The following figure shows the logical architecture of the Oracle Big Data SQL environment. The figure includes the infrastructure component layers, the software layers, the management infrastructure, and associated tools. The infrastructure components include the HCI layer that supports compute and storage, the compute-only layer, and networking.
Figure 1. Solution architecture and components
The infrastructure layer consists of:
- PowerFlex HCI nodes: We deployed and configured four PowerFlex R840 nodes in the HCI configuration. These nodes contribute to compute and storage resources. These nodes, which were running the VMware ESXi 7.0 hypervisor, hosted four VMs running the Oracle Linux 7.6 guest operating system. We installed the Hadoop cluster and Oracle NoSQL on these VMs. These nodes also host four storage virtual machines (SVMs) which are installed during PowerFlex deployment. A Storage Virtual Machine (SVM) is a Linux-based virtual machine dedicated to PowerFlex OS that hosts different PowerFlex software components.
- Compute nodes: We configured two PowerEdge R640s as compute nodes as part of the PowerFlex infrastructure. These nodes, which were running the VMware ESXi 7.0 hypervisor, hosted two VMs running the Oracle Linux 7.6 and Windows Server 2019 guest operating systems. We installed Oracle 19c Restart and Microsoft SQL Server 2019 on these two VMs.
- PowerFlex Controller nodes: We configured three PowerFlex R640 controller nodes. These nodes, which were running the VMware ESXi 7.0 hypervisor, hosted three VMs. We deployed vCenter management, PowerFlex manager, and PowerFlex gateway VMs on these nodes. These VMs enabled the functions and management of the PowerFlex infrastructure.
- Each PowerFlex HCI and Compute node features:
- Two Mellanox ConnectX-4 LX 25 GbE SFP Adapters: Each adapter has two dual port 25 GbE network interfaces for storage data path, hypervisor management, vMotion, and the application VMs’ private interconnect
- One GbE integrated Dell Remote Access Controller (iDRAC): A dedicated iDRAC Ethernet port for out-of-band (OOB) server management
The network layer consisted of the following types of connectivity:
- A management network between vCenter and ESXi hosts. This network also connected the iDRAC Ethernet ports of all the servers
- A vMotion network for VM migration (using vMotion) between ESXi hosts
- A public network that connected various application VMs: Oracle, Oracle NoSQL, Hadoop, Microsoft SQL Server instances
- A private network that facilitated internode communication between Hadoop nodes
- A network that enabled PowerFlex storage data path communication
We implemented these networks using the following switches:
- One 1 GbE switch for management
- Two 25 GbE CISCO Nexus 93180YC-EX switches for application and storage networks
The storage layer consisted of the PowerFlex system as the storage for the different database applications, and for the data stores of the VM operating system volumes.