organizations want to: reduce infrastructure costs create more value and faster low-cost services...
TRANSCRIPT
Virtualizing Microsoft Lync Server 2013 Planning and Architecture
Campbell GunnPrincipal Lead Program Manager
OFC-B327
Understand and Size Lync Server 2013 VirtualizationMicrosoft Scenario Approach Key KHIs and Performance MetricsTesting and Stressing Approach
Session Agenda
Organizations want to:Reduce infrastructure costsCreate more value and faster low-cost services
Reduce Total Cost of OwnershipLeverage existing resources and Server ConsolidationDecrease hardware maintenance costsReduce and standardize on common platformBetter capacity planning
Operational EfficiencyImprove business continuityIncrease operational flexibilityImprove the server to administration ratio
Why Virtualize?
Virtualize Workloads
Designed to consume the full physical server capabilities
Resource Abstraction
Share infrastructure across many applications
CPUMemoryNetwork and Disk
Utilize consistency via virtualization templates
Why Virtualize?Lync Server 2013 Virtualized Workloads
Supported
Instant Messaging (IM)
IM Conferencing
Presence
Enterprise Voice (PSTN)
Audio/Video Conferencing
Web Conferencing
Application Sharing
Remote Access, Federation (Edge Server)
Response Group Service
Persistent Chat
Back-End Database
Archiving/Monitoring Databases
Hypervisor
Validated HypervisorsWindows Server 2012 Hyper-VWindows Hyper-V Server 2012
Other HypervisorsWindows Server 2012 R2 Hyper-VWindows Hyper-V Server 2012 R2
Hypervisor
Not RecommendedWindows 2008 R2 hypervisors are supported, but this is not recommended due to guests being limited to 4 cores making this unsuitable for all Lync Server roles.
Windows Server 2008 R2 Hyper-VWindows Hyper-V Server 2008 R2
KB981836 “Network connection is lost on a Windows Server 2003-based Hyper-V VM” at http://go.microsoft.com/fwlink/?LinkId=201212)
Unsupported HypervisorsWindows Server 2008
Hypervisor
OverheadTypically an overhead of 6-10%Above and beyond what the virtual guest requiresInvolves both CPU and MemoryNet increase of physical servers due to virtualization layer
Virtual Machine Image TemplatesCannot run sysprep to create machine image templates after Lync Server 2013 is installedRecommend using custom OS templates with applicable prerequisites
FEs could use templates for Internet Information Services (IIS) and Windows Media Format Runtime
Hypervisor
Live Migration and Quick MigrationNot supported
Technical challenges redirecting real time media traffic without an interruption
Can move a virtual machine running Lync Server between hostsMust be shut down prior to doing the activityAn outage will be observed by a portion of the client baseAny active conferences will be ended
MemoryDynamic Memory not supported on any rolesMemory must not be over provisioned, will starve other guest VMs
Hypervisor
Virtual DisksUse fixed-sized disks
Fixed-size VHDX delivers near native-to-physical performanceSlightly higher performance than dynamically expanding VHDX files
Differencing disks and dynamically expanding disks not supportedUnderlying Storage can become overcommitted
Allocated fixed-sized virtual disks to a static amount of disk spaceDo not store system files on the same drives dedicated to storing VMs
Creates disk contention issuesDo not use checkpoints for the virtual machines
Checkpointing creates a new secondary drive which causes reduce performance
Hypervisor
Virtual IDE and SCSIVirtual IDE controller use to boot up virtual machineAll other drives can be attached to virtual SCSI controllerVirtual SCSI can have 64 connected disks per controller and 4 controllers per virtual machine (256 Virtual SCSI disks)Virtual SCSI supports hot-add/removal of disks, Virtual IDE disk do notEnsure optimal performance and greatest flexibilityVirtualized Lync Server 2013 can use fibre channel storage solutions
