Striking a Balance on the Cloud with Rachel Stephens

Rachel Stephens is a Senior Analyst with RedMonk, a developer-focused industry analyst firm. RedMonk focuses on how practitioners drive technological adoption. Her research covers a broad range of developer and infrastructure products, with a particular focus on emerging growth technologies and markets. (But not crypto. Please don't talk to her about NFTs.)

Before joining RedMonk, Rachel worked as a database administrator and financial analyst. Rachel holds an MBA from Colorado State University and a BA in Finance from the University of Colorado.

Announcer: Hello, and welcome to Screaming in the Cloud with your host, Chief Cloud Economist at The Duckbill Group, Corey Quinn. This weekly show features conversations with people doing interesting work in the world of cloud, thoughtful commentary on the state of the technical world, and ridiculous titles for which Corey refuses to apologize. This is Screaming in the Cloud.

Corey: This episode is sponsored in part by my friends at ThinkstCanary. Most companies find out way too late that they’ve been breached. ThinkstCanary changes this and I love how they do it. Deploy canaries and canary tokens in minutes and then forget about them. What's great is the attackers tip their hand by touching them, giving you one alert, when it matters. I use it myself and I only remember this when I get the weekly update with a “we’re still here, so you’re aware” from them. It’s glorious! There is zero admin overhead  to this, there are effectively no false positives unless I do something foolish. Canaries are deployed and loved on all seven continents. You can check out what people are saying at canary.love. And, their Kub config canary token is new and completely free as well. You can do an awful lot without paying them a dime, which is one of the things I love about them. It is useful stuff and not an, “ohh, I wish I had money.” It is speculator! Take a look; that’s canary.love because it's genuinely rare to find a security product that people talk about in terms of love. It really is a unique thing to see. Canary.love. Thank you to ThinkstCanary for their support of my ridiculous, ridiculous non-sense.   

Corey: This episode is sponsored in part by our friends at Vultr. Spelled V-U-L-T-R because they’re all about helping save money, including on things like, you know, vowels. So, what they do is they are a cloud provider that provides surprisingly high performance cloud compute at a price that—while sure they claim its better than AWS pricing—and when they say that they mean it is less money. Sure, I don’t dispute that but what I find interesting is that it’s predictable. They tell you in advance on a monthly basis what it’s going to going to cost. They have a bunch of advanced networking features. They have nineteen global locations and scale things elastically. Not to be confused with openly, because apparently elastic and open can mean the same thing sometimes. They have had over a million users. Deployments take less that sixty seconds across twelve pre-selected operating systems. Or, if you’re one of those nutters like me, you can bring your own ISO and install basically any operating system you want. Starting with pricing as low as $2.50 a month for Vultr cloud compute they have plans for developers and businesses of all sizes, except maybe Amazon, who stubbornly insists on having something to scale all on their own. Try Vultr today for free by visiting: vultr.com/screaming, and you’ll receive a $100 in credit. Thats v-u-l-t-r.com slash screaming.

Corey: Welcome to Screaming in the Cloud, I’m Corey Quinn. The last time I spoke to Rachel Stephens over at RedMonkwas in December of 2019. Well, on this podcast anyway; we might have exchanged conversational tidbits here and there at some point since then. But really, if we look around the world there’s nothing that’s materially different than it was today from December in 2019, except, oh, that’s right everything. Rachel Stephens, you’re still a senior analyst at RedMonk, which hey, in this day and age, longevity at a company is something that is almost enough to occasion comment on its own. Thanks for coming back for another round, I appreciate it.

Rachel: Oh, I’m so happy to be here, and it’s exciting to talk about the state of the world a few years later than the last time we talked. But yeah, it’s been a hell of a couple years.

Corey: Really has, but rather than rehashing pandemic stuff because I feel like unless people have been living in a cave for the last couple of years—because we’ve all been living in caves for the last couple of years—they know what’s up with that. What’s new in your world? What has changed for you aside from all of this in the past couple of years working in one of the most thankless of all jobs, an analyst in the cloud computing industry?

