Technological advances have paved ways for data analytics to become the essence of entire data center operations. It is now possible to provide application level integration by accessing meta-data from data center components.


Applying data for better performance

Data center efficiency as well as ability to augment problem resolution can be significantly improved by analyzing performance data that is being stored in large volumes by teams of data center employees.

Conceptually, the meta data generated by data center is submitted to a big data platform that can be expected to make recommendations to application or resource clustering software by analyzing the data. This can involve two different approaches as discussed in this post. We will touch upon the two distinct data analytics approaches adopted by CloudPhysics and Intel for performance enhancement of data center infrastructures of the future.

Use of detailed analytics by CloudPhysics

CloudPhysics is able to provide access to wide range of sophisticated attributes of vSphere including its advanced resource management facilities. However when it comes to actual implementation of automated resource placement, the customers are sometimes not able to execute the same with desired speed.

In principle, vSphere distributed Resource Scheduler (DRS) has ability to make sure that workloads are provided with physical resources that are essential for application performance. It is also observed that features such as DRS can be implemented with success if these are executed according to the best practices.

It is however not easy to implement the best practices since it can be prove to be an overwhelming challenge. In order to use the detailed analytics of data obtained from vCenter for the identification of performance trends, simplification of DRS configuration is achieved by comparing anonymized performance data to CloudPhysics best practices.

The detailed analysis of data procured from vCenter is used by CloudPhysics for identification of performance trends. In order to facilitate customers understand the performance of physical infrastructure according to data from CloudPhysics global clients, customers can also be empowered with custom reports.

CloudPhysics has maintained clear focus on cluster provider. It should be noted that vCenter has the distinction of being the only supported cluster provider.


Use of telemetry data by Intel

Intel is currently working on its new open source project Snap. It is designed to facilitate a broader focus and aims at providing telemetry data to cluster controllers including an application or to Mesos and Kubernetes.

‘Snap’ is being developed more as a telemetry framework than a platform for analytics as explained by Matthew Brender, developer advocate at Intel data center practice. It aims to provide a universal framework for analysis of low level raw data.

The commonest instance of telemetry data is the volume of L1/L3 data being consumed on a processor. You can get more in-depth information associated with data including CPU utilization from cache consumption details. This data is then fed to cluster management application so that the cluster manager can identify where to place the loads by analyzing the telemetry data.

It is being hoed by Intel that ‘Snap’ is able to take off as the basic framework. Unless there is a huge contribution from vendors, it is not possible for Snap to add value to cloud infrastructures. Knowledgeable technical teams can be expected to provide telemetry data from systems including interfaces to systems.

In conclusion

We can expect data analytics to grow in terms of significance as the data centers keep moving to cloud in order to sustain efficiency of data centers. The examples of CloudPhysics and Intel are from among a plethora of vendors that are intensely pursuing the use of data analytics in this domain.

Appropriate placing of workload is the key to greater performance of data centers that are moving to cloud. Proper application of data center analytics is going to be the most vital attribute of next generation infrastructures.