IT operations teams are exploring new IT analytics tools like Splunk to help them improve efficiency and align...
themselves more closely with business operations. Juniper Networks Inc. is answering this trend with the release of its new Cloud Analytics Engine, a product that collects and crunches data flowing from its network equipment, as well as applications running on servers.
"More and more of our customers are asking us, 'How can I get visibility into the application layer and how can I correlate that to the statistics that the network is feeding me?" said Calvin Chai, Juniper's director of data center marketing.
Most network management tools that collect data from network devices -- such as ingress and egress interface statuses, buffer utilization, hop statistics and time stamping -- compile that data and keep it isolated in a query-based system. Juniper's Cloud Analytics Engine combines that network machine data with data taken from applications, and provides analysis and visualization of that combined information, Chai said.
Visibility across networks and virtualized servers is always a challenge, said Rohit Mehra, vice president for network infrastructure research at IDC. Cloud Analytics Engine "provides correlative visibility into the physical and virtual," he said. "We have analytics platforms that do a good job with virtual infrastructure and we have platforms that do a good job with physical network infrastructure. This is one place that provides that correlative visibility across both. But, it also has API integration into SDN and network virtualization platforms."
Many vendors have started responding to increasing demand for improved IT analytics. Performance management specialist Extrahop Networks Inc., for instance, has integrated its machine data feed with Splunk and some open-source analytics engines.
Juniper's Cloud Analytics Engine has four basic components:
1. A network device agent. Engineers can deploy this agent on any device running Juniper's Junos operating system. In the first phase of the IT analytics product's rollout, Juniper will support this agent on its QFX 5100 switches. Future releases will expand to other Juniper products.
2. A compute agent. This agent can deploy on both bare-metal and virtualized servers to expose data from the server and its resident applications. This agent supports KVM, VMware ESXi and Microsoft Hyper-V hypervisors. The Juniper Contrail vRouter will also support this agent, giving visibility into Juniper's network virtualization overlay.
3. A data learning engine. This IT analytics engine works as a central controller and an aggregation point for data forwarded by the network and compute agents. The data learning engine is fully integrated with Juniper's Junos Space Network Director.
4. Visualization software. Engineers can use this software to identify unexpected connections between application and network machine data.
"This allows you to correlate data from the network itself from an end-to-end view, and tie that in with host information from the virtual machines," Chai said. "You can set up probes that mimic application behavior. Once you tie all this together, you are able to get more fine-grained visibility into what's actually happening. As applications become more dynamic and move around data centers and cloud environments, you want be able to tie all these bits of information together."
Juniper's analytics tool will aid with troubleshooting, performance management and capacity planning, Mehra said. And, since the technology is integrated into Junos Space Network Director, engineers can identify issues in the analytics tool and make changes to the network from within the same environment.
The various components of the Cloud Analytics Engine are connected via REST APIs and the data model it uses is based on an open schema, which will allow engineers to integrate the engine with their own tools. For instance, an operations team could replace Juniper's visualization software with its own preferred tool. And, it's possible Juniper could integrate the engine with other SDN and network virtualization technologies through these same APIs, IDC's Mehra said.