Rachel: Well, the job stuff is all excellent and I’ve had wonderful time working at RedMonk. So, RedMonk overall is an analyst firm that is focused on helping people understand technology trends, particularly from the view of the developer or the practitioner. So, helping to understand how the people who are using technologies are actually driving their overall adoption. And so there has been all kinds of interesting things that have happened in that score in the last couple of years. We’ve seen a lot of interesting trends, lots of fun things to look at in the space and it’s been a lot.

On a personal side, like a week into lockdown I found out that I was pregnant, so I went through all of locking down and the heart of the pandemic pregnant. I had my maternity leave earlier this year and came back and so excited to be back in. But it’s also just been a lot to catch up on in the space as you come back from leave which I’m sure you are well familiar with.

Corey: Yes, I did the same thing, slightly differently timed. My second daughter Josephine was born at the end of September. When did your kiddo arrive on the scene to a world of masked strangers?

Rachel: So, I have an older daughter who just turned four, and then my youngest is coming up on his first birthday. He was born in December.

Corey: Excellent. It sounds like our kids are basically the same age, in both directions. And from my perspective, at least looking back, what advice would I give someone for having a baby in a pandemic? It distills down to ‘don’t,’ just because it changes so much, it’s no longer a trivial thing to have a grandparent come out and spend time with the kid. It’s the constant… drumbeat of is this over? Is this not over?

And that manifested a bunch of different ways. And I’m glad that I got the opportunity to take some time off to spend time with my family during that timeframe, but at the same time, it would’ve great if there were options such as not being stuck at home with every rambunctious—at the time—three-year-old as I went through that entire joy of having the kid.

Rachel: Yeah. No, for the longest time, my thing was like, okay, like, there’s no amount of money you could pay me to go back to middle school. I would never do it. And my new high bar there is no amount of money that you could pay me to go back to April 2020. That was the hardest month of my entire life was getting through that, like, first trimester, both parents at home, toddler at home, nowhere to go, no one to help. That was a [BLEEP] hard month. [laugh] that was bad.

Corey: Oh, my God, yes, and we don’t talk about this because we’re basically communicating with people on social media, and everyone feels bad looking at social media because they’re comparing their blooper reel with everyone else’s highlights. And it feels odd on some level to complain about things like that. And let’s be very clear as a man, I wind up in society getting lauded for even deigning to mention that I have children, whereas when mothers wind up talking about anything even slightly negative it’s, “Oh, you sound like a bad mom.” And it is just one of the most abhorrent things out there in the world, I suppose. It’s a strange inverted thing but one of the things that surprised me the most when I was expecting my first kid was looking at the different parenting forums, and the difference in tone was palpable where on the dad forums, everyone is super supportive and you got this dude it’s great. You’re fine. You’re doing your best.

Sure, these the occasional, “I gave my toddler beer and now people yell at me,” and it’s, “What is wrong with you asshole?” But everyone else is mostly sane and doing their best. Whereas a lot of the ‘mommy’ forums seem to bias more toward being relatively dismissive other people’s parenting choices. And I understand I’m stereotyping wildly, not all forums, not all people, et cetera, et cetera, but it really was an interesting window into an area that as a stereotypical white man world, I don’t see a lot of places I hang out with that are traditionally male that are overwhelmingly supportive in quite the same way. It was really an eye-opening experience for me.

Rachel: I think you hit on some really important trends. One of the things that I have struggled with is—so I came into RedMonk—I’ve had a Twitter account forever, and it was always just, like, my personal Twitter account until I started working at RedMonk five years ago. And then all of a sudden I’m tweeting in technical and work capacities as well. And finding that balance initially was always a challenge.

But then finding that balance again after having kids was very different because I would always—it was, kind of, mix of my life and also what I’m seeing in the industry and what I’m working on and this mix of things. And once you started tweeting about kids, it very much changes the potential perception that people have of who you are, what you’re doing. I know this is just a mommy blogger kind of thing.

