Scale-Up vs. Scale-Out Architecture

By CIOReview | Wednesday, December 21, 2016

With the burgeoning amount of data being generated every day, it’s becoming more and more critical for enterprises to develop new and improved ways to store data in a secure environment without affecting the performance of the system in any way. The debate has been raging on for quite some time about the supremacy of ‘Scale-Up or Scale-Out’ architectures for primary storage.

Cost and performance constitute the driving force in deciding the architecture solution to enhance performance and address the needs of any organization. Initially, vendors were obsessed with increasing the performance and reinforcing their storage by adding additional capacity options. In today’s business world, caching and tiering are being used to optimize the performance of flash by combining it with disk drives.

All–Flash Scale-Up and Scale-Out Approaches

Scale-Up refers to the more traditional architecture that uses a fixed controller (or a dual controller) to do all the processing work. New shelves are added to increase the capacity of the disk drives to the maximum level. Unfortunately, it cannot support the spinning hard drives which only support a couple of hundreds of IOPs (Input/Output Operations per Second) each. Secondly, as many users have observed that the scale-up architecture is often inefficient, the maximum capacity of the controller is attained long before the maximum capacity of the disk drive is achieved. In the scale-up architecture, traditionally, a dual controller is used. Due to its cost-effective nature, scale-up architecture offers a much viable option. But this architecture can also lead to performance problems as the controller resources are insufficient for the ever growing data.

On the contrary, scale-out is the type of architecture that does not depend on single controllers and scales up by adding more storage space. Scale-out architectures have got a number of nodes, with each node having a controller and storage capacity of its own. This enables them to scale up the process as the need arises.

Scale-Out all-flash arrays consist of clusters of storage nodes that are added to increase the capacity, which can go well over a million IOPs. This is an ideal choice for high performance, as more workloads can be added when the capacity is increased. For long term datacenter performance and flexibility, scale-out is the ideal choice. The use of all-flash scale-out solution provides mixed storage workload consolidation and has become an ideal choice for organizations.

Challenges while Opting Scale-out Architecture

The problem lies in the fact that with the addition of nodes, the architecture can grow really big and affect the overall performance. Most importantly, for businesses, the more nodes are added to meet specific demands and this leads to a proliferation of nodes, where the capacity of nodes go unutilized and are not fully leveraged. So, it becomes imperative for organizations to fully tap the potential of each node before opting to increase the capacity of each node.

With the price of all flash arrays plummeting at a rapid pace, it’s becoming clearer by the day that the traditional approaches for data storage are becoming obsolete and it’s better for organizations to shift to all flash arrays, which offers a better mix of flexibility and performance. Before one considers a scale-up or scale-out architecture, one should look carefully at the needs and requirements of the organization and choose wisely. Increasing the performance and scalability is a tough challenge for any IT organization and with the right infrastructure in place it can be solved effectively.