Following up on the recent Amazon Web Services conference, re:Invent, regrettably held in Las Vegas, the three monks who attended got together to chat about what we saw and the conference and what it means for AWS, developers and the rest of the industry. The conversation is lightly edited for content and for incompatible emoji.
sogrady (Stephen O’Grady): Ok, so welcome to the RedMonk re:Invent Slack Chat. What were your respective impressions of the event?
rachel (Rachel Stephens):: I feel like the metaphor you used in your recap post about approaching it as an event streaming service was a pretty great one. re:Invent is so jam packed (both with people and announcements) that it’s simply too hard to summarize succinctly.
kelly (Kelly Fitzpatrick): This was only my second re:Invent, so for me there were a lot of automatic comparisons to the 2018 event. One common theme: the attendance numbers, the number of events, and the sheer amount of announcements are all still intimidating.
sogrady: Indeed, the event is still overwhelming. Trying to get from place to place makes you feel like a salmon.
rachel: But my high level take away was that Amazon is a machine with product updates.
I didn’t walk away with any feeling of “this is the most significant announcement” as in previous years when they introduced a new compute primitive like Firecracker.
This year felt like there were a bunch of interesting but smaller product announcements, and unrelenting march of marginal improvements to their existing products.
sogrady: I thought the best evidence of how high expectations are for AWS was a woman in the elevator at the Wynn who, in a conversation with a colleague, said “Well, there’s no Lambda this year.”
As if every year will or should see the release of a market creating product.
kelly: For me, it has been interesting to see how different folks have processed all of the announcements in the week that followed. We all have broad coverage areas, but for folks who have more specialized interests, it has to require a lot of patience to sort through all of the noise and find the information that is most relevant to you.
sogrady: On that note, what was the number of announcements even before the event, @kelly? 250+ right?
Last year we announced 124 new AWS services and features in the 30 day run-up to re:Invent (pre:Invent). This year we are already at 251 and still have 4 more days to go… hold on to your hats folks, this re:Invent is going to be a belter. #awsreinvent #reInvent2019 #seeyouthere pic.twitter.com/vw11jqtTBf
— Paul Maddox (@paulmaddox) November 28, 2019
So at least 251.
sogrady: I’m genuinely curious as to how people whose day job isn’t just keeping up with news like ours, people who have actual things to build, process all of this.
kelly: I feel like that has to help narrow down the announcements you care about.
sogrady: That’s a fair point. Part of our problem is that we cover all of it, or at least most of it. Still, 251 announcements before the event has even started…
rachel: Agreed. re:Invent is the time of year when I am most jealous of people with narrower coverage areas.
kelly: At least that you care about immediately; then you probably panic about what else is new that you need to know.
This is a great event for FOMO on many different levels.
rachel: Roadmapping from this would be brutal.
kelly: Although even for folks with more narrow coverage areas, it has to be intimidating. One of the recaps I caught this week was in Jeremy Daly’s weekly serverless newsletter, and even with that relatively narrow scope, there is a LOT to read and process:
Issue #67 of Off-by-none is out! This week, we look at the big #serverless announcements from @AWSreInvent 2019, and share plenty of great content from the community. Plus, we recognize @esh as our ⭐️ of the week! #offbynone https://t.co/FWb9bEE2oC
— Jeremy Daly (@jeremy_daly) December 11, 2019
BTW, Jeremy includes your “Five Things to Take Away From re:Invent 2019” post, sogrady (even though you do not talk about serverless).
sogrady: Ah, cool, I’d missed that. Anyway, we’ve already largely covered it, but anything notable to add re: event logistics or what the event was like?
rachel: I think just a plug for my annual request to take the event out of Las Vegas, but other than that I think we’re covered.
kelly: Does anyone have official attendance numbers?
rachel: I think it was 65,000.
sogrady: I heard 75, but who knows. It was a lot.
And second the request to hold the event anywhere – literally anywhere – but Mos Eisley. Not that I expect that to change.
kelly: I was all good with Las Vegas for the event until we had a day of all rain.
Navigating Vegas is one thing; doing it in a downpour is another.
sogrady: The problem with navigating Vegas is that you’re in Vegas. But anyway, let’s get to our next question: what were your most important takeaways?
rachel: This is my time to throw out my super amazing definition of Amazon!
kelly: Yes, please.
rachel: Amazon thus far has very little appetite for providing an integrated abstraction layer, other than smaller forays with products like Beanstalk or Amplify, and in fact they have an astounding ability to not fear cannibalization. Teams build products and solutions that openly compete with one another simply by providing different approaches to problem solving.
Amazon is not a platform in a PaaS sense, but rather they are a platform of primitives.
sogrady: No argument here; I came to pretty much the same conclusion in my re:Invent takeaways piece.
kelly: Also agreed.
rachel: Not novel. I was just really excited about the phrase.
kelly: It is a great phrase!
And for me, I was impressed by the emphasis on the tech, tools, and processes that AWS is using to run and create those primitives. For instance, there was a lot of time devoted to chips in Andy Jassy’s keynote on Tuesday. We saw Clare Liguori on the keynote stage on Thursday discussing the redesigned Fargate dataplane: Firecracker microVMs on Nitro bare metal instances. We also heard about the Amazon Builders’ Library: “a collection of living articles that describe how Amazon develops, architects, releases, and operates technology.”
sogrady: But their willingness to not fear cannibalization was on display once again. DynamoDB has been firmly positioned as a superior alternative to Cassandra, but in the face of customer demand Amazon did what Amazon does – they spun up a managed Cassandra service – competition with its own DynamoDB service bedamned.
