Your app might be happy in the cloud, but it’s possible your storage might not be. Find out why ‘noisy neighbors’ can make cloud storage so tricky and what to do about it.
One of myths surrounding the cloud is that once applications are created or ported to them, storage isn’t an issue. With vendors selling multi-tenant services that promise scalability and flexibility, it’s easy for organizations to be lulled into thinking there won’t be any problems with storing data in the cloud — but that can’t be further from the truth. As experts have noted, there is a definite gap between cloud applications and cloud storage, and ignoring this fact could result in performance problems.
The noisy neighbor dilemma is one that dogs cloud applications. Withvirtualization meaning 10 different applications may run on the same server, they’re all competing for the same storage resource, according to Dan Florea, director of product management at Tintri, a storage provider specializing in the cloud. For example, a database with a very I/O-intensive workflow may need more resources to read data from storage. At the same time, a virtual desktop may be trying to access that same storage, and without some sort of regulation mechanism to ensure fairness, one user may be waiting for an unacceptable amount of time for the data.
Load balancing can alleviate latency woes
Load balancing can help distribute application resource use across servers, according to Michael Pardue, senior architect at Accusoft, which makes imaging processing tools. “One of the first things we try to do is break away from the local file system to various cloud service providers like AWS [Amazon Web Services],” he said. Otherwise, there needs to be more load balancing on the client side so that when the request comes through the server, all the resources are still available.
“We want to have a number of servers online to handle a large number of requests,” Pardue said. That means maintaining the notion of the session and coordinating requests across servers. Accusoft engages in “round robining,” sending requests to indiscriminate servers, he said.
Keeping application data out of the blender
Another issue with storing data in the cloud is mixing data from applications, Florea said. If data from two or more applications is all stored in one partition, it’s possible for the information about the applications to be mixed up and not useful for the business.