by Kevin OBrien

EMC Unified Storage VNX Series Hands On [VNX 5300]

Most storage solutions like the EMC VNX 5300 come to our enterprise test lab for review. On occasion though, we get an opportunity for hands-on time that requires us to bring our lab to on-site. That was the case last week as we spent hands on time with a pair of VNX 5300 unified storage systems. The VNX 5300 is one of EMC's mainstream unified storage offerings that are tuned for virtual applications with VMware and Hyper-V integration. The VNX 5300 can be provisioned with up to 125 disks; combining SAS, nearline SAS and flash storage. Like most of the VNX family, the 5300 supports FC, iSCSI and FCoE block protocols and NFS, CIFS, MPFS and pNFS file protocols.

For this on-site mini-review we brought along a Lenovo ThinkServer RD240 equipped with a Mellanox ConnectX-3 EN Dual-Port 10GbE adapter and our Mellanox SX1036 36-port 40GbE switch.

StorageReview On-Site Lab Specifications:

  • Lenovo ThinkServer RD240
    • 2 x Intel Xeon X5650 (2.66GHz, 12MB Cache)
    • Windows Server 2008 Standard Edition R2 SP1 64-Bit
    • 8GB (2 x 4GB) 1333Mhz DDR3 Registered RDIMMs
    • Mellanox ConnectX-3 Ethernet Dual SFP+ Port Adapter
  • Mellanox SX1036 36-port 40GbE Switch
    • 36 40GbE ports in 1RU
    • Up to 64 10GbE ports
    • Up to 2.88Tb/s of non-blocking throughput

For the purposes of this test, we had one port on the switch configured in fan-out mode to convert a single 40GbE connection into four 10GbE connections. Flow-control was enabled on the ports we were interfacing with, as well as an MTU of 9000.

The EMC VNX 5300 we evaluated is set up with the following storage configuration:

  • 4 600GB 15K SAS (Array Software)
  • 11 600GB 15K SAS + 8 2TB 7200RPM SAS (RAID 5 Storage Pool)
  • 3 100GB Flash SAS (RAID1 + hotspare, FAST-VP)

While not a complete in-depth review, we did spend enough time with the VNX system to run workloads against it. Keep in mind this particular VNX 5300 was in active use, so our results show what happens even with dozens of data stores operational. To keep things fairly simple, we applied a static QD128 load using IOMeter with our Database, Workstation, File Server, and Webserver profiles. The database profile includes a 67% read and 33% write workload mix primarily centered on 8K transfer sizes. The Workstation also uses 8K transfer sizes, but has less write activity and more sequential access. The File Server workload has a spread of transfer sizes ranging from 512-byte to 64KB, with 80% read and 20% write. The Webserver profile has a transfer spread of 512-byte to 512KB, and is read only.

In this scenario we were connected to two 100GB LUNs, with one 10GbE iSCSI connection per LUN. The speeds shown below are the aggregate of two workers individually accessing their respective LUN. These tests show sustained burst performance, not long-term endurance performance from what would be found in our standard in-depth reviews.

It is interesting measuring performance on a production platform, outside of our normal testing environment where equipment is clean of external activity. For many users, these conditions are more representative of the real world, where you have many servers or virtual machines accessing the same storage pool. In our configuration with a bit of Flash Cache and a mixed RAID5 disk pool consisting of 7,200RPM and 15,000RPM SAS drives, the EMC VNX 5300 offered very good performance, with low average latency in these production-environment tests.

As StorageReview continues to evaluate both custom storage arrays and high-end storage array offerings like this EMC VNX solution, the site editors endeavor to bring deeper array testing and evaluation to our readers. 


Discuss This Story