Hypervisor
Guest StorageDepends on a number of factors
Storage is locally attached, or use of SANTwo Methods provided in Windows Server 2012 Hyper-V
In-guest iSCSI and Virtual Fibre Channel
Common best practices for Guest StorageOperating system boot drive must be mapped to virtual IDE driveSQL Server workloads/binaries use a virtual IDE drive or the operating system driveGuest servers need to move between hosts then iSCSI targets are directly exposed to the guest server
Hypervisor
Host Requirements
Resource Over-AllocationDo not over-allocate the CPU, only 1:1
Mixing Physical and Virtual ServersMixing of physical and virtual servers in the same pool is not supportedVirtual Front End server pool and physical Back End Servers is supported
Physical to Virtual CPU RatioPhysical to virtual CPUs ratio should not be oversubscribedOversubscription of CPU cores of host running Lync Server media workloads not supported and not recommended on other workloads
AntivirusOnly deploy antivirus software on VM host root partition (OS)
Host Requirements
Host RedundancyEnvironment should be full redundant
Ensure that similar roles are spread across multiple physical platforms
Host NetworkingEach host must have a dedicated network adaptor Share usage not supportedCorrect sizing should be used to handle the workload on the host server
KB968703, “Microsoft Support Policy for NIC Teaming with Hyper-V” at http://go.microsoft.com/fwlink/?LinkId=211073
Enable VLAN tagging on the host NICImplement (multiple) VLANs on the guests to optimize network traffic from FEs to clients and FEs to Bes
Host Requirements
Guest Requirements
Guest Core RequirementsHyper-V has not limitations to the number of CPU Cores on Windows Server 2012Other third-part hypervisors may be limited to 8 coresLync Server 2013 running on physical hardware does not have any limit to CPU cores
Guest Operating SystemWindows Server 2012Windows Server 2012 R2 preferredWindows Server 2008 R2Windows Server 2008 SP2 is not supported
Guest Requirements
Legacy versus Synthetic Virtual NICUsing Legacy adapter results in lower data transfer
Emulates Intel 21140-based PCI Fast Ethernet Adapter (usually 1GBps or faster)
Synthetic Adaptors preferredDedicated VMBus to Virtual and Physical NIC Communication
Single Root IO Virtualization (SR-IOV)Available in Windows Server 2012Provides extensions to PCI Express devices like network adapters to separate access which creates a virtual switchLegacy network adapter is not supported
Guest Requirements
IPv4 and IPv6Not recommended to mix IPv4 with IPv6
Disable Virtual DVD/CD Drives in GuestPrevents allocation resources being allotted for their function
Guest Requirements
Back End Database Role(BE DB)BE Database is no longer the real-time data store in Lync poolProvides overall persistent blob storage of data from Front End ServersDifferent in comparison to Lync Server 2010
Front End Server Role(FEs)Handles transactions for users
A/V Conferencing ServerPart of the Front End Server roleNo longer separate Monitoring nor Archiving Server role
Lync Server 2013Architectural Changes
Environment Sizing
Sizing your environmentHyper-V Guests
GuestProcessor: 12 Cores
Memory: 32 GB
Disks: 2-C: 4-D:
Network: 1GB
GuestProcessor: 8 Cores
Memory: 16 GB
Disks: 2-C: 2-D:
Network: 1GB
GuestProcessor: 4 Cores
Memory: 8 GB
Disks: 2-C:
Network: 1GB
Sizing your environmentHyper-V Guests
GuestProcessor: 12 Cores
Memory: 32 GB
Disks: 2-C: 4-D:
Network: 1GB
Used for Lync Roles:• Front ends• Back Ends• Standard Edition
Hardware Requirements: http://technet.microsoft.com/en-us/library/gg398835.aspx
Sizing your environmentHyper-V Guests
