WSV 423 Cluster Shared Volumes Reborn in Windows

WSV 423 Cluster Shared Volumes Reborn in Windows Server 2012 Gareth James Senior Consultant Hani Adhami Senior Consultant

Challenges of Managing Hyper-V Clusters Manageability • Multi-path • Masking several LUN’s Flexibility • LUN - smallest unit of failover Capacity Scalability • Poor SAN space utilisation • Complexity with drive letters

Windows Server 2008 R 2

What it delivers

Under the hood CSV Architecture

I/O Synchronisation Overview Metadata Simultaneous read/write access on all Cluster Nodes Read/Write Shared LUN VHD VHD Shared Storage Server side metadata synchronisation - Avoids I/O interruptions

Key Takeaways Metadata updates - small operations, infrequent for VMs Parallel metadata updates - non-disruptive for applications

Share VM VM Share Node 1 Share Coordination Node 2 VM Node 3 CSVFS CSV Proxy File System Server / SMB MUP/RDB SS/SMB CSV Proxy File System CSV File System Filter NTFS CSV Volume. M gr Volume Manager LBFO/RDMA Disk Storage Connection Broken or not present Direct I/O SAN

Configuring a CSV Disk

demo Setting up Clustered Shared Volumes

Consistent view across the cluster

Mount Points

Resiliency How CSV Enables Even Higher Availability

CSV Resiliency

I/O Redirected via network VM running on Node 2 is unaffected Coordination Node SAN Connectivity Failure VHD VM’s can then be live migrated to another node with zero client downtime

New Coordinator Node Failure VM running on Node 2 is unaffected Coordination Node Brief queuing of I/O while volume ownership is changed Volume relocates to a healthy node VHD

Metadata Updates Rerouted to redundant network VM running on Node 2 is unaffected Volume mounted on Node 1 Network Path Connectivity Failure VHD Fault-Tolerant TCP connections make a path failure seamless

demo Cluster Shared Volume Resiliency

Performance Improved CSV Performance

Improved CSV I/O Performance

demo Cluster Shared Volume Block Cache in action

CSV Optimisations Direct I/O Redirected I/O • Block level I/O performance parity • Remote file system (SMB) performance parity

High Performance fault condition I/O redirection Share Node 1 VM Share Coordination Node VM Node 2 CSVFS Server / SMB CSV Proxy File System MUP/RDB SS/SMB CSV Proxy File System Avoids traversing file system stack twice CSV File System Filter NTFS CSV Volume. Mgr 2 x performance over File System redirection Volume Manager CSV Volume. Mgr LBFO/RDMA Disk Storage Connection Broken or not present

Inherit gains for CSV redirection performance SQL TPC-C transactions/s 4270 (98%) 4315 (100%) 4500 4000 3500 3000 1175 (27%) 2500 2000 1500 1000 500 0 SMB + DAS

SMB 3. 0 integration VM with I/O being redirected 10. 10. X 20. 20. X CSV Streaming I/O Across Multiple Networks VHD Coordinator Node

CSV Redirected mode enhancement summary Many Pieces Come Together for Radical Improvement I/O redirection needed less often Improved SMB performance ~2% off block Direct (RDMA) CSV Block Level Redirection SMB multi-channel

Deployment CSV Deployment Considerations

The requirements to enable a disk to be CSV

Deployment considerations

Backup CSV Backup

Key Takeaways INFRASTRUCTURE as a SERVICE

Breakout Sessions (session codes and titles) VIR 312 - What's New in Windows Server 2012 Hyper-V, Part 1 VIR 315 - What's New in Windows Server 2012 Hyper-V, Part 2 VIR 314 - WS 2012 Hyper-V Live Migration and Live Storage Migration WSV 326 - The Path to Continuous Availability with WS 2012 WSV 332 - Cluster-Aware Updating and the New Generation of WSUS Find Us Later At the ‘Speaker Lounge’ (12: 45 – 13: 45)