*Edited information from P4000 Multi-Site HA/DR Solution Pack user guide
The Multi-Site SAN features enable synchronously and automatically mirroring data between geographic sites.
1: Designing a Multi-Site SAN
Multi-Site clusters can:
• Span up to three sites
• Span multiple subnets
• Have multiple VIPs
• Geographical awareness enabled by designating storage systems as members of a particular site.
• Synchronously replicated (mirrored) data between sites, based on volume data protection level.
• Site information that ensures that data is mirrored between sites for high availability and disaster recovery.
• I/O path preferencing, so that application servers connect to storage systems located in the same site.
• Failover Manager support for automatic failover/failback in three-site configurations without requiring a physical storage system in the third site.
• Failover Manager for quorum management if the network connection between the two sites becomes unavailable.
• A feature key for each storage system in the management group that is also participating in a Multi-Site cluster
• All sites must have the same number of storage systems
3: Protection Offered
Common types of data center failures a Multi-Site SAN protects against include:
• Site power outage
• Site network outage
• Site disaster
Data protection level Number of sites supported:
Network RAID-10 → 2 sites supported
Network RAID–10+1 → 3 sites supported
Network RAID–10+2 → 2 sites supported
Network RAID-10+2 offers a level of data protection that enables the Multi-Site SAN to sustain a complete failure at one site and still maintain replicated data at the other site. With Network RAID-10+2 there are four copies of data for the volume on the SAN.
Table → Common configurations of Multi-Site SANs and managers
4: Designing the network for the Multi-Site SAN
• Adequate bandwidth – plan for 50 MB/sec of bandwidth for each storage system in each site (example – if each site contains 4 storage systems, then need 200 MB/sec throughput which translates into two Gigabit Ethernet links (125MB/sec) or more.)
• Low round-trip latency – in order to not impact disk I/O to the application server, the round-trip latency between the sites must be no more than 2 ms (which implies a theoretical maximum distance between sites of 299km)
• Redundant links – have multiple physical connections (media) between the sites for redundancy (the network should be configured so that a link failure does not cause the Multi-Site SAN to go offline.)
• Assign servers to sites – avoid high-latency connections by assigning application servers to the site where the server is
Commonly used designs:
• Dual redundant links between the sites
• Full-mesh triangular (three) redundant links between sites
• Hub and spoke central network core with redundant links out to each site
• Combination full-mesh core network with triangular (three) links between sites
Using multiple subnets:
Can use multiple subnets with a Multi-Site SAN. Multiple subnets let you have multiple VIPs, one VIP per subnet. The advantage of using multiple VIPs is that you can connect to a volume without having to leave the local subnet. Alternatively, you can assign an application server to a site.
If using multiple subnets within the iSCSI cluster and implementing one subnet per site, can take advantage of the following Multi-Site SAN features:
• One Virtual IP address per site – iSCSI discovery sessions are terminated on a storage system that is local to the application server in that site.
• Virtual IP addresses are hosted only on storage systems that reside in the same subnet as the VIP
• Read and Write I/O requests are forwarded from the application server to a storage system that is in the same site as the server. This ensures that the I/O path is optimal and multiple network hops across the network link between the sites are eliminated.
5: Other considerations
1: Stretched vSphere HA clustering
2: Stretched subnets spanning the sites