This post is a transcript of an internal chat about AWS re:Invent. The chat covers the following topics:
- Hybrid event format (and what that means for the hallway track)
- S3 announcements
- Abstractions vs. primitives
- Egress costs and their effect on the competitive landscape
- Serverless announcements
- Sustainability
- Graviton3 and C7g instances, the momentum of ARM, and what that means for Intel
- Other impacts to competitors
- AWS’ move to additional markets
- Thoughts about Adam Selipsky’s first re:Invent as CEO
rachel: (Rachel Stephens): On occasion the RedMonk team will come together in Slack to discuss a tech event. We wanted to take apart some of the announcements from AWS re:Invent and figured there was no better person to discuss AWS news with than Corey Quinn.
Corey takes apart AWS bills, talks about tech news, and shitposts on Twitter for a living. Corey, we’re super excited to have you join us. 🙂
cquinn: (Corey Quinn): Thanks for the Quinnvitation; it’s always great to chat with RedMonk folks.
monkchips: (James Governor): @cquinn at what point did you realise Quinn rhymed with everything else?
cquinn: In 2010, when I changed my last name to “Quinn.” You say this like it was a happy accident or something rather than an intentional decision. 🙂
monkchips: Quinntentional
rachel: Awesome, let’s dive Quinn! The first topic I’d love to talk about is the hybrid event format itself. Let’s get people’s insights on attending virtually or in-person.
kelly (Kelly Fitzpatrick): I attended virtually, and part of that experience was wondering what the in-person event was like while also being happy that I was not in Vegas with 20K of my closest friends a few days after the Omicron variant was announced. It was part “what hallway track conversations am I missing?” and part “so glad to see folks wearing masks but also OMG that is a crowd.”
cquinn: I attended in person, which is why I’m having this chat from the hotel room in San Francisco where I’m isolating in the hopes of not killing my young children. It was something of a rebuilding year for the event; there were a lot of little issues along the way that stemmed from not remembering how to Event properly. The AWS events team were the unsung heroes of the event, to be clear.
sogrady (Stephen O’Grady): From my perspective, the virtual experience left a lot to be desired, though I feel for the AWS re:Invent team. When they originally planned the event, it must have looked like they’d be able to do it primarily in person – much as it did when we decided to hold the Monktoberfest. Where we just cancelled, however, AWS had to pivot late to a hybrid in-person / virtual event, and that seemingly broke some things.
Still, I at least got to attend sessions, unlike last year when Chime was the bane of my existence.
cquinn: I was surprised to see the virtual event wasn’t as polished as last year’s; I think you’re onto something with regard to the hybrid model breaking things.
monkchips: Ooh yes, the hybrid event format. The production was good on the keynotes. The talks were good. And attending at home means the room didn’t stink of cigarettes, cigars and cologne.
But yes re:Invent is indeed the Big One when it comes to hanging out with friends. I am happy to avoid flying, so I am into the hybrid option, but it does create logistical challenges.
I really think AWS needs to think about the virtual hallway. We actually ran our own Discord channel for that. A place to hang out, maybe shitpost a little bit, and enjoy talks with like minds. We named the rooms after bars and coffee shops on the Strip. I like it. I don’t think AWS gives that enough attention, but as Stephen says, AWS wasn’t expecting to do so much virtual this time around
kelly: And also it is a different framing to have talks that are designed to be consumed by virtual audiences only vs. talks that happen in person but then are live-streamed to virtual audiences. And I think those of us attending virtually felt that difference.
monkchips: @kelly you are so right
rachel: I think the virtual platform had the right idea of “people will want to consume this on their own time” (which is sometimes true! a benefit of virtual is that you don’t have to attend things on someone else’s schedule), but the platform’s design for real-time virtual viewership for sessions outside of the keynotes had a lot of friction.
kelly: That said, you could definitely feel a different energy from the talks that were in front of a live audience. In some ways I am so jealous of those speakers.
monkchips: People cheering every time someone said Nasdaq…
rachel: @cquinn, what was your favorite part of the in-person experience?
cquinn: It was wonderful to catch up with folks I hadn’t seen in two years. That forgives an awful lot of sins!
