ESG's Mike Leone reviews the results from Esg Lab's recent testing of hyper-converged Infrastructure with Dell EMC ScaleIO software-defined storage.
Announcer: The following is an ESG Lab video.
Mike: Welcome to another ESG Lab video summary, where I'll be reviewing the results from a recent testing of hyper-converged infrastructure with Dell EMC's ScaleIO software-defined storage.
In order to meet the challenges of doing more with less, many organizations have turned to modern data center architectures, such as hyper-converged infrastructures. ScaleIO from Dell EMC is a software-defined, distributed, block storage system that converges storage and compute resources on physical servers, forming a single layer, hyper-converged architecture. This simplifies management while still enabling the solution to scale over a thousand nodes.
ScaleIO can be deployed on industry-standard servers, on pre-optimized Dell EMC ScaleIO-ready nodes, or on a fully integrated VCE VxRack FLEX solution.
We evaluated ScaleIO on a test bed that demonstrated the typical use of ScaleIO as an all flash, hyper-converged cluster, with each physical server in the cluster being used both for storage services and as a general purpose hypervisor host for guest virtual machines.
We used the vSphere plugin to enter the required configuration information and ScaleIO automatically installed across the cluster, aggregating the internal disks into shared pools of storage. The whole process took less than an hour.
We used both the vSphere plugin and ScaleIO's standalone GUI to provision volumes and map them to hosts. The GUI dashboard provides the user with a wealth of status, configuration and performance information, with drilldowns to get to detailed component level data.
For the benchmarks, application response time was restricted to 1 millisecond in order to show how much IO can be driven at very low latency, which is important for sensitive OLTP workloads. Baseline FIO performance tests show that a 4-node cluster sustained more than 429,000 read IOPS. After scaling up, in 8-node clusters, sustained more than 653,000 IOPS, a 46% gain, demonstrating linear performance increases when adding more nodes. For mixed read to write workload that mimics common OLTP transactions, the 8-node cluster showed a steady performance of more than 408,000 IOPS. Again, the tests were configured to measure sustained IOPS with latency kept under one millisecond, the industry standard performance threshold for all flash storage systems. In an additional test, however, we took brakes off a little, and the 8-node cluster impressively sustained a 906,000 IOPS with a response time of just 1.46 milliseconds.
Using the widely adopted Silly Little Oracle Benchmark kit, to simulate a real-world OLTP environment, the 8-node cluster sustained over 562,000 read IOPS and 351,000 IOPS for mixed read to write workloads. This demonstrates the ability of the ScaleIO solution to simultaneously process storage transactions and run complex demanding database applications.
An 8-node cluster running both the FIO and the Oracle benchmarks sustained over 424,000 IOPS. This compares favorably with the 409,000 sustained IOPS we saw when running only FIO and demonstrates that multiple heavy application loads do not impact storage system's performance.
But what happens when you lose a server in a ScaleIO cluster? We simulated a catastrophic node failure with both FIO and SLOB running on an 8-node cluster. When there are disks or node failures, ScaleIO automatically rebuilds the copies and rebalances data on the remaining devices. And we specified, "Rebuild throughput limits of 140 megabytes per second."
After simulating the loss of one of the eight nodes during the rebuild, the system was able to sustain 391,000 IOPS, a mere 8% reduction from the normally operating clusters' 426,000 IOPS, while response time only increased from 0.9 milliseconds to 1.27 milliseconds.
Faster storage provisioning and increased simplicity of implementation and management alleviate the administrative burden, enabling IT to focus on optimizing performance efficiency and reliability. Customers gain peace of mind knowing they can consolidate workloads onto ScaleIO-based hyper-converged infrastructure with confidence in its sustained performance and knowledge that it will scale with their capacity and workload growth needs.
Integrating the latest and greatest storage, networking, and server technologies into a flexible architecture for multiple enterprise workloads can be a significant challenge. As a hyper-converged infrastructure, ScaleIO from Dell EMC takes the pain out of the equation, enabling organizations to deliver performance, reliability, availability in an optimal user experience from a single solution. ScaleIO is ideally suited for database infrastructure, consolidation, and virtualization at organizations ranging from the smallest businesses to the largest enterprises.