The challenge
A motor parts manufacturer with several branches across the US ran its virtual machines on VMware vSAN, where each cluster pooled the local disks inside its own servers. The hardware was aging, storage and compute were locked together, and every host event (a failure, a reboot, routine maintenance) became a storage event that left the cluster running with reduced protection while data rebuilt. Capacity was stranded in per-cluster islands, vSAN consumed CPU, memory and network in every host, and the platform was due for a full refresh. They needed a modern, resilient foundation delivered without disrupting production.
Our approach
- Assessment and design: documented the existing vSAN clusters, workloads and network, then designed a target architecture that separates storage from compute.
- Vendor-neutral selection: as a multi-vendor partner (NetApp, Dell and others), VeeMost engineered the design to fit the customer rather than a single brand; the customer selected the Dell PowerStore platform for the capabilities that fit their environment.
- Latest-generation servers: replaced the aging server hardware with current-generation hosts, reclaiming the CPU, memory and network capacity that vSAN had consumed.
- No-single-point-of-failure SAN: built an isolated 10G iSCSI storage network on two independent fabrics, giving every host four active paths to a dual-controller, active/active array; repurposed existing switches as dedicated SAN switches and stood up a new Cisco Nexus core.
- Non-disruptive migration: brought the new storage online alongside vSAN and moved VMs live with Storage vMotion, cluster by cluster, retiring vSAN on each cluster only after it passed acceptance tests that deliberately failed every switch, path and controller to prove resilience.
The solution
VeeMost refreshed the servers and moved the manufacturer off distributed vSAN onto a central Dell PowerStore array reached over an isolated, dual-fabric 10G iSCSI SAN. Storage now scales independently of compute as one shared pool across all clusters; a failed or serviced host no longer affects storage health; and routine maintenance such as switch firmware, controller service and cabling can be done one path at a time during business hours with no downtime. Production, storage and out-of-band management run as three networks that share no links, so no event on the LAN can reach storage. Every virtual machine was migrated live, with no interruption to production.