Hardware Requirements: http://technet.microsoft.com/en-us/library/gg398835.aspx
Used for Lync Roles:• Edge Servers• Directors
GuestProcessor: 8 Cores
Memory: 16 GB
Disks: 2-C: 2-D:
Network: 1GB
Sizing your environmentHyper-V Guests
GuestProcessor: 4 Cores
Memory: 8 GB
Disks: 2-C:
Network: 1GB
Used for Lync Roles:• Stress Agents
Hardware Requirements: http://technet.microsoft.com/en-us/library/gg679084.aspx
Sizing your environmentScaling the roles
Front End
Edge Server
On the server side Frontends scale the fastest • 3,333 Users / Front End
Back End
Stress Agent
Sizing your environmentScaling the roles
Front End
Edge Server
The backend will scale with Pool count.• 40,000 Users / Lync
Pool
Stress Agent
Back End
Sizing your environmentScaling the roles
Front End The edge server will scale with Pool count.• 40,000 Users / Lync
Pool
Stress Agent
Back End
Edge Server
Sizing your environmentScaling the roles
Front End The stress agents will scale the fastest• 2,500 Users / Agent Back End
Edge Server
Stress Agent
Lync Server 2013 Lab Scenario
ObservationsMultiple tests were conductedOptimal solution found
10,000 user pool with three Front End ServersSolution produced a high resource utilizationWas within critical KHI LimitsTesting is required
Due to different hardware and user model
Lync Server 2013 Scenario
Lync Server 2013 Scenario
Lync Server Topology
Host Storage Configuration
Lync Server 2013 Scenario
Storage Location
Internal External
Disks Disks 0-3 Disks 0-3 Disks 4-7 Disks 8-11
Array Array 0 (RAID 10)
Array 1 (RAID 10) Array 2 (RAID 10) Array 3 (RAID 10)
Capacity 600 GB 1.2 TB 1.2 TB 1.2 TB
IOPS (100% Read %)
700 IOPS 700 IOPS 700 IOPS 700 IOPS
IOPS (100% Write %)
350 IOPS 350 IOPS 350 IOPS 350 IOPS
Virtual Machine PlacementGuest virtual machines were spread across hosts in order to test performanceCollocation of specific server roles was not tested. Instead, we used a logical separation of rolesDid not place FEs together on same hostCan keep virtual servers in the same pool improves resilience
Lync Server 2013 Scenario
Logical Proc
Host01 Host02 Host03 Host04 Host05 Host06 Host07 Host08
00-03 FE01 FE02 FE03 BE01 BE02 Edge01 Edge02 ST0504-07 08-11 DC01 DC02 ST0612-15 16-19 OWAS01 BE03
(Witness)
OWAS02 ST07
20-23 24-27 ST01 ST02 ST03 ST04 ST0828-31 Reserved Reserved Reserved Reserved Reserved Reserved Reserved Reserved
LSS Ramp UpThroughout testing process required a ramp up period of about 30minsDuring Ramp-Up time, the following occurred:
LSS Configuration files are loadedUsers are logged inLync Server completes some background tasks
Lync Server 2013 Scenario
LSS Ramp Up
Lync Server 2013 Scenario
SQL ExpressFour processors remained in high utilization over time
Associated with the RTCLOCAL SQL Server Express Edition instanceSQL Server Express Edition can use only four logical processors
http://technet.microsoft.com/en-us/library/ms143760.aspx>90% utilization indicates Enterprise with 3 FEs is already at full capacity
From a SQL Server Express viewpoint and can be accommodated by addition Front End Servers
Lync Server 2013 is optimized to utilize the four cores assigned for SQL Express ComputeKey KHI is Queue Latency and Sproc Latency
Lync Server 2013 Scenario
SQL Express
Lync Server 2013 Scenario
Stress ToolingUser Provisioning Tool
• Create users and groups
• Enable users for Lync• Generate contact lists• Create test location
info.
Changing “Average Contacts per User” will increase subscriptions per user and increase processor utilization on the Lync Frontends.
• Specify Stress Agents• Service discovery• Enable Modalities to
test• Set load variables such
as number and frequency of PSTN calls
Be mindful of the features you enabled and how you configure them. For instance the Reach client can increase processor utilization on the Frontends.