You have to be really cognizant of that balance and making sure that you continue to put yourself in a place where you can still be your authentic self, but you really as a mom in the workspace and especially in tech have to be cognizant of not leaning too far into that. Because it can really damage your credibility with some audiences which is a super unfortunate thing, but also something I’ve learned just, like, I have to 
be really careful about how much mom stuff I share on Twitter.

Corey: It’s bizarre to me that we have to shade aspects of ourselves like this. And I don’t know what the answer is. It’s a weird thing that I never thought about before until suddenly I find that, oh, I’m a parent. I guess I should actually pay attention to this thing now. And it’s one of those once you see it you can’t unsee it things and it becomes strange and interesting and also more than a little sad in some respects.

Rachel: I think there are some signs that we are getting to a better place, but it’s a hard road for parents I think, and moms, in particular, working moms, all kinds of challenges out there. But anyways, it’s one of those ones that is nice because love having my kids as a break, but sometimes Mondays come and it’s such a relief to come back to work after a weekend with kids. Kids are a lot of work. And so it has brought elements of joy to my personal life, but it has also brought renewed elements of joy to my work life as really being able to lean into that side of myself. So, it’s been a good year.

Corey: Now that I have a second kid, I’m keenly aware of why parents are always very reluctant to wind up—the good parents at least to say, “Oh yeah, I have X number of kids, but that one’s my favorite.” And I understand now why my mom always said with my brother and I that, “I 
can’t stand either one of you.” And I get that now. Looking at the children of cloud services it’s like, which one is my favorite?

Well, I can’t stand any of them, but the one that I hate the most is the Managed NAT Gateway because of its horrible pricing. In fact, anything 
involving bandwidth pricing in this industry tends to be horrifying, annoying, ill-behaved, and very hard to discipline. Which is why I think it’s probably time we talk a little bit about egress charges in cloud providers.

You had a great analysis of Cloudflare’s R2, which is named after a robot in Star Wars and is apparently also the name of their S3 competitor, once it launches. Again, this is a pre-announcement, yeah, I could write blog posts that claim anything; the proof is really going to be in the pudding. Tell me more about, I guess, what you noticed from that announcement and what drove you to, “Ah, I have thoughts on this?”

Rachel: So, I think it’s an interesting announcement for several reasons. I think one of them is that it makes their existing offering really compelling when you start to add in that object store to something like the CDN, or to their edge functions which is called their Workers platform. And so, if you start to combine some of those functionalities together with a better object storage story, it can make their existing offering a lot more compelling which I think is an interesting aspect of this.

I think one of the aspects that is probably gotten a lot more of the traction though is their lack of egress pricing. So, I think that’s really what took everyone’s imaginations by storm is what does the world look like when we are not charging egress pricing on object storage?

Corey: What I find interesting is that when this came out first, a lot of AWS fans got very defensive over it, which I found very odd because their egress charges are indefensible from my point of view. And their response was, “Well, if you look at how a lot of the data access patterns work this isn’t as big of a deal as it looks like,” and you’re right. If I have a whole bunch of objects living in an object store, and a whole bunch of people each grab one of those objects this won’t help me in any meaningful sense.

But if I have one object that a bunch of people grab, well, suddenly we’re having a different conversation. And on some level, it turns into an interesting question of what differentiates this with their existing CDN-style approach. From my perspective, this is where the object actually lives rather than just a cache that is going to expire. And that is transformative in a bunch of different ways, but my, I guess, admittedly overstated analysis for some use cases was okay, I store a petabyte in AWS and use it with and without this thing. Great, the answer came out to something like 51 or $52,000 in egress charges versus zero on Cloudflare. That’s an interesting perspective to take. And the orders of magnitude in difference are eye-popping assuming that it works as advertised, which is always the caveat.

Rachel: Yeah. I remember there was a RedMonk conversation with one of the cloud vendors set us up with a client conversation that want to, kind of, showcase their products kind of thing. And it was a movie studio and they walked us through what they architecturally have to do when they drop a trailer. If you think about that thing from this use case where all of a sudden you have videos that are all going out globally at the same time, and everybody wants to watch it and you’re serving it over and over, that’s a super interesting and compelling use case and very different from a cost perspective.

