Episode Summary

iland Director of Cloud Market Intelligence Brian Knudtson is joined by guests Rebecca Fitzhugh, David Klee, and Jason Shiplett for a conversation around the concerns customers have with performance in the cloud. They discuss the assumptions we can (and can’t) make about cloud performance, the impact on expectations and monitoring when moving from a dedicated to shared environment, and how to best assess application performance. The cloud might be chock full of performance, but what’s the secret to extracting it? Listen and find out.

Panel

Cloud Conversations

Topic 1

[03:25] Performance came in straight in the middle of my polling of concerns with moving to the cloud. Should we just assume all clouds have enough performance that customers shouldn’t really worry about it while considering providers?

Topic 2

[08:29] While moving to the cloud, most customers will be moving from a dedicated environment to a shared environment. How will this change performance expectations and monitoring?

Topic 3

[16:01] How should customers assess performance of their applications in the cloud? Are existing on-premises tools sufficient?

Cloud Bites

[02:40] “If you don’t architect it properly, you’re going to end up with some pretty substantial performance bottlenecks.” — David Klee

[04:23] “You can get the performance you need, but the onus is on you to architecture applications in an optimized manner.” — Rebecca Fitzhugh 

[05:23] “What is the optimal architecture for this service or application and what is the appropriate right sizing.” — Rebecca Fitzhugh 

[05:54] “Cloud’s got all the performance that you want as long as you’ve got an unlimited budget and a thick wallet.” — David Klee 

[07:13] “For most people, one millisecond or five milliseconds is an order of magnitude on price. Nobody’s going to notice the difference.” — David Klee

[07:42] “If you don’t know your application stack, if you don’t know how all the disparate pieces fit together, how are you ever going to determine where your bottlenecks are?” — Jason Shiplett

[10:42] “We really need to establish what KPIs we’re talking about to make sure that we’re really speaking a consistent language, because what I consider good performance, what you consider good performance, it’s really subjective at the end of the day.” — Rebecca Fitzhugh 

[11:19] “If those SLOs aren’t in place, and if you haven’t worked out an agreement with your provider for a minimally acceptable level of performance, then if you’re running into issues, they might say everything’s great on our end, so it’s on your end, so shove off.” — Jason Shiplett

[15:06] “Talk to end users, get their requirements around how they use the application, how they use data, how they interact with the system and the changes that you’re making, moving something from on prem to a shared cloud environment.” — Jason Shiplett 

[16:55] “The toolset that you would have to leverage in that case would be something either intrinsic to the service or something where you can monitor more end user centric metrics like application load time or Web page load time, things of that nature.” — Jason Shiplett 

[18:38] “It’s just different. So you have to retool your processes.” — David Klee

[21:08] “Throw away your on prem tools and start over. That would be my hot take. But again, nuance within that.” — Rebecca Fitzhugh

[22:34] “My big thing is I don’t like reactive tools, and that’s what traditional monitoring has always been. I want proactive tools.” — David Klee

[25:38] “Identify the metrics that matter. Simply put, you’re going to collect a lot of data that just doesn’t matter. So what matters?” — Rebecca Fitzhugh