We learned of some interesting big data use cases and met many people embarking on exciting projects last week at the Strata Data Conference in San Francisco. Cloud migration has been an especially hot topic for enterprises lately, and not surprisingly, it came up often at the show. To remain competitive, organizations need real-time access to their data, and the reality is that a multi-cloud strategy is likely inevitable. So it makes sense that, along with helping attendees understand the importance of unified big data performance management, cloud migration was a familiar topic of discussion at the show.
We recently identified Three Crucial Considerations When Assessing Your Big Data Cloud Migration, which addressed pre-migration analysis. Today we’ll look at the bigger picture, from pre- to post-migration, to help ensure applications in the cloud perform reliably and meet their SLAs.
Pre-Migration Baseline and Workload Mapping
Migrating to the cloud without first looking at on-premises performance data is risky. If a migration is not done right, it’s more likely you’ll have to devote time, resources and money to re-engineer applications and complaints (and missed SLAs) will be inevitable. Assess your current on-premises environment carefully and look closely at what’s going on. Collect and analyze data, see how applications are performing, gauge your SLAs, and determine where there are typically usage bursts and performance dips. The only way to do that is to collect and measure actual metrics over time: Look at what’s happening for a month so you can see where the spikes are, when you are more likely to see issues, and what you need to ensure SLAs are maintained when you move to the cloud.
Each cloud service provider offers various instance types and each with different compute, memory, and storage capabilities. You’ll want to analyze your baseline assessment with the various instance types to identify which workloads, with their baseline and burst levels, are best suited for static or on-demand instances based on CPU and memory requirements. These results will help you calculate costs for each workload in the various cloud providers to ensure the most cost-effective implementation of your hybrid or multi-cloud strategy.
One of the biggest cloud migration hurdles is ensuring application SLAs and performance are maintained after migrating to the cloud. Skipping this crucial step could be disastrous.
Once the migration is complete, assess the applications in the cloud and compare the performance to the results of your on-premises baseline. Collect performance metrics again once everything is in the cloud and compare the results to your original baseline assessment. This will indicate whether performance levels have been maintained in the cloud or if more work needs to be done to make improvements.
Leverage Automation and Expertise
You could do the assessment manually but it’s difficult, time-consuming, and will likely cause delays in migration. Applying automation to pre- and post-migration discovery will accelerate the process and ensure accuracy.
Pepperdata can automatically analyze and profile every workload in your cluster to accurately determine your projected cloud costs, and provide the most appropriate instance recommendations for workloads, queues, jobs, and users (learn more). We map your big data workloads to various instance types to meet SLA requirements and enable you to compare services and costs across AWS, Azure, Google Cloud, and IBM Cloud (start a free assessment here).
Additionally, if your cloud service provider is AWS, Pepperdata has partnered with the big data experts at Cloudwick to ensure migration success. Pepperdata provides Cloudwick migration services customers with a baseline of on-premises performance, maps workloads to optimal static and on-demand instances, and diagnoses any issues that may arise during migration. Once the migration is complete, Pepperdata collects and analyzes the same operational metrics from the cloud to assess performance results and validate migration success.