Hybrid cloud strategy: 4 signs yours needs a refresh
Blog: The Enterprise Project - Enterprise Technology
To paraphrase a very old poem, “the best laid plans of IT leaders and architects sometimes go awry.”
Sure, the original verse concerned a mouse, but the underlying principle has become timeless: Things can go amiss with even the most carefully developed strategies.
The key then is to regularly review your plans and be willing to revise when the yellow and red flag start indicating that you might be veering off-track. This is an important principle for hybrid cloud success, too. Yes, it’s a good sign that you have a plan in the first place. But that strategy isn’t necessarily evergreen.
[ Learn the do’s and don’ts of hybrid cloud: Get the free eBook, Hybrid Cloud Strategy for Dummies. ]
Hybrid cloud strategy: 4 warning signs
1. Your end users are grumbling
People sometimes seem to have a natural tendency to complain. But when the volume of complaints – quantity or decibels could both apply here – starts to spike, there’s probably an underlying reason.
“The first indicator that something isn’t right is likely going to come in the form of complaints from your users,” says Scott Sneddon, senior director and multi-cloud evangelist at Juniper Networks. “Oftentimes, the additional latency that comes from running an application in a location other than on-premises will have a performance impact that’s difficult to predict.”
Network latency and other potential performance issues should be part of any enterprise’s evaluation of whether to migrate particular workloads to a cloud. And they’re worth continuing to keep tabs on post-migration, too. Again, if you’ve got unhappy users – especially of an application that never used to send many people in search of the complaint department – then that’s a sign that it might be time for a strategy review. That’s all the more true in the current moment.
“In this time of remote working, latency can be made even worse if end users are forced to connect over a VPN back to a central location before that traffic is sent back out over the internet to the cloud location where the application is running,” Sneddon says. “End-to-end performance monitoring is key.”
While the work-from-home context is specific to internal users, latency issues in general can also impact external or customer-facing applications.
[ Want to improve your hybrid cloud strategy? Get the four-step hybrid cloud strategy checklist. ]
2. You start to expect the unexpected in your cloud bills
Over the long haul, one or two surprises in your cloud bills aren’t always cause for immediate alarm. But if you’re regularly or even semi-regularly caught off-guard by your bills – and especially if you can’t explain why they’re higher than you anticipated – then something’s wrong.
“Cloud cost control is a common complaint and it only gets harder as clouds proliferate,” says Gordon Haff, technology evangelist at Red Hat.
Haff adds that getting a handle on public cloud costs can be particularly tricky, enough so that there are consultants whose entire business is helping customers reduce their bills. But even if you bring in outside help, this is a sign that your strategy might need to be reviewed.
“IT organizations need to at least keep close watch on potentially big-ticket items like data egress, expensive resources that were spun up for some purpose and never shut down, and services running in multiple locations,” Haff says. “Billing alerts are one useful tool but, more generally, watch for costs that seem to be growing faster than the demands being put on your clouds.”
Sneddon similarly sees a skyrocketing public cloud bill as a sign that you need to revisit your hybrid cloud plan, particularly in areas like monitoring, governance, and your application’s characteristics.
“If deployments aren’t monitored, and if constraints aren’t placed around how applications are deployed, application sprawl can happen fast and costs can rise,” Sneddon says. “Network I/O and storage costs can also be a surprise if these variables associated with an application are not fully understood in advance, or are not closely monitored.”
While unexpected bills can always be a sign that it’s time to review your strategy, they can be especially important to keep a close eye on when your team is less experienced with public cloud platforms.
“These costs can be especially surprising to IT and finance departments who are used to the more predictable costs normally associated with building private infrastructure,” Sneddon says.
Let’s look at two more warning signs that deserve attention:
3. A once-reliable app falters in a new environment
Reliability and resiliency are common general goals of hybrid cloud and multi-cloud strategies. If an application starts experiencing more outages or other problems than it did previously, then that should obviously be a sign that something might be amiss in your strategy, especially as it pertains to matching the right workloads to the right environments.