Corey: You’ll notice the video streaming services all do business with something that is not AWS for what they stream to end-users from. Netflix has its own Open Connect project that effectively acts as their own homebuilt CDN that they partner with providers to put in their various environments. There are a bunch of providers that focus specifically on this. But if you do the math for the Netflix story at retail pricing—let’s be clear at large scale, no one pays retail pricing for anything, but okay—even assuming that you’re within hailing distance of the same universe as retail pricing; you don’t have to watch too many hours of Netflix before the data egress charges cost more than you’re paying a month than subscription. And I have it on good authority—read as from their annual reports—that a much larger expense for Netflix than their cloud and technology and R&D expenses is their content expenses.

They’re making a lot of original content. They’re licensing an awful lot of content, and that’s way more expensive than providing it to folks. They have to have a better economic model. They need to be able to make a profit of some kind on streaming things to people. And with the way that all the major cloud providers wind up pricing this stuff, it’s not tenable. There has to be a better answer.

Rachel: So, Netflix calls to mind an interesting antidote that has gone around the industry which is who can become each other faster? Can HBO become Netflix faster, or can Netflix become HBO faster? So, can you build out that technology infrastructure side, or can you build out that content side? And I think what you’re talking to with their content costs speaks to that story in terms of where people are investing and trying to actually make dents in their strategic outlays.

I think a similar concept is actually at play when we talk about cloud and CDN. We do have this interesting piece from my coworker, Steve O’Grady, and he called it “Convergent Evolution of CDNs and Clouds.” And they originally evolved along separate paths where CDNs were designed to do this edge-caching scenario, and they had the core compute and all of the things that go around it happening in the cloud.

And I think we’ve seen in recent years both of them starting to grow towards each other where CDNs are starting to look a lot more cloud-like, and we’re seeing clouds trying to look more CDN-like. And I think this announcement in particular is very interesting when you think about what’s happening in the CDN space and what it actually means for where CDNs are headed.

Corey: It’s an interesting model in that if we take a look at all of the existing cloud providers they had some other business that funded the incredible expense outlay that it took to build them. For example, Amazon was a company that started off selling books and soon expanded to selling everything else, and then expanded to putting ads in all of their search portals, including in AWS and eroding customer trust.

Google wound up basically making all of their money by showing people ads and also killing Google Reader. And of course, Microsoft has been a software company for a lot longer than they’ve been a cloud provider, and given their security lapses in Azure recently is the question of whether they’ll continue to be taken seriously as a cloud provider.

But what makes Cloudflare interesting from this approach is they start it from the outside in of building out the edge before building regions or anything like that. And for a lot of use cases that works super well, in theory. In practice, well, we’ve never seen it before. I’m curious to see how it goes. Obviously, they’re telling great stories about how they envision this working out in the future. I don’t know how accurate it’s going to be—show, don’t tell—but I can at least acknowledge that the possibility is definitely there.

Rachel: I think there’s a lot of unanswered questions at this point, like, will you be able to have zero egress fees, and edge-like latencies, and global distribution, and have that all make sense and actually perform the way that the customer expects? I think that’s still to be seen. I think one of the things that we have watched with interest is this rise of—I think for lack of a better word techno-nationalism where we are starting 
to see enclaves of where people want technology to be residing, where they want things to be sourced from, all of these interesting things.

And so having this global network of storage flies in the face of some of those trends where people are building more and more enclaves of we’re going to go big and global. I think that’s interesting and I think data residency in this global world will be an interesting question.

Corey: It also gets into the idea of what is the data that’s going to live there. Because the idea of data residency, yes, that is important, but where that generally tends to matter the most is things like databases or customer information. Not the thing that we’re putting out on the internet for anyone who wants to, to be able to download, which has historically been where CDNs are aiming things.

