Design recommendations for metro node
General best practices for the Dell metro node back-end
On converged systems, take into account that:
- Each metro node director must have redundant physical connections to the back-end storage array across dual fabrics.
- Metro node systems require that 100 percent of all front-end and back-end ports are connected.
- Metro node does not support vVols.
- Moving the ESXi host boot LUN behind metro node is not supported. Boot from SAN LUNs needs to be provided to the host directly from the storage array, and not from metro node.
Zoning guidelines and examples
Metro node zoning is based on the following design principles:
- Zone metro node director A ports to one group of four storage array ports.
- Zone metro node director B ports to a different group of four storage array ports.
- Each metro node director is treated as a VMware ESXi cluster (host). Follow the host connectivity guidelines in the Tier-3 Logical Build Guide. Look for the document on the Dell Support website. In a metro node environment, the same array ports are zoned to metro node director-A, with two active paths per fabric. The storage array ports that are reserved for the second VMware ESXi cluster are zoned to metro node director-B.
- The even back-end port (IO-02) connects to Fabric A. The odd back-end port (IO-03) connects to Fabric B.
See the Dell Integrated Data Protection Zoning for metro node spreadsheet for back-end zoning examples with Dell UnityXT and PowerStore arrays. To obtain a copy of the spreadsheet, contact your Dell Technologies sales representative.
Array-specific best practices for the Dell metro node back end
Adhere to the following best practices for specific storage arrays.
As an ALUA array, the Unity XT array requires the four active and four passive connectivity rules. Each director node of the metro node cluster must have four active and four passive paths to all LUNs.
Each metro node director has two back-end ports. For compliance, each metro node director back-end port must be zoned to two ports on each Unity XT storage processor (SP). The SP that owns the LUN provides the active paths to the LUN, while the second SP provides the passive paths.
When configuring back-end connectivity with Unity and Unity XT storage arrays, ensure that:
- Each metro node director has logical and physical connectivity to storage controllers.
- The minimum supported configuration is eight paths (four active and four passive) per metro node director for any given LUN.
For more information, see the Dell EMC Integrated Data Protection Zoning for metro node spreadsheet on the Dell Technologies Support website.
When configuring back-end connectivity with PowerStore storage arrays, ensure that:
- Each PowerStore node has two IOM slots.
- Each node has at least one FC IOM.
- The even FC ports (0 and 2) are cabled to the Cisco MDS switch A fabric.
- The odd FC ports (1 and 3) are cabled to the Cisco MDS switch B fabric.
- Each metro node engine is zoned to a port group on each appliance: FC00 is zoned to two Fabric A ports, and FC01 is zoned to two Fabric B ports.
- Each metro node engine is zoned to all appliances in the array cluster using the same FC ports. Each LUN is owned by one appliance, enabling LUN migration between appliances in the same array cluster.
- Each metro node director must have logical and physical connectivity to storage controllers.
- The minimum supported configuration is four paths per metro node director for any given LUN.
General best practices for the Dell metro node front end
When configuring front-end zoning for metro node on converged systems, ensure that:
- The front-end I/O module on each metro node director has a minimum of two physical connections, one to each fabric.
- Even port IO-00 connects to Fabric A, odd port IO-01) connects to Fabric B.
- Each host has at least one path to metro node director-A and one path to metro node director B on each fabric, for a total of four logical paths.
- Four paths are configured from each host to the metro node cluster to complete an NDU.
- PowerPath provides load-balancing and failover on converged systems.
- PowerPath adaptive policy is used for non cross-connect configurations on converged systems.
- All VMware ESXi host initiators are registered with a Host Type of default.
Metro node WAN-COM connectivity
In a metro node configuration:
- Intracluster connectivity refers to director-to-director communication within the metro node cluster.
- Inter-cluster connectivity refers to communication between the metro node clusters.
IP WAN-COM connectivity
Connect the IP WAN ports to the Cisco Nexus 9000 switches. Take into account that:
- Each of the four metro node directors has two WAN ports configured into two port groups, WC-00 and WC-01.
- Independent WAN links are strongly recommended for redundancy.
- Two separate non-vPC VLANs for the WC-00 port group must be defined on Switch A and for the WC-01 port group on Switch B.
- IP WAN-COM ports cannot participate in a port channel, so attaching to a member switch in a vPC pair results in orphan ports. The vPC peer-link between Cisco Nexus switches in a converged system is not sized to account for metro node Metro traffic.
- Separate uplinks are configured for the non-vPC VLANs. You can attach IP WAN-COM ports directly to the customer's dark fiber network switches.
- As optical 10 Gbps ports, the IP WAN-COM ports do not automatically negotiate down to slower speeds and must connect to 10 Gbps SFPs. Provide 10 Gbps connectivity locally where the IP WAN-COM ports attach to the network. Other network segments can run at speeds other than 10 Gbps.
- All WC-00 ports are assigned to the same VLAN in each site.
- All WC-01 ports are assigned to the same VLAN in each site.
- The IP WAN ports do not support 802.1Q (VLAN) tagging. Configure the switch interfaces as access ports.
- The MTU size attribute affects IP WAN-COM performance.
- The default IPv4 MTU size for network switches is 1,500.
- Increasing the size of the MTU increases performance over the WAN.
- Metro node supports a maximum MTU size of 9,000. Use the highest MTU size that is supported on the network.
- Every network switch in the path between the metro node clusters is configured to support jumbo frames. Otherwise, the frame is fragmented into multiple smaller frames with an MTU of 1,500, which negatively affects performance.
- The supported network round-trip latency is less than or equal to 10 milliseconds.
- The correct socket buffer size (socket-buf-size) is set for your anticipated workload.
To start your base lining process, use the following socket buffer size values:
- 1 MB for an MTU of 1,500 with an RTT of 1 millisecond
- 5MB for an MTU of 1,500 with an RTT of 10 milliseconds
- 5 MB for an MTU of 9,000 with an RTT of 1 millisecond and 10 milliseconds
For information about changing the MTU size or the socket buffer size, look for the Metro node best practices document on the Dell Technologies Support website.
Metro Cross-Connect best practices
When configuring a host Cross-Connect for metro node Metro, take into account that:
- The maximum round-trip latency for Cross-Connect is 1 millisecond with all metro node Metro configurations.
- The maximum round-trip latency for Cross-Connect is 5 milliseconds with VMware vSphere.
- Metro node Witness server is mandatory for Cross-Connect.
- Dell Technologies does not support Cross-Connect for converged systems that do not include Cisco MDS FC switches. This is referred to as the separated networking option.
- Host Cross-Connect requires front-end SAN connectivity between converged systems. This ensures that hosts in the primary converged system can communicate with front-end ports of the metro node cluster that is connected to the secondary converged system, and conversely.
- PowerPath/VE must be configured with active paths to the local metro node cluster.
- PowerPath/VE provides an auto-standby feature that groups logical paths by the metro node cluster and determines which has the lowest path latency to identify local and remote.
- PowerPath/VE autostandby must be enabled for each VMware ESXi host in a Cross-Connect configuration.
Autostandby with the proximity trigger determines and selects the preferred paths to a distributed volume and places nonpreferred paths in autostandby mode (asb:prox).
Here is the syntax of the command to enable autostandby: rpowermt set autostandby=on trigger=prox host=x.x.x.x
- The auto-resume feature must be set to true for all metro node Metro CGs.
- The thin-rebuild feature must be set to true for all metro node Metro distributed volumes.
- Cross-Connect zoning follows the rules that are defined in metro node front-end (FE) logical port groups.
Storage array cross-connect (back-end SAN connectivity) is sometimes implemented for adding protection to system volume