Cloud credibility took a hit in 2023, there’s no question about it. It’s still taking one in 2024 too. One troubling truth is that there are still signs that the lack of thorough evaluation of cloud benefits is (no pun) clouding current cloud planning. Might the same thing that led to too much cloud may now lead to too little? Let’s see.
I got cloud comments from 274 enterprises so far this year, and 255 said they believed that they had underestimated cloud costs for their projects in 2023. Of that group, 208 said that they should have done something different and would have if they’d been more thorough. All of this group said they would be assessing cloud opportunities differently in 2024, and 194 said they had changed their basic approach to the cloud at least somewhat. Only 28 had or planned to repatriate some applications, but 111 said they expected to actually modify some cloud applications to reduce costs.
The software architects and application planners who commented on the topic (187 of them) believed the primary reason they’d failed to assess cloud cost/benefits properly was senior management expectations. They suggested that development mangers, line management, and even C-suite executives has been “indoctrinated” and launched cloud projects without much in the way of assessment. In particular, they blamed the CIO and CFO for jumping into cloud usage with less planning than would have been done for an in-house technology change. Several, in fact, said that they’d gotten pushback from CIOs/CFOs on changes to data center middleware that the tech experts had “carefully explained and justified”, and the very same people “jumped on the cloud bandwagon” without careful technology planning and financial justification.
This wasn’t entirely a management problem, though. Of our 187 tech types, 124 admitted that they hadn’t pushed back hard enough. The cloud was hot, cloud experience was valuable, and they had their own career paths to think about. Of the 124, 98 believed that they might have been able to turn some of the particularly bad cloud decisions around had they tried harder.
The underlying driver of all of this, according to all 194 who said they’d changed their approach to the cloud for 2024, was a combination of hype and sales strategies that bordered on misleading. Of the 194, 170 said that a review of their documentation on cloud assessment showed that they were presented with inaccurate cost assessments, customer references that were unrealistic or incorrect, and documentation from vendors and third party sources that was incomplete and inaccurate. In 143 of the cases, the companies admitted that the services initially proposed were not the ones implemented; that project changes had introduced higher costs that were never really assessed properly. About a third said they blamed the cloud providers for not telling them that their changes would raise costs significantly.
Before we blame everything on predatory cloud sales practices, though, we should reflect on a basic truth. Of those 194 companies, 164 said that they got similarly inaccurate, incomplete, or obsolete information from data center software vendors too. Of the 208 companies who said they’d do something different with the cloud if they had to do it over, 191 said that they routinely had to discount information provided by their data center vendors. They hadn’t done that for the cloud, and 177 said the reason was that the cloud was different, more so than they’d expected. Their “normal” technology assessment strategies, honed for decades on the data center, just missed the boat with regard to the cloud because it didn’t work (and wasn’t priced) the same.
It’s also hard to truly learn the cloud. Going back to the 208 companies, 190 said that they didn’t find online articles about the cloud were both accurate and comprehensive. Among the same group, 173 said they didn’t have adequate books or ebooks on the cloud, and 77 said that even certifications didn’t cover all the topics they needed for proper cloud evaluation, though they were helpful in development techniques. Not so much for cost management, though. In the group 49 companies had used professional integrator services, and 30 of that group said they were more costly than expected, and overall didn’t deliver a higher success rate than companies achieved on their own.
CIO magazine has a decent piece on the great cloud re-think, and it makes many of the same points I heard from enterprises. I think it even includes a key technical point, though it uses different terminology than I use, and that I hear from enterprises themselves. The article includes a subhead “Reassessing, one workload at a time”, and I like that if we change “workload” to “workflow”.
Workflows are, IMHO, the big missing ingredient in application planning today. Many enterprise applications today, particularly the long-standing core business ones, are really monolithic. It’s popular to criticize the monolith model simply because recent programming practices encourage componentization, but the real point with monolithic applications don’t have workflows; work stays within the application. When they’re modernized, it’s usually done by adding in a front-end element that allows web and smartphone access. That adds a component, but it also introduces workflow.
In multi-component applications in general, and in particular when the components are distributed between cloud(s) and data centers (the hybrid model), workflows are the single most critical concept to consider. Almost everything related to the cloud involves workflows, because scaling, resiliency, hybrid and multi-cloud, all involve movement of information among components. Workflows are what generate latency, influence overall reliability, and in the cloud form what’s almost always a source of incremental costs.
The biggest technical problem I’m seeing in enterprise discussions of their cloud usage is the failure to associate the process of optimizing cloud value with the optimization of cloud costs. What makes the cloud useful is what makes it different from the data center, and those things are also typically what introduce additional costs. Development effort is reduced through the use of cloud web services, and those cost money. Scalability and resilience also introduce costs, but they’re the essence of the cloud value proposition. The only way to balance these two competing factors is through analysis of workflows, and that means keeping workflows out front during both the initial application development and any evolution to that approach, soon or later on.
If you dig down in the 77 enterprises who have faithfully followed workflow-centric cloud planning, you find something interesting. The cloud is never cheaper than the data center when it’s used for fairly static applications, ones that don’t see major variations in resource requirements through normal use. Where the cloud is valuable is where workloads vary significantly, both in terms of volume and in terms of the workflow paths through components. Workflow analysis, by letting these enterprises associate traditional cloud benefit management with cost management, has put these enterprises on track. They believe, and I agree, that their approach could be almost universally successful if applied broadly.
But can it be? We’ve had “cloud-native” and “cloud-first”, but not “workflow-first”, or even “hybrid-first”. IBM is actually the only vendor that enterprises (those who use IBM’s stuff, at least) consistently say have given them realistic cloud and hybrid advice. I’ve talked to Dell and HPE people who have the right idea too, but these two score well below IBM’s level, and other vendors don’t match those two. The good news is that enterprises are no longer looking at the cloud as the place where everything ultimately ends up, and that’s the path to the right kind of cloud decisions.