Yes, of course, they can restrict it to people with logins and the rest, but that type of object storage in my experience is not usually subject to heavy regulation around data residency. We’ll see because I get the sense that this is the direction Cloudflare is attempting to go in, and it’s really interesting to see how it works. I’m curious to know what their stories are around, okay, you have a global network. That’s great. Can I stipulate which areas my data can live within or not?

At some point, it’s going to need to happen if they want to look at regulated entities, but not everyone has to start with that either. So, it really just depends on what their game plan is on this. I like the fact that they’re willing to do this. I like the fact they’re willing to be as transparent as they are about their contempt for AWS’s egress fees. And yes, of course, they’re a competitor.

They’re going to wind up smacking competition like this, but I find it refreshing because there is no defense for what they’re saying, their math is right. Their approach to what customers experience from AWS in terms of egress fees is correct. And all of the defensiveness at, well, you know, no one pays retail price for this, yeah, but they see it on the website when they’re doing back-of-the-envelope math, and they’re not going to engage with you under the expectation that you’re going to give them a 98% discount.

So, figure out what the story is. And it’s like beating my head against the wall. I also want to be fair. These networks are very hard to build, and there’s a tremendous amount of investment. The AWS network is clearly magic in some respects just because having worked in data centers myself, the things that I see that I’m able to do between various EC2 instances at full network line rate would not have been possible in the data centers that I worked within.

So, there’s something going on that is magic and that’s great. And I understand that it’s expensive, but they’ve done a terrible job of messaging that. It just feels like, oh, bandwidth in is free because, you know, that’s how it works. Sending it out, ooh, that’s going to cost you X and their entire positioning and philosophy around it just feels unnecessary.

Rachel: That’s super interesting. And I think that also speaks to one of the questions that is still an open concern for what happens to Cloudflare if this is wildly successful. Which, based off of people’s excitement levels at this point, it’s seems like it’s very potentially going to be successful. And what does this mean for the level of investment that they’re going to have to make in their own infrastructure and network and order to actually be able to serve all of this?

Corey: The thing that I find curious is that in a couple of comment threads on Hacker News and on Twitter, Cloudflare’s CEO, Matthew Prince—who’s always been extremely accessible as far as executives of giant cloud companies go—has said that at their scale and by which they he’s referring to Cloudflare, and he says, “I assume that Amazon can probably get at least as decent economics on bandwidth pricing as we can,” which is a gross understatement because Amazon will spend years fighting over 50 cents.

Great, but what’s interesting is that he refers to bandwidth at that scale as being much closer to a fixed cost than something that’s a marginal cost for everything that a customer uses. The way that companies buy and sell bandwidth back and forth is complex, but he’s right. It is effectively a fixed monthly fee for a link and you can use as much or as little of that link as you want. 95th percentile billing aficionados, please don’t email me.

But by and large, that’s the way to think about it. You pay for the size of the pipe, not how much water flows through it. And as long as you can keep the links going without saturating them to the point where more data can’t fit through at a reasonable amount of time, your cost don’t change. So, yeah, if there’s a bunch of excitement they’ll have to expand the links, but that’s generally a fixed cost as opposed to a marginal cost per gigabyte.

That’s not how they think about it. There’s a whole translation layer that’s an economic model. And according to their public filings, they have something like a 77% gross margin which tells me that, okay, they are not in fact losing money on bandwidth even now where they generally don’t charge on a metered basis until you’re on the Cloudflare Enterprise Plan.

Rachel: Yeah. I think it’s going to just be really interesting to watch. I’m definitely interested to see what happens as they open this up, and like, 11 9s of availability feels like a lot of availabilities. It’s just the engineering of this, the economics of this it feels like there’s a lot of open questions that I’m excited to watch.

Corey: You’re onto my favorite part of this. So, the idea of 11 9s because it sounds ludicrous. That is well within the boundaries of probability of things such as, yeah, it is likely that gravity is going to stop working than it is that’s going to lose data. How can you guarantee that? Generally speaking, although S3 has always been extremely tight-lipped about how it works under the hood, other systems have not been.