I’d forgotten that the rodeo comes into town just as the AWS folks are leaving. This gives rise to some honestly hilarious moments.
kelly: @cquinn I loved your walk through the expo hall as well.
sogrady: Did you say hi to Atomic Liquors for us?
cquinn: Atomic Liquors is apparently the oldest free standing bar in Las Vegas; you inadvertently started a new tradition by holding a drink-up there two years ago.
monkchips: Love that place. It was not Quinnadvertant though. It was very intentional.
kelly: It is a great place to get a break from the maze of casinos.
rachel: Mazes of casinos are suboptimal. Let’s talk about mazes of product announcements instead. Any favorite announcements from the week? What caught your eye?
monkchips: Oh my god though so much solid, incremental functionality. Where do we start?
cquinn: While FSx supporting OpenZFS was clearly the most noteworthy announcement, it was the S3 announcements that are going to have the most “quality of life” improvements for customers as I see it.
“Throw S3 ACLs away” and “Intelligent Tiering should now be your default choice of storage class” more specifically.
monkchips: Tell us more about Intelligent Tiering.
cquinn: With Glacier Instant Retrieval, combined with the “no monitoring fee for small objects” and “no 30 day minimum storage charge” changes from earlier in the year, unless you’ve got incredibly deep insight into your data lifecycle there’s no reason not to use it for most workloads. You’re well into the corner cases before that shifts.
monkchips: Good stuff.
sogrady: Unrelated, while the OpenZFS news was interesting as someone who goes back with that technology a ways, to me the most interesting announcement was actually what wasn’t announced: more abstractions. With apologies to Amplify Studio, this was very much a primitives conference which seemed at odds with both their recent history and the tone that Selipsky set up front.
kelly: And one that we are collectively responsible for:
.@Werner says all of these “primitives” are our fault because we “asked for them!” 😀 #reinvent pic.twitter.com/liu1S3pt6L
— Jeremy Daly (@jeremy_daly) December 2, 2021
sogrady: Indeed. It’s all our fault.
rachel: There seemed to be some internal politics at play between primitives and abstractions. Also, it’s not often you see C-level leadership from any company, let alone Amazon, blaming their customers on stage (even jokingly.)
cquinn: Yeah, that didn’t land super well.
rachel: I think for me it was the slide about Primitives Not Frameworks, right before the introduction of Amplify Studio, that was most jarring though.
monkchips: They’re good primitives, though, Brent.
cquinn: I don’t know what the job of the future is that’s responsible for determining which of AWS’s 40 managed databases is the correct one for a given application, but that job is not going to be pleasant.
sogrady: Also, which legal team is tasked with keeping the DOJ at bay for that as well.
monkchips: Isn’t that what the Swami Subramanian’s Machine Learning keynote was about? The one that wasn’t a machine learning keynote, but now a Data, Analytics and Machine Learning keynote.
Just put all your data in AWS, and worry about storage engines, models and stuff later. I felt like there was some moat building with Snowflake continuing its hot streak.
Which sort of answers the DoJ question. AWS does have impressive competitors, even though it’s crushing it. I mean MongoDB just beat its numbers and saw a 16% spike in its share price
cquinn: “Put all of your data into AWS, worry about it later” is more than a little self-serving as far as messages go.
monkchips: Of course @cquinn, AWS job number one is to win workloads.
kelly: I mean it’s not like egress costs are ever an issue…
rachel: I was kind of surprised to not really see egress costs addressed at all given what’s been evolving in the market around AWS (notably Cloudflare).
At some point they’re going to have to directly acknowledge their egress margin is someone else’s opportunity. But apparently not this year.
cquinn: “We’re seeing strong adoption of DocumentDB (With MongoDB compatibility).” Yes except that nobody likes it; the thing driving that adoption is in no small part “free cross-AZ replication traffic” in the way that only first-party AWS services get to privilege themselves. Everyone else is staring a 2¢ per GB charge in the face.
