SQL transaction log and database files on

physically separate disk pools

Place transaction log files on RAID1/0 volumes/pools for high write

performance and faster rebuilds

Most of SharePoint data (content databases) can use

RAID5 volumes RAID5 for more read intensive workloads (common, mainly publishing farms) RAID1/0 for higher random write workloads (heavy collaboration, tempdb,
search)

RAID 6 usually for higher availability with large amount of drives

(Virtual Pools)

28

Additional Considerations for SQL Server

Network topology requirements

Be sure to establish LAN-class bandwidth and latency between Web servers,
application servers, and the SQL Servers. Network latency should be <= 1ms

Understand WAN bandwidth requirements if you are planning to use SQL

Server mirroring or log shipping to keep a remote site up-to-date

Microsoft recommends that Web and application servers have two network
adapters (or vices), one for end-user traffic and the other for communication with the servers running SQL Server. Placing Web or application servers on the same ESX/ESXi host as the SQL Server can improve performance by using internal virtual switches

Guidelines for scaling out SQL Server

Add an additional database server when you have more than four Web servers
that are running at full capacity

Add an additional database server when your content databases exceed 5TB Add an additional database server to promote secure credential storage when
you are running the Secure Store service application
29

Deploying to ESX/ESXi

30

Virtual Machine Resource Allocation

Virtual CPUs
Allocate the minimum requirement and adjust as needed; use HotAdd. If overcommitting processors, monitor %RDY, %MLMTD, and %CSTP Keep NUMA node size in mind with sizing virtual machines

Virtual Memory
Right-size memory allocations for efficient use of host memory Use vSphere 4.1 or above to take advantage of memory compression If overcommitting memory, monitor SWAP /MB: r/s, w/s and MCTLSZ

Storage
Understand I/O requirements for each application tier to avoid performance
degradation due to under-provisioned storage

Use redundant paths to storage Dual host-bus adapters or teamed network

interface cards connected to separate switching infrastructures

Avoid partition misalignment by creating VMFS partitions from within the

vSphere client If creating VMFS from the CLI use fdisk to align
31

Sample Architecture on vSphere

Based on Microsofts departmental collaboration environment

technical case study (SharePoint Server 2010) at http://technet.microsoft.com/en-us/library/ff758649.aspx
Departmental Collaboration Environment Technical Case Study on vSphere Sample Architecture

32

Improving Consolidation Ratios

According to the case study, actual utilization of memory and

processor are fairly low, offering the opportunity to improve consolidation ratios by making adjustments to the environment

If this were a real production environment on vSphere, you could

Reduce the number of vCPUs allocated to the application Servers, freeing up
more cores and allowing increased virtual machine density on each host

Utilize vCPU shares to throttle the amount of actual processor each virtual
machine is consuming without having to change its configuration

Departmental Collaboration Case Study Processor Utilization

33

Agenda
Introduction and Benefits SharePoint on vSphere Performance SharePoint on vSphere Capacity Planning
Workload Modeling and Architectural Design SQL Server Capacity and Performance Deploying to ESX/ESXi

SharePoint on vSphere Availability and Recovery

High Availability Disaster Recovery Backup and Recovery

More Information
34

SharePoint 2010 Availability

What to protect? (Service Level Agreements)

Recovery Point Objectives (RPO) Recovery Time Objectives (RTO) Recovery Level Objectives (RLO)

How to protect?
Tools and technologies available from SharePoint 2010 natively VMware vSphere additions

35

High Availability

36

SharePoint 2010 High Availability

Web and application server role

Software or hardware load balancing SharePoint 2010 search
Multiple crawl servers/databases Multiple query components with mirror
index partitions

Database server role

Synchronous database mirroring or
AlwaysOn

Failover clustering

Use VMware HA, vMotion, DRS

to increase host availability

Use DRS affinity/anti-affinity

rules to spread server roles across hosts
37

SharePoint 2010 with Application-Aware HA

Protects all SharePoint server roles from hardware and application

failure

Does not require complex cluster setup or standby resources Fully integrated with VMware HA and vCenter

