The Path to Continuous Availability with Windows Server 2012Luke NotleyTechnical Consultant – Red Ember
WSV326
Availability Matters…
Why now? What you said
Keep my data safe
Keep my services running
Keep my business running
Server virtualization
Server consolidation
User expectations
Availability Spectrum
Address planned and unplanned downtime
Reduce frequency of failures
Manage failures or interruptions
Focus on the availability of the service
Single server
Scale up and out
Multi-site
cost
SL
A
Continuous Availability
Transparent failover is ability to survive planned moves or unplanned failures – without errors, without losing data, and while performing well at scale
Continuously available
software and hardware platforms are designed to
support transparent failover without data loss
introducing
Windows Server 2012 'Active On'
Disk Error Protection
Data recovery from unreadable disk errorsDirectory corruption repair in seconds
With disk online with CSVNo ChkDsk at all with ReFS
Storage Virtualisation and Resiliency
Data redundancy with either n-way mirroring (clustered or unclustered) or parity mode (unclustered)
Designed for resiliency on industry standard storage hardware – JBOD with SAS (clustered) or SATA (clustered or unclustered)
Integrated with Windows Failover Clustering & Cluster Shared Volumes (CSV)
Software-only storage virtualization of individual disks to manage as pools
NTFS Online Scan and Repair aka chkdskDisk scanning process separated from repair process
Scanning done with volume online, taken offline only for time to repair – based on number of errors to fix rather than size of volume
Zero offline time when used with CSV
Windows Server 2012 will support volumes up to 64 TB.
100
Millio
n File
s
200
Millio
n File
s
300
Millio
n File
s0
50
100
150
200
250
300
350
400
Windows Server 2008 R2
Windows Server 2012
Min
ute
s
Resilient File System (ReFS)
Data corruption avoidedAllocate-on-write eliminates corruption from in-place update errors Checksums detect whether the data on disk has changed since it was last written
Data integrity protectionFile system metadata is always protected, maximizing online operationOptional user data protection (integrity streams) is configurable on a per-volume, per-directory, and per-file basis
Rapid recovery from file system corruption without affecting volume availability
If redundant copy available, automatic correctionIf not, salvage done online and only corrupt portion of namespace affected
Managing Memory Errors
Server memory errors isolated to single application or VM Other VMs and host server continue running
Memory Fault IsolationOS support for consumed memory error recovery on capable x64 processors
Application Assisted Memory Error Recovery
Gives applications an opportunity to free memory that is marked for offline
Hyper-V supports recovery from uncorrected memory errors
Machine check signaling of uncorrected memory error injected into the affected guest VM
If the affected memory page is not in use, or can be freed by the application or OS, then page is unmapped, marked as bad, and taken out of service
Event log
Processor
User mode process
WHEA2
3
5
KERNEL
4
Fetch data or execute code from memory location with uncorrected error
1
Mem
ory
Machine Check Exception
Handle Exception
Recovery Actions
Recovery Actions:User Mode
• Terminate the process• Take memory offline
Kernel Mode• BugCheck the system
2
Clustering Applications and Services
Scale HA applications and VM infrastructure both up and outEasy setup experience for clustered appliancesNative highly-available DHCP service
Windows Failover Clustering
Increased scale out and scale up
Scale out to 64-nodesScale up to 4,000 VMs per cluster4x > Windows Server 2008 R2
Much faster failover timesCluster Shared Volumes (CSV) v2 supported for general use
. . .
Sca
le u
p
Scale out
. . .