Demo
User Provisioning ToolLync Load Configuration Tool
Testing
Verify your environment is functionalRun tests for 24 hoursRun the tests twiceUse control clients
Testing Methodology
Running TestsLync Load Configuration Tool
• Once desired configuration determined Generate files
• Directory created for each stress agent
• Each directory should be copied the corresponding agent
Running TestsLync Performance and stress tool
• Runs all the tool generated tests
• Runs a specific test such as instant messaging
Validation ToolsetPerform stress testing of a targeted deployment infrastructureCan target a specific usage patternSimulate individual workloads, such as IM and presenceUse to provision users and simulate user loadsEnsure proper design and deploymentPerformed after technical design, before pilotingDeploy into lab (controlled environment)Requires clients for stress testing
Stress and PerformanceLync Server Stress and Performance Tool (LSS)
10 Critical Key Health IndicatorsProvide the basis for a “Good Pass”Ease of checking all the indicators in each runIf testing reveals no issues in the 10 KHIs
Further analysis of the full KHI set is recommendedNetwork Planning, Monitoring, and Troubleshooting with Lync Server documentation http://www.microsoft.com/en-us/download/details.aspx?id=39084
Key Health IndicatorsGood Pass
10 Key Health IndicatorsCounter Description Healthy Range\Processor Information(*)\ % Processor Time
CPU Utilization < 80% (highest core)
\Memory\Available Mbytes Displays the amount of physical memory, in bytes, available to processes running on the computer. >20% total system MB
\PhysicalDisk\Avg. Disk Sec/ReadA key measure of disk latency representing the average time, in milliseconds, of each read to disk where > 20 is poor, <20 is good/fair, <12 is better, <8 is best
<10 ms Green< 10-25 ms Yellow> 25 ms Red
\PhysicalDisk\Avg. Disk sec/Write
A key measure of disk latency representing the average time, in milliseconds, of each write to disk, where non-cached writes.
(>20 poor, <20 fair, <12 better, <8 best), differing significantly from cached writes (>4 poor, <4 fair, <2 better, <1 best ).
For online transaction processing (OLTP) databases, the lower this number the better, especially for disks holding the transaction log.
< 10 ms Green< 10-25 ms Yellow> 25 ms Red
LS:Usrv - DBStore\Usrv – Queue Latency (msec)
Queue latency is the amount of time (in milliseconds) that it takes for a request to leave the Front End Server’s queue toward the back-end database.If this value by itself is high, you may have be encountering a network problem between the Front End Server and the SQL Server.If this value is high and Sproc Latency is high, it is most commonly a bottleneck with the Back End Server.
<100ms sustained
LS:USrv – DBStore\Usrv – Sproc Latency (msec)
Sproc latency is the amount of time (in milliseconds) that it takes for the SQL Server database to process the request. This performance value is collected from the time the request leaves the Front End Server queue until that the request returns.If this value by itself is high, it is likely a bottleneck with the Back End Server.
<100ms sustained
\SQL Server Buffer Manager\Page life expectancy
Tells, on average, how many seconds SQL Server expects a data page to stay in cache. The target on an OLTP system should be at least 300 (5 min). When under 300, this may indicate poor index design (leading to increased disk I/O and less effective use of memory) or, simply, a potential shortage of memory.
A low value, below 3600 seconds (together with high lazy writes/sec and checkpoint pages/sec) indicates memory pressure.
< 300
\MSSQL$RTC:Memory Manager\Total Server Memory(KB)
Shows the amount of memory that SQL Server is currently using. This value should grow until it is equal to Target Server Memory, as it populates its caches and loads pages into memory. When it has finished, SQL Server is said to be in a “steady-state”. Until it is in steady-state, performance may be slow and IO may be higher.
