ESG Test Report Huawei OceanStor Dorado V3 Delivers Excellent Performance for Mainstream All-Flash Applications
The ESG lab is a well-known IT consulting organization based in the United States. It performs IT analyses, research, and verification, and is renowned for its independent, rigorous, and practical laboratory tests.
In September 2017, the ESG lab tested the performance, reliability, cost-effectiveness, and total cost of ownership (TCO) of Huawei OceanStor Dorado V3. The ESG set up a real service environment and tested the performance and reliability of Huawei all-flash storage for applications such as VMs, databases, and mailboxes.
Figure 1 shows the test environment used in the ESG lab. The VMware vSphere 6.0 virtualization infrastructure is deployed on 10 Huawei FusionServer RH series rack servers, with two dual-controller Huawei OceanStor Dorado5000 V3 storage systems connected to the back-end.
According to the ESG lab test data, Huawei Dorado V3 all-flash storage delivers a consistent latency of 0.3 ms under the pressure of 5000 Exchange users, 1000 VMware Horizon virtual desktops, and hundreds of Oracle users. Even if storage devices are faulty, upper-layer applications can automatically recover within 3 seconds by using the active-active solution, ensuring service continuity. In addition, the data reduction technology reduces the five-year total cost of ownership (TCO) by 75%, making Huawei Dorado V3 an ideal choice for mission-critical workloads.
Stable High Performance
The ESG lab has verified that the OceanStor Dorado V3 can integrate the most challenging workloads and mission-critical tasks into a high-performance and high-availability platform. When 5000 Exchange users, 1000 VMware Horizon virtual desktops, and hundreds of Oracle users are integrated into one OceanStor Dorado5000 V3 system, the integration environment has an average response time of 0.3 ms for these services.
Figure 2 shows the test results of five thousand 1 GB mailboxes and 9.6 TB Oracle OLTP databases. In this case, the system supports about 105,000 continuous IOPS, while maintaining an average response time of just 0.3 milliseconds.
In the next test, personnel simulated a boot storm of 1000 virtual desktops. The I/O soared to about 160,000 IOPS in approximately one minute. At this time, the peak response time increased but remained lower than 0.5 ms. Shortly after this, the two indicators returned to the same level as before the boot storm (see Figure 3).
In addition, the ESG lab created multiple snapshots and then rolled these back, which did not affect system performance. Deduplication and compression are always in the enabled state, and the deduplication and compression ratio is 7:1 for the three types of hybrid workloads. Together with Thin Provisioning, this reduces the total space required by 14:1.
Huawei OceanStor Dorado V3 all-flash platform ensures high availability and stable performance during planned maintenance and unexpected interruptions. The HyperMetro feature provides a 100-kilometer, gateway-free, active-active solution to ensure service continuity. RAID-TP technology ensures services continue to run stably even if three disks in a single disk array fail.
In the ESG lab, HyperMetro is first tested in a campus environment. Two OceanStor Dorado V3 storage systems are configured with HyperMetro and run the Oracle RAC database to simulate OLTP transaction database traffic. The total traffic generated by the two servers is 50000 IOPS, the processing volume of a single disk array is 25000 IOPS, and the average I/O response time is 0.25 ms.
The ESG simulated the failure of array 1 by removing its power supply, reducing the total IOPS of array 2 to 0. After about 3 seconds, the total IOPS of array 2 was restored to provide services for all requests in the cluster. A total of 50,000 IOPS are supported, and during the switchover the maximum latency on the controller of array 2 was 4 ms and 8 ms. The latency was reduced to 0.25 ms after 3 seconds (see Figure 4).
The ESG also tested the performance of the Dorado RAID TP technology when three disks are faulty. Three SSDs were removed from the disk array at intervals of one to two minutes to simulate the simultaneous failure of three SSDs. The disk array was still available after three SSDs were shown to be faulty without significantly reducing performance (see Figure 5).
The RAID-TP reconstruction time was also tested. The ESG lab removed an SSD from a disk array that was configured with nine SSDs and 7.4 TB volumes. It took just nine minutes for the RAID-TP to reconstruct the 7.4 TB volume. There is a 7.4 TB capacity and there are nine SSDs, while the reconstruction data volume of each SSD is about 822 GB. Therefore, the reconstruction time of 1 TB of data was only 11 minutes.
The ESG lab simulated the expected storage TCO of a company that required a high availability hybrid service environment. The OceanStor Dorado5000 V3 has an obvious advantage in this area when compared with Huawei's hybrid storage arrays and the first-generation all-flash storage devices of other mainstream vendors. The five-year TCO was shown to be reduced by more than 75% (see Figure 6).
The ESG lab acknowledges that Huawei OceanStor Dorado V3 provides stable high performance and high reliability for its users. Huawei OceanStor V3 perfectly adapts to high-performance service applications and can run in a virtualization environment with high performance requirements.