Mainframe as a service: Myth Buster 005 - HASSEM PRAG
Myth003 - Mainframes are secure and have advanced features. Mainframe security has advanced so myth not busted.
Mo, Hassem, Mohammed, IT, CIO, Fintech, Advisory, Consulting, Leadership, Technology, Tech, IT, ISFAP,Jay Prag,Farieda Mayet,Jayesh,Bank Zero
16783
portfolio_page-template-default,single,single-portfolio_page,postid-16783,ajax_leftright,page_not_loaded,,qode-title-hidden,qode-content-sidebar-responsive,qode-theme-ver-16.7,qode-theme-bridge,qode_advanced_footer_responsive_1000,wpb-js-composer js-comp-ver-5.5.2,vc_responsive,elementor-default,elementor-kit-16169

Myth005 – Horizontal vs Vertical scaling, which is better

I guess scalability can be argued in many different ways. The key aspects of scalability must be the ability to automatically detect and provide capacity on demand with elasticity. So how do large organizations prepare for a sudden arrival of workloads that may cause serious damage to your reputation or may result in many customer queries and frustrations. With the advancement made in cloud technology and the use of frameworks such as Kubernetes and Docker containers, the ability to automatically increase elasticity to CPU and Memory becomes a lot easier to manage. This caters for the concept of horizontal scaling whereby a Kubernetes Master Node controls the many worker nodes that are configured. When workloads change, these worker nodes can be dynamically scaled up or down.

 

With vertical scaling as in the case of mainframes, CPUs and Memory are dynamically added to the max configuration for that particular hardware. This allows for capacity on demand and can be controlled in the same manner as with Kubernetes clusters. The only difference is that you are not allocating more servers to dynamically create elasticity. As demand increases for CPUs and Memory elasticity allows for workloads to be managed and later as demand reduces, we have a drop in CPU capacity. The disadvantage here is that you are physically limited to the hardware size of the physical mainframe. Our experiences have shown that if you plan correctly, you will never be in a situation where you outgrow this single physical footprint. So the question is which one is better. The answer is it depends on your type of workload and what you want to achieve and whether you have a cloud strategy or not.

 

There is no clear winner here so one has to determine the workloads and configurations needed in order to determine hosting requirements. The only issue is with z/OS where you are bound by rules that govern the platform. So if you have COBOL applications, you are bound by mainframe configurations. All other workloads can revert to horizontal scaling. There is a caveat to mainframes and horizontal scaling and that is if you implement a SYSPLEX DATA SHARING environment allowing you to scale horizontally.