And it looks an awful lot like the idea of Reed-Solomon erasure coding, where for those of us who spent time downloading large files of questionable legality due to copyright law and whatnot off of Usenet, they had the idea of parity files where they’d take these giant media files up—they’re Linux ISOs; of course they are—and you’d slice them into a bunch of pieces and then generate parity files as well.

So, you would wind up downloading the let’s say 80 RAR files and, oh, three of them were corrupt, each parity file could wind up swapping in so as long as you had enough that added up to 80, any of those could wind up restoring the data that had been corrupted. That is almost certainly what is happening at the large object storage scale. Which is great, we’re going to break this thing into a whole bunch of chunks. Let’s say here is a file you’ve uploaded or an object.

We’re going to break this into a hundred chunks—let’s say arbitrarily—and any 80 of those chunks can be used to reconstitute the entire file. And then you start looking at where you place them and okay, what are the odds of simultaneous drive failure in these however many locations? And that’s how you get that astronomical number. It doesn’t mean what people think of does. The S3 offers 11 9s of durability on their storage classes, including the One Zone storage class.

Which is a single availability zone instead of something that’s an entire region, which means that they’re not calculating disaster recovery failure scenarios into that durability number. Which is fascinating because it’s far, like, you’re going to have all the buildings within the same office park burn down than it is all of the buildings within a hundred square miles burn down, but those numbers remain the same.

There’s a lot of assumptions baked into that and it makes for an impressive talking point. I just hear it as, oh yeah, you’re a real object-store. That’s how I see it. There’s a lot that’s yet to be explained or understood. And I think that I’m going to be going up one side and down the other as soon as this exists in the real world and I’m looking forward to seeing it. I’m just a little skeptical because it has been preannounced.

The important part for me is even the idea that they can announce something like this and not be sued for securities fraud tells me that it is at least theoretically economically possible that they could be telling the truth on this. And that alone speaks volumes to just how out-of-bounds it tends to be in the context of giant cloud customers.

Rachel: I mean, if you read Matt Levine, “Everything is Securities Fraud?“ so, I don’t know how much we want to get excited about that.

Corey: Absolutely. A huge fan of his work. 

Corey: You know its important to me that people learn how to use the cloud effectively. Thats why I’m so glad that Cloud Academy is sponsoring my ridiculous non-sense. They’re a great way to build in demand tech skills the way that, well personally, I learn best which I learn by doing not by reading. They have live cloud labs that you can run in real environments that aren’t going to blow up your own bill—I can’t stress how important that is. Visit cloudacademy.com/corey. Thats C-O-R-E-Y, don’t drop the “E.” Use Corey as a promo-code as well. You’re going to get a bunch of discounts on it with a lifetime deal—the price will not go up. It is limited time, they assured me this is not one of those things that is going to wind up being a rug pull scenario, oh no no. Talk to them, tell me what you think. Visit: cloudacademy.com/corey,  C-O-R-E-Y and tell them that I sent you!

Corey: So, we’ve talked a fair bit about what data egress looks like. What else have you been focusing on? What have you found that is fun, and exciting, and catches your eye in this incredibly broad industry lately?

Rachel: Oh, there’s all kinds of exciting things. One of the pieces of research that’s been on my back-burner, usually I do it early summer, and it is—due to a variety of factors—still in my pipeline, but I always do a piece of research about base infrastructure pricing. And it’s an annual piece of looking about what are all of the cloud providers doing in regards to their pricing on that core aspect of compute, and storage, and memory.

And what does that look like over time, and what does that look like across providers? And it is absolutely impossible to get an apples-to-apples comparison over time and across providers. It just can’t actually be done. But we do our best [laugh] and then caveat the hell out of it from there. But that’s the piece of research that’s most on my backlog right now and one that I’m working on.