Much as they’ve done with the Kafka service and Kinesis, EKS/ECS, etc in the past.
kelly: See, if I was building things, those are the types of announcements that would trigger my FOMO: if I have already decided to use a certain type of tech AND to use it on AWS, I now have to worry if I am using the RIGHT version of that tech on AWS.
rachel: I also think the GA of Outposts and the introduction of Local Zones will be interesting to watch. It really starts to stretch commonly understood definitions of cloud, edge, and local data centers.
rachel: Excellent point.
sogrady: Agreed, good point. Cisco has pushed this term called “Fog Computing,” which is a term to be honest that I actively dislike, but between AZs, Local Zones, and Outpost things are definitely getting foggier.
kelly: Ooo, and don’t forget Wavelength!
sogrady: As for Wavelength, I thought we agreed that this was going to be a 5G free safe zone?
kelly: I can’t help it. We saw 5G on the KubeCon AND the Re:Invent keynote stages.
sogrady: Ok, fortunately we’re out of time for 5G, so next question: what did AWS do right at the show? And aside from hosting it in Vegas, what did they do wrong?
kelly: The winnowing down of the number of events AT the conference was smart (even if it did give us 250+ pre-announcements). I especially read the keynotes as focused.
rachel: I think it’s to their credit that they introduced an open source track this year, but I would have liked to see more open source narrative on the main stage.
After the year the commercial open source community has had reacting to the threat of Amazon with hybrid licenses and other moves, I think they might have benefited from having more explicit messaging there.
sogrady: That’s a fair point; too much of the OSS work is still behind the scenes.
Although I for one am glad that licensing did not make the main stage.
rachel: Yes, I don’t think licenses need stage time, but conversations around their open source contributions need to have more sunlight.
kelly: It is; even when something like Firecracker is on the main stage, the link to open source is often unclear.
Other things missing from the main stage (at least with Andy Jassy’s Tuesday keynote):
Am I mistaken or did AWS just put on a three-hour keynote that was exclusively white men on stage? I honestly don’t understand how that’s even possible.
As a customer, I’m embarrassed.
— Eric Neustadter (e) (@thevowel) December 3, 2019
sogrady: Yes to both.
kelly: I noticed this last year, but once again the only woman on stage was one of the DJs. This includes the band (and the bands the originated the songs that the band played).
Oh, and also Ginger, who is notable for her absence
sogrady: Ha! Can you explain that quickly, for people who won’t get the reference?
rachel: Jassy had a story about a customer doing a mainframe migration where the customer discovered midway through that no current employees had the credentials.
They had to go on a hunt to find a previous employee (the aforementioned Ginger) to find them.
(I have to admit I liked the story.)
kelly: I also like the story (especially the moral that “you can’t always find Ginger”). But it was painfully ironic on such a non-diverse stage.
sogrady: No argument. In interests of time, next question: what did the show mean for AWS competitors?
kelly: Well, as you note in your blog post, Microsoft was called out quite a bit, and that is definitely of interest. On the one hand, it paints Microsoft as more of a threat; on the other, it suggests they may become more of a target for various AWS competitive strategies.
rachel: I was curious if that was because Oracle is not the same target it once was now that AWS has migrated off them completely or if it’s a response to the Windows Server license changes.
Either option felt plausible to me.
sogrady: I think it’s at least possible that it’s because the Azure/GitHub combination is the biggest looming threat, but more likely it’s a combination of reasons.
Going back to our conversation about blurring lines, I think it’s also notable that AWS is expanding its competitive footprint. Azure has Stack and Google has a big push around Anthos, but AWS is really trying to widen its broaden its reach in ways that should concern both cloud and physical hardware providers alike.
rachel: I also think it’s interesting that AWS rolled out Savings Plans (improvements to reserved instances) and improved Compute Optimization visibility. They seem like small changes in the scheme of all these technology announcements, but lack of transparency in billing is a place where AWS is consistently criticized. It’s notable to see them start to tackle that because I think it indicates a more competitive posture around user experience overall.
sogrady: Ok, last question because we all have to go: a year from now, what will be the announcement from this re:Invent that we look back and recognize as the most important?
Maybe it’s just because I’d like to see my nearly decade old prediction come true and I certainly don’t have high confidences in my answer, but I’m going to go with ARM. Those numbers were bonkers, and if ARM becomes a real server player thanks in part to AWS that would be a major shift.
rachel: I’m going to go with Outposts and Amazon’s expansion of cloud into the data center.
kelly: For me, it was less a single announcement and more the shift in positioning around “serverless” to fall even MORE in line with the definition Rachel set out in the wake of last year’s Re:Invent. Tim Wagner has a good take on this –that it is a move to draw more container devs to serverless–but it also fits with an increase in the promotion of different, non-FaaS technologies as “serverless”.
sogrady: And that’s a wrap. Thanks everyone.
Disclosure: Amazon is a RedMonk customer, as are Microsoft and Oracle. Cisco is not a RedMonk customer.