They attempted to address it by expanding their 5GB a month data egress free tier to 100GB. And the entire internet went around congratulating them for it without doing the math. In their most expensive regions that’s a $13.20 per month charge, so you’ll have to forgive me if I don’t throw my hat into the air.
sogrady: Ha!
The counterpoint to the DocumentDB example, of course, is that Atlas continues to grow at a healthy clip, so I could see the AWS argument internally being: “everybody wins.”
For now.
cquinn: I can’t quite reconcile the Amazonians I know and spend time with, all of whom are fine people, with this incredibly aggressive corporate philosophy of losing sleep over the fact that someone somewhere somehow who isn’t Amazon is making money from something.
sogrady: Just as I can’t quite reconcile the commercial open source folks I know and spend time with, all of whom are fine people, with this incredibly aggressive corporate philosophy that someone somewhere somehow who isn’t them may be monetizing the project in some way 😉
monkchips: Also in the data keynote I thought the serverless offerings were interesting. Suiting the industry direction of travel. Managing databases sucks, and being able to scale to zero is :100:
cquinn: Scaling to zero is a big win even if only for development and testing environments. Whether production ever scales down that far is an open question for a given customer’s business model.
kelly: For me a key announcement was Werner’s callout to sustainability, especially announcing sustainability as a pillar of the AWS well-architected framework.
Delighted to see @Werner talking sustainability in today's #reInvent keynote. pic.twitter.com/CWhXCK2EkM
— Dr KellyAnn Fitzpatrick (@drkellyannfitz) December 2, 2021
rachel: I was in a Discord channel with Liz Fong-Jones from Honeycomb, and her excitement about C7g instances and the Graviton3 processors was infectious. That was one of my favorite announcements.
And it ties into @kelly’s favorite announcement nicely. Chip design to make power usage more efficient is a nice approach to sustainability.
monkchips: I don’t think AWS has quite had its M1 moment yet, but its investments in silicon are starting to get very interesting indeed. AWS announced Graviton3 last week. Cheaper, faster, 60% less power consumed. Bad news for Intel, but very very interesting for AWS customers and the industry at large.
The ecosystem of third parties like Honeycomb and CircleCI is rallying around. Certification and co-marketing. Open source projects are increasingly well-supported. The performance improvements so far have been impressive already, but I suspect we’ll see some breakthroughs going forward. At some point developers are likely to say “why run this workload on non-Graviton instances?”
Yeah the M1 moment is the new iPhone moment
cquinn: Naming new instance families with the “i” letter for Intel, such as c6i.xlarge, means that they’re no longer the default. I would not want to be in Pat Gelsinger’s chair right about now.
Intel has always been a huge sponsor of re:Invent; they weren’t there last week. Alternately, they’re doing the Intel roadmap thing and will be showing up for re:Invent 2021 in January.
sogrady: I WAS RIGHT ABOUT ARM! (It just took a decade.)
kelly: @sogrady we never say when things are going to happen
sogrady: @kelly true fact.
(except I kind of did)
monkchips: But yes default to Graviton is coming soon to a workload near you
cquinn: ARM’s moment is here. My laptop, phone, tablet, EC2 instance, Lambda runtime environment, and serverless container service are all using arm64 architecture.
sogrady: Right, time for me to start making RISC-V predictions, it would appear.
cquinn: At this point using x86 anywhere is the aberration; I didn’t expect to see it hitting so totally, and yet here we are.
sogrady: Agreed. It happened so slowly it was fast.
monkchips: Apparently the FTC is blocking Nvidia’s acquisition. But ARM is indeed in the cat bird seat. But that means Apple and Intel in fact are. The model is quite different. This is not a franchise like Intel was.
rachel: I feel like we’ve touched on this along the way, but let’s make it explicit: what did the announcements coming out of this year’s show mean for AWS competitors?
sogrady: Well, for once they’re not facing some brand new paradigm of a service to catch up to.
cquinn: Are we sure that the reason for that isn’t “roadmap slippage” instead of a newfound focus for AWS?
sogrady: @cquinn I wonder at the causes there. Certainly could be slippage, could be the distractions of the past 18 months (COVID, leadership and otherwise), could just be rational retrenchment, could be all of the above.
monkchips: Vercel is happy. GitHub is happy. Netlify is happy. PlanetScale is happy. Render is happy. Hashicorp is happy. Microsoft and Google are still grumpily finding their way.
