Transcript Virtual Networks - Microsoft NT konferenca

DOBRE PRAKSE POSTAVITVE
VIRTUALIZIRANEGA DATACENTRA
Gorazd Šemrov
Microsoft Consulting Services
[email protected]
AGENDA
• UVOD
• Zakaj virtualizirati, glavni motivatorji
• Glavne usmeritve in arhitekturne posebnosti pri
virtualiziranju storitev
• Kaj in kako virtualizirati (SQL, SP, EX)
• FAILOVER CLUSTER in HYPER-V konfiguracija
• Compute
• Networking
• Storage
• HEALTH CHECK
UVOD
ZAKAJ VIRTUALIZIRATI IN CENTRALIZIRATI DATACENTER
• Motivatorji
•
•
•
•
Prostorske omejitve, Napajanje, Hlajenje
Centralized data / deduplication / replication
24/7 operations
Konstantna rast zahteva optimizacijo / compliance
• …manjši stroški
• Nižji skupni stroški IT
• Nižji stroški upravljanja
• Nižji stroški operacij
• Fokusira se na čimboljši izkoristek virov
• Visoko razmerje VM/host
• Nižji stročki licenciranja
• Posebne arhitekturne komponente
• Failover clustering za zagotavljanje visoke razpoložljivosti
• Performance Resource Optimization, Dynamic Optimization, Power
Optimization
• Live Migration, Guest Clustering, Application Clustering, App-V
CLUSTER ENTERPRISE DEPLOYMENT
Library
Server
Administrator
Console
Site A
Site B
SAN
Storage
Geo-Cluster
SAN
Storage
Clustered HyperV
Hosts
Hyper-V
Hosts
KAJ VIRTUALIZIRATI
• Performančne omejitve
• Sizing/Configuration - SAN, Network, Compute
• Hyper-V performance proti fizičnem svetu dosegajo 95% do 99%
performanc fizičnih diskov in 89% do 98% performanc različnih bremen
v primerjavi z fizičnim svetom.
• Varnostne omejitve
• Arhitektura
• Kombinacija V/P - Management
• Primeri iz prakse
• Microsoft IT
• Kolektor
• GEN Energija
• Key Points
• Management
• Visoka razpoložljivost
OPTIMIZED FOR BUSINESS CRITICAL
APPLICATIONS
• 460,800 light-weight 1% concurrent users
• 80,000 simulated OLTP customers
• 20,000 Exchange 2010 mailboxes
It’s clear that Hyper-V R2 SP1 can be used to virtualize tier-1 data
center applications with confidence.
– Enterprise Strategy Group Research, Mark Bowker
HYPER-V FIXED VIRTUAL HARD
DISK PERFORMANCE OVERHEAD
ANALYSIS
Native vs. Fixed VHD IO Performance Analysis
Disk IOs per second (IOPS)
(Window 2008 R2 Physical Disk(Native) vs. Fixed VHD(VM), ESG Lab Iometer workloads)
SQL Log
OLTP Database
Physical Disk (Native)
File Server
Fixed VHD (VM)
Exchange DB
HYPER-V R2 APPLICATION
OVERHEAD ANALYSIS
Virtual vs. Physical Performance Analysis
(Relative difference in elapsed time per operation; Windows 2008 R2 data center)
Application install
Directory level copy
Subsequent copies
SQL query
0%
10%
20%
30%
40%
Native
50%
60%
70%
Hyper-V R2 VM
80%
90%
100%
SQL CONSOLIDATION
CONSIDERATIONS
Multiple SQL Instances
Multiple Virtual Machines (VM)
Shared Windows instance
Dedicated Windows instance
Number of CPUs visible to Windows
instance
Up to 4 virtual CPUs
CPU over-commit is supported
Memory
Server Limit
Dynamic(max server memory)
Statically allocated to VM (Offline changes only)
64GB limit per VM
2 TB Limit per Host
Storage
SQL Data Files with standard storage
options
SQL Data Files using
Passthrough or Virtual Hard Disks exposed to VM
Windows System Resource
Manager(process level)
SQL Server Resource Governor
Hyper-V guest VM
SQL Server Resource Governor
50
Practical limit determined by physical resources
Clustering, Database Mirroring, Log
Shipping, Replication
Live Migration, Guest Clustering, Database Mirroring,
Log Shipping, Replication
Good
Comparable with multiple instances, acceptable
overhead
Isolation
CPU Resources
Resource
Management
Number of
instances
High Availability
Performance
SQL SERVER BEST PRACTICES
•
•
•
•
Guest virtual machines are limited to 4 CPU cores
Best performance if VMs are not over-committed for CPU
Test Network intensive applications for acceptable SLAs
Use multi-pathing on host or within the VM to ensure
maximum throughput and high availability for VM
workloads
• Utilize either pass-through disk or fixed-size VHD for
guest virtual machines
• Avoid using emulated devices. Instead, ensure latest
integration components are installed and synthetic
devices are being used.
