High availability (HA) is crucial for maintaining the reliability and continuous operation of critical systems and services. By incorporating redundancy, fault tolerance, and minimizing both recovery point objective (RPO) and recovery time objective (RTO), HA solutions contribute to a robust and resilient infrastructure, reducing the impact of potential failures on business operations. IBM®HyperSwap® caters the need of HA for enterprises.
What is HyperSwap?
HyperSwap is an active-active cluster solution that ensures continuous operations by enabling seamless failover between storage systems. It guarantees HA and data redundancy which is critical for business continuity and minimizing downtime.
This blog explains the deployment configuration of HyperSwap in Ethernet data centers.
IBM HyperSwap in Ethernet environment does not require Fibre Channel over IP (FCIP) routers to make it a cost-effective solution.
Possible configurations of IBM HyperSwap in Ethernet environment include:
· Shared inter-switch link (ISL)
· Dedicated ISL
Refer to the following general guidelines and recommendations for Ethernet-based IBM HyperSwap:
· The maximum allowed round-trip time for inter-node connectivity is up to 3 milliseconds.
· For optimal performance and reliability, the backend storage systems must be interconnected with a minimum of two nodes to ensure HA.
· The minimum required connection bandwidth is more than 50 megabits per second (Mbps).
· If utilizing Layer 2 network connectivity, it is advisable to implement virtual LANs (VLANs). This requires configuration adjustments on both, the switch and node sides. L3 switches and routers are recommended for long-distance ISL-based networks because they are equipped with deeper buffer pockets.
· A minimum of two dedicated Remote Direct Memory Access (RDMA) capable Ethernet ports are required for inter-node communications to ensure best performance and reliability.
· If it is intended to use a shared switching fabric, customers would need to ensure that quality of service (QoS) guarantees that inter-node traffic is not affected by any other traffic present in the same switching fabric.
Refer to the following flowchart to find out which deployment configuration is suitable for the customer environment:
Notes:
· IBM HyperSwap in Ethernet environment configurations are qualified for general availability for iWARP connectivity with release version 8.5.2.0
· IBM HyperSwap in Ethernet environment configurations are not qualified for RoCE connectivity. Contact IBM Support for RDMA over Converged Ethernet (RoCE) configuration.
· Before deployment of Ethernet data centers, you need to thoroughly plan how many RDMA-capable Ethernet adapters are needed to support the data center that includes inter-node and host-to-node connectivity.
· Ensure that each switch in your network configures the VLAN on the switch port to Trunk mode. Also, specify the VLAN ID for RDMA ports to align with the corresponding VLAN on the switch side port.
· The gateway is not utilized for HyperSwap internally; it merely serves as a parameter for internal network configurations.
· Refrain from setting the gateway as 0.0.0.0 using the chnodeip command as this action is invalid.
· Users can disable the Link Layer Discovery Protocol (LLDP) on the switch ports used.
References:
· IBM Spectrum Virtualize inter-node communication support over Ethernet-based RDMA https://www.ibm.com/downloads/cas/WQAYR4XE
· Deployment of Active - Active High Availability system using Ethernet RDMA https://www.ibm.com/downloads/cas/8AG67JVY
· Ethernet-based IBM HyperSwap: Cost-effective, active-active high availability solution over converged long-distance links https://community.ibm.com/community/user/storage/blogs/abhishek-jaiswal1/2020/09/02/ethernet-based-ibm-hyperswap-converged-links
About authors
@Shrirang Bhagwat is architect in the IBM Flash system team at IBM Systems Development Labs, India. He specializes in block storage for Ethernet environment. He can be reached at: shbhagwa@in.ibm.com.
@Abhishek Jaiswal is development lead in the IBM Flash system team at IBM Systems Development Labs, India. He has expertise in block storage and working on the Ethernet mission for the IBM Spectrum Virtualize product. He can be reached at: ajaiswa9@in.ibm.com.
@Veerendra Risbud is system software developer in IBM Flash system team at IBM Systems Labs, India. He can be reached at: Veerendra.Risbud@ibm.com
@Harshit Jain is system software developer in IBM Flash system team at IBM Systems Labs, India. He can be reached at: harshit.jain7@ibm.com
#Featured-area-2#Highlights#Highlights-home