“Enterprises often don’t have a full understanding of how their applications might behave in different environments,” Sneddon says.
This has become one of the driving appeals of containerization, CI/CD, and Kubernetes – the ability to ensure that a developer’s code runs well in any environment. But, of course, not every application was built in a cloud-native manner – far from it. So you’ll need to review and revisit how different workloads behave in different environments.
[ Related reads: How to explain Kubernetes in plain English and 3 reasons to use an enterprise Kubernetes platform. ]
Sneddon notes that this warning sign can pop up sometimes after a legacy application is migrated to a cloud environment. It may be a sign that you need to revisit an initial decision to take a lift-and-shift approach (in which the application’s code and components remain relatively unchanged), instead of refactoring to take fuller advantage of the new environment’s features.
“Applications can also suffer from unexpected outages in public cloud infrastructure that may not have been experienced in a private cloud if those applications are not refactored,” Sneddon says.
One reason for this, according to Sneddon, is that private cloud infrastructure is typically built to “five nines” of reliability and usually not oversubscribed.
“Building this way can be expensive, but the reliability is undeniable,” Sneddon says. “Public cloud infrastructure is usually built to accommodate more users on less capacity, and is often built with lower reliability to reduce the cost of the infrastructure. The expectation is that applications that are built to run on cloud are more distributed and resilient than traditional monolithic enterprise applications, and can withstand failures at the infrastructure layer.”
On a related note, latency isn’t just an application-to-user (or vice versa) issue. Latency can also occur between various components of an application, and cause similar issues in terms of performance. If you’re experiencing latency problems, Sneddon notes that it could be a sign that you need to revisit your architecture – to see if there are slowdowns between various components of an application that may be running in a more distributed manner than before.
“The consideration of how applications access storage, where that storage needs to be located for compliance reasons, and how that location affects performance is [also] very important,” Sneddon says. “It may be relatively simple to move or start an application module in a new location to address a user complaint. It’s not so simple to move the data that the application needs to run.”
4. You don’t have criteria for evaluation and measurement
Another potential warning sign: You have no warning signs. If you have no criteria for making initial and ongoing decisions about your hybrid cloud strategy, then you don’t really have a healthy way of ensuring things are going according to plan. Sure, outrageous bills or massive performance issues will eventually still tell you something’s up – but the point here is that you shouldn’t wait for more urgent problems to occur.
“Well-defined objectives and success metrics should be a key part of your system audit and the hybrid solution selection process,” says Daniel Herndon, senior product manager for cloud infrastructure at Laserfiche. “The objectives and success metrics should represent improvements over the existing system, and play a role in the justification and benefit of a new system.”
Herndon notes that this is a solid approach to evaluating new platforms, and also to measure outcomes post-adoption.
“Success metrics are also a critical tool for gaining executive sponsorship, maintaining confidence in the project, as well as providing status updates over the life of the project,” Herndon says.
Dave Newell, director of the North American cloud center of excellence at Capgemini, advises setting benchmark goals for any environment in your hybrid cloud architecture. These criteria include things like: cost, technology, license management, support capabilities, data security, application longevity, performance, and internal/external interfaces, Newell says.
Initial evaluations of a particular vendor or of a workload’s fit for a particular environment are important, of course. Newell also advises using metrics or KPIs across any environment in your hybrid cloud on a continuous basis. The aforementioned example of increasing outages or downtime (that doesn’t have a ready explanation) is a fundamental one, but there are others. These become your own built-in early warning system that things might not be going according to the plan, and you can then act accordingly.
“Your hybrid cloud objectives need to be realigned when service level agreement (SLA) failures begin to increase, escalations begin to consume more management involvement, unplanned technology refreshes impact your cost models, training requirements significantly increase for your IT resources, or security controls fail to meet your business expectations,” Newell says.
[ Learn more about hybrid and multi-cloud workload strategy: Get the free eBook, Multicloud Portability for Dummies. ]