SHAREPOINT ROLES &
VIRTUALIZATION CONSIDERATIONS
Virtualizatio
n
Decision
Considerations and Requirements
Ideal
• Easily provision additional servers for load balancing and fault
tolerance
Query Role
Process Search
Queries
Ideal
• For large indexes, use physical volume over dynamic
expanding VHD
• Requires propagated copy of local index
Application Role
Excel Forms Services
Ideal
• Provision more servers as resource requirements for individual
applications increase
Role
Web Role
Render Content
Index Role
Crawl Index
Database Role
Consider
Consider
• Environments where significant amount of content is not
crawled
• Requires enough drive space to store the index corpus
• Environments with lower resource usage requirements
• Implement SQL Server® alias for the farm required
For details, see SharePoint Virtualization whitepaper at
http://www.microsoft.com/virtualization/solutions/business-critical-applications
VIRTUALIZING EXCHANGE 2010 SP1
•
Best practices for Virtualizing Exchange 2010 with Windows Server 2008
R2 Hyper V
http://www.microsoft.com/downloads/en/details.aspx?FamilyID=8647c69d-6c2c-40ca977e-18c2379b07ad
•
DAG and UM on Hyper-V cluster is now supported (Exchange 2010 SP1!)
•
•
Exchange server virtual machines (including Exchange Mailbox virtual machines that are part of
a DAG), may be combined with host-based failover clustering and migration technology, as
long as the virtual machines are configured such that they will not save and restore state
on disk when moved, or taken offline. All failover activity must result in a cold boot when
the virtual machine is activated on the target node. All planned migration must either
result in shutdown and cold boot, or an online migration that makes use of a technology
like Hyper-V Live Migration
New support policy
http://technet.microsoft.com/en-us/library/aa996719.aspx
•
STORAGE – block level only
•
•
•
•
•
Fixed size VHD
SCSI pass through disk
ISCSI disk
Cheap storage for db, archive
Snapshots arent supported
USE WINDOWS SERVER 2008 R2
HYPER-V SP1
•
•
•
Many of the scenarios discussed leverage features new to
R2 and R2 SP1
Live migration
New processor support
•
•
Enhanced scalability
•
•
•
Improve performance
10 Gb/E ready
Dynamic VM capabilities
Enhancements to server core
•
•
•
Greater VM density
Lower TCO
Networking enhancements
•
•
•
•
Improved performance SLAT
Ease management
Lower TCO
Dynamic Memory with SP1
HYPER-V CONFIGURATION
GUIDELINES
• Hyper-V Root Configuration
•
•
•
•
Plan for 1GB+ memory reserve for the management OS in the root partition
Plan for one dedicated NIC for management purposes
Separate LUNs/Arrays for management OS, guest OS VHDs and VM storage
Management OS and VHD LUNs should employ RAID to provide data
protection and performance
• HKLM\SOFTWARE\Microsoft\Windows
NT\CurrentVersion\Virtualization\MemoryReserve
• Hyper-V Guest Configuration
• Fixed-sized VHDs for Virtual OS
• Need to account for page file consumption in addition to OS requirements
OS VHD Size (minimum 15GB) + VM Memory Size = Minimum VHD size
• Account for space needed by additional files by VM (ISCSI – hot add)
• Example for SQL: OS VHD Size + (VM Memory Size) + Data Files + Log
Files
FAILOVER CLUSTER
48% of Cluster support calls are due to
configuration problems
Most failures are due to human
error
-Microsoft PSS
Configuration Issues




Cabling mistakes
SP and hotfix binaries
Driver mismatches
Inconsistent settings
Complexity
 Best practices
 Supportability
requirements
 Hardware compatibility
If we can eliminate the configuration issues up
front, we can ensure a better cluster experience
(installation and operation)
FAILOVER CLUSTER REQUIREMENTS
Operating system: Windows Server® 2008 R2 Enterprise Edition
Multiple network adapters on each node connected to different IP
subnets (production and heartbeat)
Access by each node to a shared storage device (fibre channel, iSCSI,
or SAS)
HAVM
EMPOWERED BY: FAILOVER CLUSTER
• Failover Cluster provides key features to make
Hyper-V VMs Highly Available
•
•
•
•
Live Migration / Quick Migration
CSV
Health Monitoring
Automatic Recovery
NETWORKING
• Best Scenario – 4 physical network access points
• 1 – Virtual Machine Access
• 2 – Management
• 3 – Cluster and CSV
• 4 – Live Migration
• (5th network when using ISCSI)
• Gigabit or faster NICs
• Teaming ok – Ensure latest drivers, use good Server Class NICs
• All nodes must be on the same ‘logical’ subnet
• Configure Network Priority
• Consider QoS policy for Live Migration Network
• Ensure networks are aptly named
http://technet.microsoft.com/en-us/library/ff428137(WS.10).aspx
LIVE MIGRATION
• Identical Processors
• May enable Processor Compatibility
• Must be same processor manufacturer
• Identically named ‘Virtual Switches’
• Ensure that no ISOs are mounted – script
• VMM PS script:
get-vmmserver <vmmservername> | Get-VM | Get-VirtualDVDDrive | where
{$_.Connection -ne "None"} | Set-VirtualDVDDrive -NoMedia
•
•
•
•
VM may need to be ‘refreshed’ in Failover Cluster
Specify the network for the VMs to use for Live Migration
Don’t use IPSec on LM network unless required for security
Linux HAVM needs Static MAC Address
DEMO
DEMO NASLOV:
NIC HA SCRIPT
http://social.technet.microsoft.com/Forums/enUS/ITCG/thread/5837cce8-b1b5-4654-bb590c49a7117727
HYPER-V SETUP AND NETWORKING
• Do not create virtual networks in the Hyper-V Roles Wizard in
case of failover cluster scenarios!
