One of the primary goals of the Storage and Availability team at VMware is to validate business critical application running on Virtual SAN continuously. It is of the utmost importance for us to deliver the necessary information to customers, so they feel confident about the storage platform they are considering or plan to use to run their business critical applications. At the same time, we want to provide the necessary data points for them to understand how the platform will deliver the capacity, performance, and availability services that are demanded by their applications.
Below is a sample of the information that can be found on a performance study performed on SAP IQ, a mission critical application running on VMware Virtual SAN.
SAP IQ is an intuitive, cost-effective, and highly optimized RDBMS that is fast and efficient for extreme-scale data warehousing and big data analytics. SAP IQ is a distributed application with multiplex nodes, which may have different roles with different capabilities. This is unlike other database cluster architectures, which usually follow either a shared-everything or shared-nothing architecture. The multiplex server configuration can be described as an “asymmetrical cluster” One node is designated as the coordinator; the remaining nodes are query nodes and may be either Readers or Writers. In addition to its role of handling transaction management, the coordinator can also serve as a Reader or Writer in the multiplex.
Distributed Query Processing uses the available memory and CPU resources of all available nodes to process queries. Performance is therefore determined by the overall workload in the cluster as a whole at any given time. In a single run of a long-running query, the work distribution may change over the course of a query execution as the load balance changes across worker nodes.
The node at which a query is submitted becomes the leader node for that query, and the remaining nodes assume the worker role. Any node in the cluster can be the leader node; similarly, a worker node is any node that is capable of accepting distributed query processing work. Work is performed by threads running on both the leader and worker nodes, and intermediate results are transmitted between nodes by one of two mechanisms: through a shared disk space, or over an inter-node network.
The goal of this study is to showcase that vSphere and VMware Virtual SAN are a suitable platforms to deploy distributed and multi-tier applications. Testing showed that SAP IQ — a distributed database platform— virtualizes well by putting individual nodes of a multiplex server on vSphere virtual machines, which are supported by a virtual storage area network (Virtual SAN) that provides a robust backend to ensure all data is accessible from all nodes. The image below illustrates a logical representation of the test environment utilized for the study.
Note: This performance study was performed VMware vSphere 5.5 with Virtual SAN 5.5. The performance results would be greatly improved from a performance standpoint on the latest version of Virtual SAN 6.x. The performance improvements is an assumption based on the performance improvements made in the new release.
The performance characteristics and capabilities of Virtual SAN along with its ease of deployment make it an excellent storage platform for deploying distributed database applications using SAP IQ Multiplex Server.
For all the details and information on the performance study, download the white paper from the link below:
For future updates on Virtual SAN (VSAN), vSphere Virtual Volumes (VVol) and other Storage and Availability technologies, as well as vSphere Integrated OpenStack (VIO), and Cloud-Native Apps (CNA) be sure to follow me on Twitter: @PunchingClouds