CloudQoS Best Fit, Right Sizing and Migration Assessment
Sure, the migration plan looks good on paper, but how do you know which cloud service provider (CSP) is really the right fit? Or even more challenging, which configuration, out of the millions of cloud configuration, should you use? Which combination of cloud provider, location and SKU provides that optimal blend of price and performance? Once you make the choices on paper how do you rapidly test those choices to make sure expectations meet reality – were your assumptions right?
Krystallize has been running monthly cloud benchmarking on AWS, Azure, Google and IBM for more than 3 years. Comparing cloud providers and their SKU’s and measuring their performance variability. For example, in 2018 saw an average fluctuation in capability of 79% on AWS.
What would 79% performance fluctuation do to your application during a peak period? How would 79% performance degradation affect your cloud bill in an environment that auto-scaled to resolve performance issues? Would standard operational procedures increase the size of underperforming assets due to capability degradation?
These realities led us to develop cloud CloudQoS to provide transparency to the resource demand of applications and help navigate the myriad of choices todays hybrid and multi-cloud world provides. We simplify migration and sizing, through uniformly capturing an application true performance demands and reconciling that with supply available from the various cloud available.
How Do We Do It?
Using our patented CloudQoS agent, we capture actual workload (application instance) demand by using numerous standardized metrics and leverage that insight to create a synthetic digitaltwin of your workload. Once created the workload can be rapidly deployed on multiple cloud targets (internal or external clouds) to evaluate the most appropriate provider and SKU. This approach provides far more confidence in proper cloud selection – using real capability measures vs. generic metrics and assumptions.
Australian Energy Company in support of their Migration planning to Azure
What Did We Do?
In planning for their migration, our client had mapped their existing “On Premise” infrastructure using a “like for like” approach, including pricing, for their infrastructure in Azure.
After talking to Krystallize, they asked us to come in and run CloudQoS to see what, if any the delta may be.
Performance data was collected from the “on premise” hosts.The collection of the performance data of each host resulted in 7 workload demand profiles which were then used to create 7 Synthetic workloads.
Each Synthetic Load was then tested on 3-4 configurations to find the Best Fit, Best Cost solution.Each ‘best fit” solution was a combination of the best instance to handle the demand of the workload.
What Did We Find?
There were numerous opportunities for optimization, and in some cases high optimization.
In some instances, the “like-for-like” were of a similar performance, however taking into account environments, risk appetite (development environments, non-critical environments) there was the ability to provision much smaller and far less costly machines resulting in a smaller and cheaper estate.
In other instances, based on the synthetic workload, a number of original recommendations whilst on paper looked correct, would not meet the compute operations required by the native machine when needed.Understanding the workload demand prior to provisioning a cloud instance
allows the customer to find the right cloud, configuration and
provisioning requirements prior to migration.