Huawei Dorado 3000 V6 Benchmark on Proxmox VE

Storage performance often makes the difference between a sluggish virtual environment and a high‑performing one. In this benchmark, I tested the Huawei OceanStor Dorado 3000 V6 — an all‑flash enterprise storage array — within a Proxmox VE 9.0.4 virtualized environment. The goal? To compare NFS, iSCSI, and iSCSI Multipath (MPIO) configurations.

The results? iSCSI MPIO crushed it — delivering over 500 000 IOPS with ultra‑low latency.

🛠️ Test Lab Configuration

To replicate a realistic data‑center setup, the following hardware and software were used:

📈 Benchmarking Methodology

To simulate both large data transfers and typical virtualized workloads, two fio profiles were used:

Sequential Read/Write (1 M block)

fio --rw=rw --bs=1M --iodepth=16 --size=10G --runtime=60 ...

Random Read/Write (4 K, 70/30 mix)

fio --rw=randrw --rwmixread=70 --bs=4k --iodepth=32 ...

📊 Performance Summary

📌 Highlight: MPIO achieved over 500 000 IOPS with sub‑0.1 ms latency — exceptional for virtual machines and database workloads.

📉 Visual Performance Chart

Side‑by‑side throughput & IOPS comparison of NFS vs. iSCSI vs. MPIO

🔍 Observations & Insights

NFS

• Offers decent sequential performance, but random I/O becomes a bottleneck.
• Caused by single‑threaded metadata operations.
• ✅ Best suited for archival or light‑use workloads.

iSCSI (Single Path)

• Nearly fully utilizes array performance.
• Low latency (< 1 ms) — ideal for production VMs.
• Minimal overhead even with Data Protection enabled.

iSCSI MPIO

• Utilizes both 25 GbE links simultaneously.
• Delivers enterprise‑class random I/O performance.
• Best choice for latency‑sensitive VMs and clustered databases.

🧰 Recommended Tuning

Multipath Configuration (/etc/multipath.conf)

defaults {
  user_friendly_names yes
  polling_interval 5
}
devices {
  device {
    vendor "HUAWEI"
    product "XSG1"
    path_grouping_policy multibus
    path_checker tur
    no_path_retry queue
    rr_min_io_rq 1
    rr_weight uniform
  }
}

NFS Mount Example

mount -t nfs -o vers=4.1,nconnect=8,noatime,rsize=1048576,wsize=1048576 <server>:/volume /mnt/test

Performance Optimization

echo none | tee /sys/block/sdX/queue/scheduler
echo 3 | tee /proc/sys/vm/drop_caches

✅ Conclusion

The Huawei Dorado 3000 V6 proves to be a powerful and efficient storage backend for Proxmox VE environments. While NFS is easy to set up, it’s not suitable for workloads with demanding random I/O. On the other hand, iSCSI — especially when paired with MPIO — unleashes the full potential of this all‑flash array.

Summary

• Top performer: iSCSI MPIO — over 500 000 IOPS with latency below 0.1 ms
• Good choice: iSCSI single path for smaller or less demanding setups
• Use with caution: NFS for workloads involving heavy random I/O

If you’re building a high‑performance virtualized infrastructure, Huawei Dorado 3000 V6 with properly tuned iSCSI multipath is a clear winner.

🔁 (Optional) fio Benchmark Script

#!/bin/bash
DISK="/dev/mpath0"
RUNTIME=60

fio --name=seq_rw --rw=rw --bs=1M --numjobs=1 --iodepth=16 --size=10G --runtime=$RUNTIME --filename=$DISK --direct=1 --ioengine=libaio --group_reporting --output=seq_rw.log

fio --name=randrw --rw=randrw --rwmixread=70 --bs=4k --numjobs=4 --iodepth=32 --size=10G --runtime=$RUNTIME --filename=$DISK --direct=1 --ioengine=libaio --group_reporting --output=randrw.log

Want to know how to set this up in your environment? Drop a comment or get in touch!