Corey: I think that there’s a lot of question around the idea of what is the cost of a compute unit—or something like that—between providers? The idea of if I have this configuration will cost me more on cloud provider a or cloud provider B, my pet working theory is that whenever people ask for analyses like that—or a number of others, to be perfectly frank with you—what they’re really looking for is confirmation bias to go in the direction that they wanted to go in already. I have yet to see a single scenario where people are trying to decide between cloud providers and they say, “That one because it’s going to be 10% less.” I haven’t seen it. That said I am, of course, at a very particular area of the industry. Have you seen it?

Rachel: I have not seen it. I think users find it interesting because it’s always interesting to look at trends over time. And in particular, with this analysis, it’s interesting to watch the number of providers narrow and then widen back out because we’ve been doing this since 2012. So, we used to have [unintelligible 00:26:24] and HPE used to be in there. So, like, we used to—CenturyLink. We used to have this broader list of cloud providers that we considered that would narrow down to this doesn’t really count anymore.

And now why do you need to back out? It’s like, okay, Oracle Cloud you’re in, Alibaba, Tencent, like, let’s look at you. And so, like, it’s interesting to just watch the providers in the mix shift over time which I think is interesting. And I think one of just the broad trends that is interesting is early years of this, there was steep competition on price, and that leveled off for solid three, four years.

We’ve seen some degree of competitiveness reemerge with competitors like Oracle in particular. So, those broad-brush trends are interesting. The specifics of the pricing if you’re doing 10% difference kind of things I think you’re missing the point of the analysis largely, but it’s interesting to look at what’s happening in the industry overall.

Corey: If you were to ask me to set up a simple web app, if there is such a thing, and tell you in advance what it was going to cost to host, and I can get it accurate within 20%, I am on fire in terms of both analysis and often dumb luck just because it is so difficult to answer the question. Getting back to our earlier conversational topic, let’s say I put CloudFront, Amazon’s sorry excuse for a CDN, in front of it which is probably the closest competitor they have to Cloudflare as a CDN, what’ll it cost me per gigabyte? Well, that’s a fascinating question. The answer comes down to where are you visiting it from? Depending where on the planet, people who are viewing my website, or using my web app are sitting, the cost per gigabyte will vary between eight-and-a-half cents—retail pricing—and fourteen cents. That’s a fairly wide margin and there’s no way to predict that in advance for most use cases. It’s the big open-ended question.

And people build out their environments and they want to know they’re making a rational decision and that their provider is not charging three times more than their competitor is for the exact same thing, but as long as it’s within a certain level of confidence interval, that makes sense.

Rachel: Yeah, and I think the other thing that’s interesting about this analysis and one of the reasons that it’s a frustrating analysis for me, in particular, is that I feel like that base compute is actually not where most of the cloud providers are actually competing anymore. So, like, it was definitely the interesting story early in cloud.

I think very clearly not the focus area for most of us now. It has moved up an abstraction layer. It’s moved to manage services. It has moved to other areas of their product portfolio. So, it’s still useful. It’s good to know. But I think that the broader portfolio of the cloud providers is definitely more the story than this individual price point.

Corey: That is an interesting story because I believe it, and it speaks to the aspirational version of where a lot of companies see themselves going. And then in practice, I see companies talking like this constantly, and then I look in their environment and say, “Okay, you’re basically spending 70% of your entire cloud bill on EC2 instances, running—it’s a bunch of VMs that sit there.”

And as much as they love to talk about the future and how other things are being considered and how their—use of machine learning in the rest, and Kubernetes, of course, a lot of this stuff all distills down to, yeah, it runs in software. It sits on top of EC2 instances and that’s what you get billed for. At re:Invent it’s always interesting and sad at the same time that they don’t give EC2 nearly enough attention or stage time because it’s not interesting, despite it being a majority of AWS bill.

Rachel: I think that’s a fantastic point, well made.

Corey: I’ll take it even one step further—and this is one where I think is almost a messaging failure on some level—Google Cloud offers sustained use discounts which apparently they don’t know how to talk about appropriately, but it’s genius. The way this works is if you run a VM for more than in a certain number of hours in a month, the entire month is now charged for that VM at a less than retail rate because you’ve been using it in a sustained way.

