As more IT organizations begin to build out their own private clouds, many of them are struggling with I/O contention issues created when different classes of workloads are required to share access to the same storage. Storage administrators are tasked with manually trying to optimize the I/O path for each workload running in the cloud. To help with this issue, Kaminario today unfurled a cloud service that employs machine learning algorithms to identify precisely what I/O contention issues are arising.
Once identified, Chris Buckel, technology evangelist for Kaminario says, a storage administrator can then use the information generated by the Kaminario Clarity analytics service to apply the appropriate policies to eliminate that contention.
“Clarity is going to provide a suggestion about the best pathway for particular types of workload,” says Buckel.
Armed with that information, Buckel says, storage administrators can then determine whether it might be better in the future to scale up or scale out their storage systems. Designed to support either approach, Buckel says, over time the I/O characteristics of a particular workload can change. Because of that issue, the K2 all-Flash storage arrays are designed from the ground up to adapt to both scale up and out storage architecture requirements, says Buckel.
Scheduled to be available in the first half of 2017, Kaminario Clarity is able to make those recommendations by analyzing the data that Kaminario support services regularly collects via the cloud. The Kaminario Clarity application then analyzes the data collected in the Kaminario repository to identify optimal I/O paths for different classes of workloads. Based on that data, the machine learning algorithms developed by Kaminario make a recommendation.
Thanks to the rise of private cloud and microservices, I/O contention issues are only going to become more complex to manage as the number of workloads deployed per server increases. The Kaminario Clarity application isn’t intended to replace storage administrators as much as it is to make it easier for storage administrators to manage I/O contention at scale.
Obviously, the rise of all-Flash arrays has been a boon to application performance. The challenge now is optimizing the allocation of that storage across all the applications demanding to access it.