Simplified Cluster Setup
Easy setup of clustered NAS or Hyper-V appliances
Fast: Approximately 15 minutes from power-on to file sharing
Simple: Minimal set of setup steps with guided workflow
Single console: Single-pane-of-glass to configure appliance from only one node
Extensible: OEM customization framework to enable unique needs of appliance
demo
Configuring Clustering
DHCP Failover
2 Modes of Operation:Load Balance (Active-Active)
Both servers respond to clients in proportion to a configured load distribution percentageSuited for single-site deployment
Hot Standby (Active-Passive)Only one of the servers serves all the clients on the subnetStandby servers serve IP address only when the primary is downSuited for multi-site deployments
GoalsContinuous service availabilityIP address continuity to clientsLower TCO of DHCP HA deployment
Solution2 servers form part of a failover setupServers replicate lease information between each otherServers load balance the client requests between them
Configuring DHCP Failover
Continuously Available Storage
Highly available, high performance, low cost storage for Hyper-V and SQL ServerHeterogeneous storage for VMware and non-Windows applications
SMB Transparent Failover
Failover transparent to server application – zero downtime, just short IO delay during failover
Supports both planned and unplanned failovers Resilient for both file and directory operations
Witness service allows failover without waiting for TCP timeout on client
File Server Cluster
SQL Server or Hyper-V Server
Failover to Node B
2
Normal operation
Connections and handles auto-recovered; application IO continues with no errors
1 3
File Server Node A
File Server Node B
\\fs1\share \\fs1\share
Configuring SMB Failover
Both sides of SMB connection must be Server 2012 or Windows 8
SMB Multi-channel
Uses multiple TCP connections for each SMB session if available
Transparent failover – recovers network failure if another connection is available
Improved throughputBandwidth aggregation with multiple NICsUtilizes multiple CPUs for network processing with multiple or RSS-capable NICsCan combine with LBFO teams
Simple automatic configuration - SMB detects and uses multiple network paths
Multiple 10GbE/IB RSS-capable NICs
Multiple 1GbE NICs
SMB Server
SMB Client
Switch1GbE
SMB Server
SMB Client
NIC1GbE
NIC1GbE
Switch1GbE
NIC1GbE
NIC1GbE
Switch10GbE/IB
NIC10GbE/IB
NIC10GbE/IB
Switch10GbE/IB
NIC10GbE/IB
NIC10GbE/IB
Sample Configurations
SMB Multi-channel
No user configuration necessary – connections are made automatically and dynamically
Both ends of the SMB connection must be running Windows Server 2012 (or Windows 8)
Parallel data transfer requires similar network interfaces; failover can occur across dissimilar interfaces
Scale-Out File Server
Active/Active access to the same shares through any cluster node
Simpler deployment and management
Distributed Net Name: Single DNS name for clusterCSV: Cluster-wide file system available on all nodesConsistent cluster-wide file server configuration
Zero-downtime client failover – both planned and unplanned
Single File System Namespace
Cluster Shared Volumes
Single Logical Server (\\Foo\Share)
SMB
Hyper-V Cluster
File Server Cluster
Clustered StorageNFS v3 and iSCSI Target
Continuously available:iSCSI block storage for diskless boot & application (initiators)NFS v3 shares for UNIX/Linux clients and VMware VMDKs
Transparent & fast:Failover transparent to client – zero downtime, just short IO delay during failoverSupports both planned and unplanned failovers
Clustered iSCSI Target or NFS v3 server
iSCSI Initiator(s),NFS v3 clients
1
23
Node A Node B
Normal storage access through Node A
Failover of iSCSI Target or NFS connections to Node B
Client automatically reconnects to Node B and application IO continues
demo
Clustered File Shares
Non-disruptive Reconfiguration & Updating
Reconfiguring VM storage without downtimeAutomated patching of clustered servers without downtime
Hyper-V Storage Live Migration
Move VHDs for VM from one disk location to another
Zero VM downtime
New storage location must be reachable from Hyper-V host
Can use SMB file server as target
VHD
Cluster Aware Updating
Simple orchestration of cluster node updates
Determines updates needed, moves workloads off nodes for updates
Uses Windows Update AgentIdentifies node with least loadPuts node in maintenance modeVerifies success, then moves to next node
Maintains service availability and without impacting cluster quorum
Windows Update
UpdateCoordinator
demo
Cluster Aware Updating
VM Multi-site Recovery
Easy basic DR replication of VMs
Hyper-V ReplicaSimple, secure DR replica of VMs across standard networksReachable to any reachable Windows Server 2012 Hyper-V server
Storage types can be differentVendor independent
Can set for application-consistent or crash consistent snapshots (VSS inside VM)
What does this all mean?
Windows Server 2012 storage can deliver enterprise-class storage availability, reliability and performance – at a fraction of the cost
Windows Server 2012 continuously available storage
Flexible, low-cost solutions for Hyper-V and SQL ServerEasy to set up and configure
More options
for storage
Related Content
Find Me tomorrow morning at the Speaker Community Lounge
WSV325 - WS2012 File and Storage Services Management – Jeff Alexander
WSV326 - The Path to Continuous Availability with WS2012 – Luke Notley
WSV332 - Cluster-Aware Updating & the New Generation of WSUS – Orin Thomas
© 2012 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista 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.