All you have to do to capture that is don’t turn it off. You know, what everyone’s doing already. And sure if you commit to usage on it you get a deeper discount, but what I like about this is if you buy some reserved instances is or you buy some committed use discounting, great, you’ll save more money, but okay, here’s a $20 million buy. You should click the button on, people are terrified to click at that button because I don’t usually get to approve dollar figure spend with multiple commas in them. That’s kind of scary. So, people hem and they haw and they wait six months. This is maybe not as superior mathematically, but it’s definitely an easier sell psychologically, and they just don’t talk about it.

It’s what people say they care about when people actually do are worlds apart. And the thing that continually astounds me because I didn’t 
expect it, but it’s obvious in hindsight that when it comes to cloud economics it’s more about psychology than it is about math.

Rachel: I think one of the things that, having come from the finance world into the analyst world, and so I definitely have a particular point of view, but one of the things that was hardest for me when I worked in finance was not the absolute dollar amount of anything but the variability of it. So, if I knew what to expect I could work with that and we could make it work. It was when things varied in unexpected ways that it was a lot more challenging.

And so I think one of the things that when people talked about, like, this shift to cloud and the move to cloud, and everyone is like, “Oh, we’re moving things from the balance sheet to the income statement.” And everyone talked about that like it’s a big deal. For some parts of the organization that is a big deal, but for a lot of the organization, the shift that matters is the shift from a fixed cost to a variable cost because that lack of predictability makes a lot of people’s jobs, a lot more difficult.

Corey: The thing that I always find fun is a thought exercise is okay, let’s take a look at any given cloud company’s cloud bill for the last 18 to 36 months and add all of that up. Great, take that big giant number and add 20% to it. If you could magically go back in time and offer that larger number to them as here’s your cloud bill and all of your usage for the next 18 to 36 months. Here you go. Buy this instead.

And the cloud providers laugh at me and they say, “Who in the world would agree to that deal?” And my answer is, “Almost everyone.” Because at the company’s scale it’s not like the individual developer response of, “Oh, my God, I just spent how much money? I’ve got to eat this month.” Companies are used to absorbing those things. It’s fine. It’s just a, “We didn’t predict this. We didn’t plan for this. What does this do to our projections, our budget, et cetera?”

If you can offer them certainty and find some way to do it, they will jump at that. Most of my projects are not about make the bill lower, even though that is what is believed, in some cases by people working on these projects internally at these companies. It’s about making it understandable. It’s about making it predictable, it’s about understanding when you see a big spike one month. What project drove that?

Spoiler, it’s almost always the data science team because that’s what they do, but that’s neither here nor there. Please don’t send me letters. But yeah, it’s about understanding what is going on, and that understanding and being able to predict it is super hard when you’re looking at usage-based pricing.

Rachel: Exactly.

Corey: I want to thank you for taking so much time to speak with me. If people want to hear more about your thoughts, your observations, et cetera, where can they find you?

Rachel: Probably the easiest way to get in touch with me is on Twitter, which is @rstephensme that’s R-S-T-E-P-H-E-N-S-M-E.

Corey: And we will, of course, put links to that in the [show notes 00:34:08]. Thank you so much for your time. I appreciate it.

Rachel: Thanks for having me. This was great.

Corey: Rachel Stephens, senior analyst at RedMonk. I’m Cloud Economist, Corey Quinn, and this is Screaming in the Cloud. If you’ve enjoyed this podcast, please leave a five-star review on your podcast platform of choice, whereas if you’ve hated this podcast, please leave a five-star review on your podcast platform of choice, along with an angry comment, angrily defending your least favorite child, which is some horrifying cloud service you have launched during the pandemic.

Corey: If your AWS bill keeps rising and your blood pressure is doing the same, then you need The Duckbill Group. We help companies fix their AWS bill by making it smaller and less horrifying. The Duckbill Group works for you, not AWS. We tailor recommendations to your business and we get to the point. Visit duckbillgroup.com to get started.

Announcer: This has been a HumblePod production. Stay humble.