38

VMware HA and SQL Server Database Mirroring / AlwaysOn

SharePoint 2010 is mirroring-aware Provides redundancy for

SharePoint 2010 databases

Protection against HW/SW failures

and DB corruption

Storage flexibility (FC, iSCSI, NFS) RTO in few seconds VMware HA + Database Mirroring
Seamless integration, virtual machines
rejoin mirroring session after VMware HA recovery

Can shorten time that database is in

unprotected state

Reduces synchronization time after

virtual machine recovery
39

VMware HA and Failover Clustering

Supports two-node cluster Failover cluster nodes can be

physical or virtual or any combination of the two

Host attach (FC) or in-guest (iSCSI) Supports RDM only VMware HA + failover clustering
Seamless integration, virtual machines
rejoin clustering session after VMware HA recovery

Can shorten time that database is in

unprotected state
Failover clustering now supported with VMware HA with vSphere v4.1 and above
http://kb.vmware.com/selfservice/microsites/search.do?language=en _US&cmd=displayKC&externalId=1037959
40

Disaster Recovery

41

Disaster Recovery with Site Recovery Manager (SRM)

Relies on storage or vSphere Replication Allows creation, maintenance, and execution of automated process to facilitate
site recovery

Safe testing without impacting production environment Improves hardware utilization with co-located test/dev with DR Self-documenting

42

Disaster Recovery with Site Recovery Manager (SRM)

Relies on storage or vSphere Replication Allows creation, maintenance, and execution of automated process to facilitate
site recovery

Safe testing without impacting production environment Improves hardware utilization with co-located test/dev with DR Self-documenting

43

Backup and Recovery

44

What to Backup

SharePoint 2010 Server Farm

Servers
Front End, Application, Index, Search, SQL

SQL Server Databases

Web Applications

Site Collections

Sites

Configuration, Search, Services, and so on

Content Databases
Lists (document libraries, events, contacts, and the like)

Documents and Items

45

VMware Data Recovery (VDR)

Quick, simple, and complete data protection for your SharePoint VMs
with VDR, a disk-based backup and recovery solution

Integrated with vCenter to enable centralized and efficient

management of backup jobs

Useful for small environments Can be used for SQL Server if the service is STOPPED

46

SharePoint 2010 Backup using EMC Replication Manager

47

Summary

vSphere provides the foundation for high performance SharePoint

environments

Virtualized SharePoint instances perform very well compared to

equally sized physical instances

Tests of both Web front-end and SQL virtual machines show

scaling out can provide increased throughput

Monitoring virtualized SharePoint remains the same as a physical

deployment with additional visibility into the underlying infrastructure

Use VMware HA to protect SharePoint from downtime; for higher

availability, consider: Symantec Application HA for more granular control at the service level Combining VMware HA with SQL Server Mirroring

Use SRM for site recovery; co-locate test/dev and recovery VMs
48

Agenda
Introduction and Benefits SharePoint on vSphere Performance SharePoint on vSphere Capacity Planning
Workload Modeling and Architectural Design SQL Server Capacity and Performance Deploying to ESX/ESXi

SharePoint on vSphere Availability and Recovery

High Availability Disaster Recovery Backup and Recovery

More Information
49

Resources

Visit us on the Web to learn more about specific apps

http://www.vmware.com/solutions/business-critical-apps/

Visit the BCA Blog site at: http://blogs.vmware.com/apps/

Includes best practices, design guidance, and success stories for:

Microsoft Apps
Exchange SQL Server SharePoint

Oracle SAP Java Applications

50

Questions

51

Learn more about VMware for SMBs

vmware.com/go/smb

blogs.vmware.com/smb
bit.ly/VMworld_SMB

VMwareCloudContest.com

52

FILL OUT A SURVEY

EVERY COMPLETE SURVEY IS ENTERED INTO DRAWING FOR A $25 VMWARE COMPANY STORE GIFT CERTIFICATE

APP-BCA2930

Virtualizing SharePoint Best Practices

Scott Salyer, VMware, Inc.