< 80% of physical memory
\ASP.NET Apps v2.0.50727(*)\Requests Rejected
< 80%
Performance MonitoringKey Health Indicators
CPUCPU Utilization < 80%
DiskAvg. Disk Write < 10 msAvg. Disk Read < 10 ms
MemoryAvailable MB
>20% System Total
NetworkQueue Length < 2
Discarded (in / out) = 0
SQL
AS/AV/IM MCUMCU Health State <2
Web Components
LySS
Page life expectancy > 300 Sec. Batch requests / sec < 2500
% of space used by Storage Service DB < 80 # of replica replication failures = 0# of data loss event = 0
Distribution List expansion AD timeouts <0ABWQ failures = 0LIS failures = 0Authentication Errors < 1/secASP.NET v4 Requests Rejected = 0
SIP StackAvg. Incoming Message Processing < 1 secIncoming Responses Dropped < 1/secIncoming Requests Dropped < 1/secQueue Latency < 100 ms Sproc Latency < 100 msThrottled Requests = 0
Authentication Errors < 1/secIncoming Messages Timed Out < 2Avg. Incoming Message Hold < 1 secFlow Controlled Connections < 2Avg. Out Queue Delay < 2 sec
• 2 categories, service & resource instrumentation
• Add to performance collections
• Analyze to determine services are in a healthy state
Performance MonitoringPerformance Counters
CPUCPU Utilization < 80%
DiskAvg. Disk Write < 10 msAvg. Disk Read < 10 ms
MemoryAvailable MB
>20% System Total
NetworkQueue Length < 2
Discarded (in / out) = 0
SQL
AS/AV/IM MCUMCU Health State <2
Web Components
LySS
Page life expectancy > 300 Sec. Batch requests / sec < 2500
% of space used by Storage Service DB < 80 # of replica replication failures = 0# of data loss event = 0
Distribution List expansion AD timeouts <0ABWQ failures = 0LIS failures = 0Authentication Errors < 1/secASP.NET v4 Requests Rejected = 0
SIP StackAvg. Incoming Message Processing < 1 secIncoming Responses Dropped < 1/secIncoming Requests Dropped < 1/secQueue Latency < 100 ms Sproc Latency < 100 msThrottled Requests = 0
Authentication Errors < 1/secIncoming Messages Timed Out < 2Avg. Incoming Message Hold < 1 secFlow Controlled Connections < 2Avg. Out Queue Delay < 2 sec
Actual Counter Names• LS:DATAMCU - MCU Health And Performance\DATAMCU - MCU Health
State• LS:AVMCU - MCU Health And Performance\AVMCU - MCU Health State• LS:AsMcu - MCU Health And Performance\ASMCU - MCU Health State• LS:ImMcu - MCU Health And Performance\IMMCU - MCU Health State
Understand and Size Lync Server 2013 VirtualizationMicrosoft Scenario Approach Key KHIs and Performance MetricsTesting and Stressing Approach
Session Summary
Resources
Learning
Microsoft Certification & Training Resources
www.microsoft.com/learning
Developer Network
http://developer.microsoft.com
TechNet
Resources for IT Professionals
http://microsoft.com/technet
Sessions on Demand
http://channel9.msdn.com/Events/TechEd
Technical Network
Join the conversation!Share tips and best
practices with other Office 365 expertshttp://aka.ms/o365technetwork
Managing Office 365 Identities and Services
5
Office 365
Deploying Office 365 Services
Classroomtraining
Exams
+
Introduction to Office 365
Managing Office 365 Identities and Requirements
FLC
40041
Onlinetraining
Managing Office 365 Identities and ServicesOffice 365 Fundamentals
http://bit.ly/O365-Cert
http://bit.ly/O365-MVA
http://bit.ly/O365-Training
Get certified for 1/2 the price at TechEd Europe 2014!http://bit.ly/TechEd-CertDeal
MOC
20346 Designing for Office
365 Infrastructure
MOC
10968
3
EXAM
346EXAM
347
MVA MVA
TechEd Mobile app for session evaluations is currently offline
SUBMIT YOUR TECHED EVALUATIONSFill out an evaluation via
CommNet Station/PC: Schedule Builder
LogIn: europe.msteched.com/catalog
We value your feedback!
© 2014 Microsoft Corporation. All rights reserved. Microsoft, Windows, 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.