VCs investing in developer experience companies are happy.
monkchips: I kind of think AWS will go into higher level services in a different way. Never mind developer experience. It’s about the enterprise. I think it’s going to sell the hell out of Connect for multichannel comms, and do more stuff like that. Start owning swatches of the 5G market. Do more deals like the Goldman Sachs private cloud. I think AWS future is as an enterprise application provider. Domain specific, some horizontal, some vertical. It’s a different growth path.
cquinn: One challenge that I do see for AWS; they’re expanding into a whole bunch of areas with services around industrial IoT, Quantum Computing, Satellite communications, etc. “Sir, I am trying to build a website here; it even says ‘web services’ right there in the name.” It’s getting increasingly hard not to be distracted by the services that Are Not For Me(tm).
kelly: What, you don’t want a side of Private 5G with your web services?
sogrady: Sprawl and fragmentation are an issue, no doubt.
rachel: I do enjoy having 4 minute breaks interspersed in the keynote to catch my breath while they talk about cars and 5G. I think of their sprawl as my breathing room. 🙂
cquinn: That used to be what the customer stories were for.
monkchips: I loved that, saying hey Vodafone is nice, but look at this keynote slot where Dish says Vodafone is toast, or rather established network providers are like “brick phones”.
cquinn: To his credit, Adam Selipsky didn’t feel the need to pull an Andy Jassy style “get all the releases out in one breath.” They were evenly spaced throughout the week and between the various keynotes, which made for a nice change.
monkchips: Yeah the keynote was only 2 hours long.
kelly: Yes but how much of that is not having a house band?
cquinn: I will say that the press and analyst seating in the keynote was way off to one side, and the screen for that area had the warm-up band’s drum set right in front of it. This was either an unfortunate accident or some marvelous shade thrown by AWS at folks who aren’t customers so they’re allowed to dislike them.
monkchips: So what else did we notice. Not just sustainability theming, but even a nod to diversity and inclusion.
And story telling with historical allusions. That was different
rachel: Their efforts towards more representation on stage from different groups was such a welcome addition.
rachel: Okay, let’s wrap up. Any final takeaways or anything that you think we’ll look back on a year from now as having been the most important part of the conference?
kelly: For me it was thematically very disjointed, but that is so very fitting for a hybrid event in 2021.
cquinn: Cynically, depending upon what COVID does in the next two weeks it’ll either be the hallmark of large scale events coming back, or a horrible warning as to why they aren’t yet.
monkchips: Next year Selipsky will have made more changes. It will be properly his event and his company.
rachel: I’m looking forward to watching the evolution of the Selipsky era.
And I’m definitely looking forward to watching the primitives vs. frameworks investment paths.
cquinn: I think that customers can look at the re:Invent that’s now in the bag and not feel a creeping sense of dread that they’ve bet on the wrong horse. It’s hard to picture that, but it’s also no small thing.
sogrady: To me, the Selipsky-era was kicked off cautiously, and I’ll reserve further judgement on what his tenure holds given the disruption of the past several quarters.
rachel: Well, that’s a wrap! @cquinn, best wishes on a forthcoming negative COVID test and a return to your family!
Thank you so much for joining us for the slackchat today!
cquinn: Thank you for having me, despite the fact that my brain isn’t working for beans this week!
Disclosure: AWS is a RedMonk client, but this is an independent and unsponsored piece of research. CircleCI, GitHub, Google, Hashicorp, Honeycomb, Microsoft, MongoDB, and PlanetScale are also RedMonk clients.
Related links:
- re:Invent and Reinvention: Steve’s deeper analysis of “primitives, not frameworks”
- By The Numbers: Rachel’s quick take on some of the numbers that came up during re:Invent
- re:Quinnvent: Corey’s coverage of re:Invent
No Comments