• Use identical names for virtual networks on failover cluster hosts
VIRTUAL NETWORKS
• External networks
• Connect to the outside world
• Bound to a physical network
adapter
• Internal networks
• Connect to the parent partition
only
• Private networks
• Exclusively virtual (within the
parent partition)
Virtual networks in Hyper-V do not
provide any network services—
only connectivity.
VIRTUAL MACHINE WITH LEGACY
AND SYNTHETIC NIC
• Legacy Network Adapter supports PXE Boot
SELECT THE BEST CLUSTER
QUORUM MODEL
• New majority-based quorum model
• Majority-of-nodes-based quorum
• Disk is an optional witness to have a vote in deciding majority
• Three total votes, with two needed for a majority
• So the cluster can survive the loss of any one vote
Vote
Vote
Each node counts as
1 vote
Node 1
Node 2
SAN
Vote
Shared Storage
Device gets 1 vote
SHARED STORAGE
TOPOLOGY REQUIREMENTS
• Only storage that supports persistent reservations will be supported in
Windows Server 2008 R2 failover clustering
• Deprecating parallel-SCSI support
• Serial Attached SCSI (SAS)—based clusters will replace parallel-SCSI
• Application cluster with Hyper-V is only iSCSI-supported!
Fibre Channel
iSCSI
Supported Shared Bus Types
SAS
HYPER-V HIGH AVAILABLE SAN
ARCHITECTURE
Hyper-V Server
MPIO Store Port Driver
HBA1
HBA2
FC Switches
SP1
SP2
VM1.VHDs
VM2.VHDs
LUN1
LUN2
Disk Array
Group
CSV – SAN FAULT TOLERANCE
I/O Redirected
Over CSV
Network
VM running on
Node 2
Coordinator
Node
SAN
SAN Connectivity
Failure
VHD
VHD
VHD
CSV VOLUME
• Nodes running Windows Server 2008 R2
• Drive letter of the System Disk must be identical
• %SystemDrive%\ClusterStorage
• NTLM Protocol enabled on all nodes
• Identical / latest HBA/Storage/Drivers/Firmware
• SMB used by CSV
• Client for Microsoft Networks
• File and Printer Sharing for Microsoft Networks
• Check Free Space – Use PowerShell Script
http://blogs.msdn.com/b/clustering/archive/2010/06/19/10027366.aspx
Snapshots – VMM script
get-vmmserver <vmmservername> | Get-Vm | get-vmcheckpoint |
where {$_.AddedTime -ne "None"} | ft VM, AddedTime
CSV NETWORK SETTINGS
• Highest priority network = CSV network
• Network with lowest ‘metric’ is the highest priority
• Manually set CSV network if not set properly:
• Get-ClusterNetwork | ft Name, Metric, AutoMetric,
Role
• (Get-ClusterNetwork “Cluster Network”).Metric=900
DEMO
DEMO NASLOV:
CSV network
STORAGE RECOMMENDATIONS
Preveri podprtost in validacijo storage strojne opreme in funkcionalnosti
Uporabi storage vendor dokumentacijo in priporočila za planiranje in uvedbo
Zagotovi fizično varnost SAN ali iSCSI komponent
Uporabi multipathing za SAN ali iSCSI data transfer
Ensure the security of network services relies on (iSNS, DNS, Network)
Test and install the latest hotfixes, drivers, and firmware on cluster nodes
only with hardware vendor support
Use dedicated network controller for iSCSI and VMs
Use large caches on Storage Arrays to reduce IO- latencies
Storage performance has impact to minimize VM-migration time
HEALTH CHECK
HYPER-V CLUSTER BEST PRACTICE
CHECKS
Validate Report?
