After nearly 15 years of deploying HPE ML350 and DL380 servers, we're moving to Lenovo because of our growing concerns with HPE as a company. Our first Lenovo build is an SR650 V4 with 2x Xeon 6517P CPUs, 256 GB RAM, four 3.2 TB U.2 mixed-use NVMe drives intended for RAID 10, and two 480 GB NVMe drives in a PCIe boot device.
Lenovo recommended VROC Premium with U.2 drives instead of a hardware controller and U.3 drives. Their reasons were that hardware RAID is outdated and the VROC configuration is less expensive. I'm not convinced, since traditional RAID controllers have always been straightforward to replace or upgrade, and I want to understand how recovery would work if the motherboard or a CPU failed.
Lenovo has not clearly explained where the VROC metadata is stored, whether the array can be moved to replacement hardware, or how well it is monitored through XClarity. I'm also concerned about driver support and reliability under Windows Server with Hyper-V. The VROC examples I've found often involve lower-end systems, so I'm wondering whether the faster CPUs in this server make a meaningful difference.
Would you choose VROC Premium for this configuration, or switch to a Lenovo 940-series hardware RAID controller with U.3 NVMe drives? Reliability, monitoring, driver support, performance, and recovery after hardware failure matter more to us than saving on the initial purchase.
2 Answers
VROC metadata is stored on the drives, so a motherboard replacement does not automatically erase the array. However, portability is not necessarily as simple as moving disks to another system: compatible Intel platforms, firmware, drivers, and VROC versions still matter. Drive firmware can also affect stability. The fact that the configuration uses powerful CPUs does not remove those operational and support concerns.
I have seen VROC work successfully with Hyper-V, including management through Lenovo's tools, but I have also encountered difficult failures. In one case, restoring a VM that had previously run on another hypervisor caused two drives to be marked failed and the array to drop. The drives could be marked safe and rebuilt, but the problem returned when the same driver behavior occurred again. That kind of edge case would make me choose a hardware controller for an important host.
That is exactly the sort of recovery scenario I'm worried about. Even if VROC performs well normally, a hardware controller seems easier to diagnose when something unusual happens.