#vmworldapps

Backup Slides

55

vSphere Performance Enhancements

Feature Description
VMware ESX/VMware ESXi attempts to keep a virtual machine assigned to its home NUMA-node. Because memory for the virtual machine is allocated from the home node memory access is local and provides the best performance possible

NUMA Support

Transparent Page Sharing

Page sharing allows the hypervisor to reclaim the redundant copies of memory created when multiple virtual machines run the same operating system and applications The balloon driver allows the hypervisor to reclaim host physical memory if memory resources are under contention. This is done with little to no impact to the performance of the application Pages elected to be swapped that can be compressed are stored in a compression cache in main memory. When required, pages are decompressed from compression cache versus paging out from disk Applications that can benefit from large pages on native systems, such as MS SQL, can achieve similar performance improvement on a virtual machine backed with large memory pages High-performance virtual I/O adapters that can provide greater throughput while requiring lower CPU utilization As resource utilization fluctuates within a VMware vSphere cluster, workloads are migrated with no impact to performance or uptime using VMware vSphere vMotion

Memory Ballooning

Memory Compression

Large Memory Page Support Para-virtualized Network and Storage Controllers Distributed Resource Scheduler (DRS and vMotion

56

Estimating User Activity (cont.)

Enterprise Intranet Collaboration Environment Technical Case Study Example
Workload Characteristics Average daily RPS Average RPS at peak time Total number of unique users per day Average daily concurrent users Peak concurrent users at peak time Total number of requests per day Value 157 350 69,702 420 1,433 18,866,527

SharePoint 2007 User Loads from Microsoft TechNet

User Load Light Typical Heavy Extreme Request Rate 20 requests per hour. An active user generates a request every 180 seconds 36 requests per hour. An active user generates a request every 100 seconds 60 requests per hour. An active user generates a request every 60 seconds 120 requests per hour. An active user generates a request every 30 seconds Requests Per Second Per User .006 .010 .017 .034

57

Key Metrics to Monitor for ESX/ESXi

Resource Metric
%USED CPU %RDY %SYS Swapin, Swapout Memory MCTLSZ (MB) READs per second, WRITEs per second DAVG/cmd Disk KAVG/cmd GAVG/cmd MbRX/s, MbTX/s Network PKTRX/s, PKTTX/s %DRPRX, %DRPTX
58

Host / VM
Both VM Both Both Both Both Both Both Both Both Both Both

Description
CPU used over the collection interval (%) CPU time spent in ready state Percentage of time spent in the ESX host VMkernel Memory ESX/ESXi host swaps in/out from/to disk (per virtual machine, or cumulative over host) Amount of memory reclaimed from resource pool by way of ballooning Reads and writes issued in the collection interval Average latency (ms) of the device (LUN) Average latency (ms) in the VMkernel, also known as queuing time Average latency (ms) in the guest. GAVG = DAVG + KAVG Amount of data transmitted per second Packets transmitted per second Drop packets per second

Key Metrics for SharePoint

Resource Metric Description
Processor usage over a period of time. Consistently high utilization can adversely affect performance. Remember to count "Total" in multiprocessor systems. Maintain balanced performance between cores by also measuring individual core utilization Physical memory available for allocation. Insufficient memory leads to excessive use of page file and increase in page faults Rate at which faults occur when a page is sought in the file system cache and is not found. Effective use of cache for read and write operations can have a significant effect on performance Rate at which pages are read from or written to disk to resolve hard page faults. Increases in page faults indicate system-wide performance degradation High page file utilization can mean an increase in hard page faults, monitor this counter along with Pages/sec and Available Mbytes to determine if allocated memory is inadequate Number of disk reads and writes per second

CPU

% Processor Time

Available Mbytes

Memory

Cache Faults/sec

Pages/sec

Paging File

% Used % Used Peak Disk Reads/sec Disk Writes/sec

Disk Avg. Disk sec/Read Avg. Disk sec/Write Network Total Bytes/sec Average latency (seconds) of reads and writes of data from disk

Rate at which data is sent and received through the network interface

59