Network adapters- Number of? Speed?
Metric (Priority) Settings? Role (1-3)? Client Access - Private/Live Migration network, AutoMetric enabled
Selected Network for Live Migration?
Network configuration - Other iSCSI in use? Verify Proper/Dedicated Network Path
Any VMs running on disks that are in "Redirected Access Mode"? Maintenance Mode?
Heartbeat monitoring for the Virtual Machine - Enabled/Disabled?
Policy for VM recovery/actions
Is Persistent Mode/Auto Start Enabled?
VMStates? Paused/Online/Offline/Failed/
CSV enabled? State of the disks? (R2 ONLY)
How many VMs per (CSV) volume? Optimization available?
GPT disk/MBR disk?
Multi-site cluster?
H/W Logo'd for Win2008/R2?
Filter Drivers Loaded?
Backup/Restore Process of (CSV or other) volumes
Perfmon counters for CSV volumes
Processors must be identical across cluster nodes - Processor Compat mode enabled?
Is Live Migration working properly?
Event logs for failed migrations?
System Drive letter must be the same across nodes
Verify supported workloads are running on each VM
Check firmware version of HBA from validation report and matches on all nodes
Verify DSM Module in use
Check if network teaming used for ISCSI
Is MPIO feature installed
Check for proper cluster Quorum Model
Storage configured to vendor tested configuration
Chkdsk set to run on any drives?
Microsoft
Approved!
HYPER-V BEST PRACTICES
ANALYZER
REQUIREMENTS
•
•
•
• Update 977238 required on Hyper-V Hosts +
update for DM SP1 - KB 2485986
•
•
• http://gallery.technet.microsoft.com/scriptcenter/enus/84d3e608-1404-4ee9-b5e2-24452ecc8b94
WHY HYPER-V BEST PRACTICES
ANALYZER ?
Integration
Components
Number of logical
processors
Number of VPs
on each VM
VHD type
detection
• Installed? What version? Are all of the ICs added to the VMs?
• Supported LPs number on host
• Inventory and supported number of VPs for each OS
• Fixed, Dynamic or Differencing Disk inventory
Legacy Network
Adapters
• Only use it if it’s mandatory (i.e Citrix provisioning VDI scenarios)
Much more…
• http://technet.microsoft.com/en-us/library/ee941122(WS.10).aspx
OTHER
•
•
•
•
Use Possible and Preferred Owners with VM resources
Persistent Mode, Autostart
Ensure proper amounts of RAM per node
Base performance on least powerful node (as applicable)
• Each node must be able to handle the workload it may
be given
• SCVMM PRO and SCOM feature for load balancing
• Use VMQ in place of Chimney Offload
• More efficient copying between networks
• Allows threads to be scheduled on multiple processors
• http://blogs.technet.com/b/kaitling/archive/2011/02/10/hav
m-cluster-configuration-recommendations.aspx
DEMO
DEMO NASLOV:
HYPER-V Best Practices Analyzer
WHY CLUSTER VALIDATION
WIZARD?
Shared
Storage
Configuration
• Configuration and functional testing
Cluster
Network
Configuration
• Heartbeat, CSV and LiveMigration network
configuration tests
Inventory and
Compare
• Hardware, updates and drivers
Cluster
Configuration
• Active Directory settings, DNS, Quorum,
dependencies and many more…
CLUSTER VALIDATION WIZARD
• Windows Server 2008 or 2008 R2
• Failover Cluster Feature enabled
• Storage and Network Connections Configured
• Run Validate immediately after setup of Virtual Switches
• Validate does not currently check Hyper-V specific config
More info:
http://technet.microsoft.com/en-us/library/cc732035(WS.10).aspx
WHAT TO MONITOR
•
•
•
•
•
•
•
CSV PERFORMANCE BASELINE
• What is the data telling us?
• Counters from Parent Partition:
• Object: Physical Disk
• Disk Reads/sec, Disk Writes/sec, Avg Disk Queue, Current
Disk Queue
• Multiple nodes are accessing the same CSV
• Aggregate data across nodes
• Reads/Sec, Writes/Sec
• Disk latency issues
• Check counters within VMs
http://blogs.technet.com/b/kaitling/archive/2011/02/11/decipheringperformance-data-csv-volumes.aspx
OPERATIONS MANAGER
NADZOR IN POROČILA
Nadzor operacij, ki se izvajajo v
virtualiziranem okolju. Samodejno
odkrivanje komponent. End to end
monitoring
VPRAŠANJA?
Po zaključku predavanja prosim izpolnite vprašalnik.
Vprašalniki bodo poslani na vaš e-naslov, dostopni pa bodo tudi preko profila na spletnem
portalu konference. www.ntk.si .
Z izpolnjevanjem le tega pripomorete k izboljšanju konference.
Hvala!