Welcome to Build re:Invent Next!
The lights go down. It’s time for your preferred cloud vendor’s annual keynote to start. An inspiring video plays to what was charitably deemed “corporate dance music” but in practice sounds like a sped-up Enya on quaaludes.
Satya / Andy / a robot walks on stage to welcome us all to their conference. First, there’s the obligatory Serious Face about These Uncertain Times. That’s followed by the nod to racial justice by the CEO of a company whose entire Black senior leadership team could comfortably fit inside a Denny’s booth without crowding. The Serious Topics now suitably dispensed with, they turn to fantastical new creations that are or will be released within the coming months.
It’s odd. They do this every year, but you don’t see your bill reflecting any of these fantastical services. You wonder what’s being sold and to whom.
They’re selling a fantasy, and you’ve been buying it all along.
Anecdata science
I fix AWS bills for a living, generally for companies with annual cloud spend ranging from “a couple of million dollars” to “north of Tonga’s annual GDP.” I don’t pretend that this is a perfect sampling of the cloud customer base, but I see enough varied cloud environments to identify trends as they cross my desk. One of these days, I’ll perform a proper statistical analysis of the data and subject myself to the Well Actually Chorus. But until then, I’m afraid you’ll have to take my impressions on what I’ve seen as the next best thing.
A majority of spend across the board is and always has been the direct cost of EC2 instances. The next four are, in order, RDS, Elastic Block Store (an indirect EC2 cost), S3, and data transfer (yup, another EC2 cost).
That’s right. The cloud really is a bunch of other people’s virtualized computers being sold to you. The rest of it is largely window dressing.
Sure, there are exceptions: I’ve found some fascinating savings opportunities for companies making heavy use of DynamoDB, and some folks are seeing great success with transitioning things like Kafka from EC2 instances to Amazon MSK.
I’m not talking about YOU and your precious unicorn company; of course, YOUR bill is different. You might be acloud.guru last year, a company who proudly claimed at ServerlessConf that their total compute bill was between $1,000 and $2,000 per month! Statistically though, you’re not. You’re closer to acloud.guru this year after acquiring LinuxAcademy and its $2.7 million a year AWS bill.
It’s okay. You’re in good company.
The lies we tell ourselves
The promise of the cloud is that we can leverage a host of higher-level services. We can elastically scale up to meet demand then scale back down when the spike has passed. We can take advantage of a bunch of managed services built and run by experts and free ourselves from the undifferentiated toil that adds no direct business value.
On balance, we do none of those things. We treat the cloud like it’s an expensive data center, and we highlight the few exception cases that tell almost any other story than that sad truth.
This is what feeds directly into the narratives that shape (effective) cloud marketing. We imagine that somehow we’re going to break the cycle of making bad IT decisions and begin making good IT decisions just after this next sprint.
We’re lying to ourselves. But it feels better than the unvarnished truth.
Large enterprises so encumbered with process that they can’t break free long enough to innovate imagine themselves as transforming into nimble startups. Startups strangely don’t imagine the opposite—and that’s where we see the marketing magic come to life.
The Kubernetes shuffle
If you look carefully along the seams, the increased focus by all of the major cloud players (and, to be frank, the minor ones too) on Kubernetes highlights this.
Here is something new and trendy; companies talk about their container orchestration stories or their service mesh (which pluralizes to “service meesh”) offerings. Here’s something new and differentiated that detracts from the old-and-busted perception of VMs.
Except for the minor point that what Kubernetes and its ecosystem really unlock is the ability to shove existing workloads into containers, have no real idea what’s going on inside of them, and put them onto (you guessed it!) the compute that the cloud company is poised to sell you.
It seems like there’s a trend here.
Be sure to use the conference mobile app…
“…and all of these new services are available for your use today!”
Almost in passing, they mention a new generation of compute instances. Everyone nods and waits for the next slide.
But that “minor” release is going to cause the single greatest change in its customers’ unit economics of everything announced this entire dog-and-pony show of a conference.
The fact is that running virtual machines in the cloud is boring, and people crave something new and exciting.
Kubernetes was born out of this craving; it lets people running workloads on virtual machines cosplay as if they were cloud providers themselves. They shuffle and rebalance the deck chairs on the Titanic, working hard enough at keeping abreast of the exploding world of container orchestration that they can ignore the grim realization that their entire careers are now spent working on a product made by a company that doesn’t pay them.
Storing data, crunching that data, moving that data from place to place: That’s what fuels the cloud engines of the hyperscalers. That’s what you spend your entire IT budget on and that’s what the elaborate stage show you’re watching instead of working is artfully concealing from you.
Anyone who says otherwise is selling you something.
