Comment by TechBro8615
4 years ago
It's hard for me to know whether to feel bad for ES in this case. Did they bring it on themselves? Is Amazon too big and a bully?
From my perspective, Amazon has made most of its profit price gouging consumers on bandwidth after vendor locking them into their ecosystem, where they bootstrap new services by wrapping open source software with some provisioning scripts, management dashboards and cookie-cutter API / console templates. Indeed, most of this is templated -- AFAIU, for example, each AWS service autogenerates its Boto bindings and parts of its console frontend via code generators. Amazon has really mastered the factory process of churning out new services, and when they find a popular one, they can invest more resources into developing it than the original team ever could.
And therein lies the rub. If Amazon is improving the software in a way that the original team couldn't, it's hard to say that the community isn't benefiting. I think what strikes me the wrong way is that Amazon is not doing it for any altruistic reason. In fact, Amazon contributes very little to open source in general, considering how much they take from it. Compare them to Facebook (React, etc) or Google (tons of dev tools) or Microsoft (VSC, TypeScript). What does Amazon have? Firecracker, kind of? And now a fork of ES because that's the only way they could continue making money off it without violating the license a small startup put in place to stop them?
Well, good for Amazon, I suppose, but I find myself instinctively disliking them for this. I'm not sure what the solution is. Hopefully technologies like Kubernetes and Terraform will encourage big customers to become at least cloud-agnostic, if not cloud-independent. At the very least it would be great if Amazon / Google / Microsoft stopped gouging bandwidth at such absurd margins. Or not. Maybe it will be their downfall as startups differentiate along those lines. That would be ironic, coming from the originators of "your margin is my opportunity."
Personally I'm doing my part by not building anything with vendor lock-in. It's great to be able to deploy to any cloud, if you value either robustness or flexibility.
One thing which surprised me: Elastic has a market capitalization of ~$11B.
I think that changes some of the more floaty ethical concerns. This is not a David vs Goliath situation. This is Goliath vs Super-Goliath.
At this point, I'm much less interested in the drama of which mega-corp is screwing over the other. I'm more interested in: how does it affect me? When the titans are done trampling over the rest of us, which side benefits me the most?
Its too early to tell, but it seems like it'll be Amazon. The product is more open. They have a demonstrated history of great support. Yeah, they gouge us on networking and everything else, but at least they're the devil we know, and buying into the OpenSearch ecosystem has a greater probability of being the more open solution into the next decade.
Uhmmm I’m pretty sure David vs Goliath is talking about scale between competition. Saying that $11B is Goliath just because you’re sitting at $1M doesn’t mean they’re not in a crazy mismatched fight against a $2T company. In the same way you could be in a David vs Goliath situation yourself if you with $1M in wealth tried to sue someone with $25K of wealth. Everything is relative. Doesn’t mean it’s not a crazy unfair mismatch that doesn’t deserve sympathy and regret.
It may be relative but it isn't proportionate. A 11B market cap company can field a similarly competent legal department as a 2T market cap company can. I, with 25k, absolutely could not afford the same lawyer someone with 1M.
3 replies →
When there are two large entities (and I take this to mean way larger than you or your company) then you’re better off rooting for the one that at least releases code you can use.
If OpenSearch is truely open then in theory you can find another provider. But for ElasticSearch you’re stuck with them.
31 replies →
I don't root for David because he's smaller. I root for him because he's small.
1 reply →
Put another way, the $11B Goliath is .55% of Super Goliaths size.
To be realistic and fair I don't believe part of Amazon that is doing fork of Elasticsearch is $2T, they have other stuff to do.
Just like the guy that is $1M worth is not going to blow all of his money just to squash some guy. He probably has more powerful friends and maybe some connections but I don't see dumping $0.5m on lawyers just because he can take piece of cake from the other guy that will make him $10k a year.
Imagine head of the department storming into Jeff Bezos office telling that he needs 20% of all Amazon worth to squash Elastic, that would be funny. Quick calc with 20% I assume would be $20B which would be only x2 of Elastic, so I don't see something like that happening.
1 reply →
This argument is part of Amazon's PR campaign to tell devs to not feel sorry for Elastic because it's now a big company and they make money in the market. So, if you built a successful OSS and start to make money then it's ethical to clone any OSS and pushes projects out of the market because now it is "Goliath vs Super-Goliath".
If Elastic's own SaaS isn't good enough to generate revenue that its keeps investors happy, and ii didn't survive, that'd be a shame. Making anti-OSS moves to salvage things though is a disgrace.
Quite a few companies have proprietary products but have learned to make money selling services. One of them even has a 200B market cap and is called Oracle.
Nobody needs to feel sorry for anyone.
22 replies →
I mean, Elastic was successful because of license arbitrage; to complain about said arbitrage when Amazon does it is ... well, it's hard to feel a lot of sympathy.
Exactly this. Elastic search took open source code and turned it into a service to profit from it while making improvements to the code. Amazon is doing the same thing, except they’re keeping their contributions truly open source.
Quick point,.., as yours is valid... these days $11Bn market cap doe not represent cash in bank for development and R&D. It just reflects what the market think its worth. R&D , of which there has been a lot, and it continues, is hugely expensive.
After a decade and a third of QE, 11 bil is not really serious money. These days the unit of power on the capital markets is a trillion.
> they can invest more resources into developing it than the original team ever could.
I know this is a popular narrative, but as someone who works on AWS, I think you would be shocked by how small the individual dev teams are that build and maintain the services that everyone uses.
I'm not going to downplay the network effects involved. Of course AWS has a tremendous advantage in being able to standardize the customer billing, IAM, and EC2 Usage.
And there are economies of scale.
But individual AWS service teams are: * incredibly lean and focused * still have to make a profit on their own terms based on the infrastructure they build and the fees they charge customers * laser customer obsessed to solve people's (developer's) direct needs.
I understand the community's concern about AWS investment and approach to OSS. But I can assure you (though you have no reason to believe me) that the goal is never to embrace, extend, then extinguish. It's all in the service of going where the customers are, and solving problems that they tell us they have. The profits are a byproduct. The "working backwards" process is no joke. We spend a lot of time figuring out what is the right thing for customers to build, start building it, and THEN we think about how do we make money from it.
I was nodding along until this:
> and THEN we think about how do we make money from it
Do you really need to think? Looking at the on-demand pricing in US East, a m5.4xlarge.elasticsearch instance costs $1.133 an hour, while a m5.4xlarge instance costs $0.768 an hour. That's 47.53% of extra money. And like you said, it only requires a small team to build and maintain the service.
It is no coincidence that all cloud providers are trying to ramp up their hosted services for open source software, even GCP, who historically only focused on their own proprietary stack. There's a lot of money to be made.
What people often forget is that cross-AZ traffic is included in those .elasticsearch instance prices. In a 3-AZ deployment and given the frequent rebalancing/replication traffic that is a significant line item if you run it yourself. I'm not arguing it's not expensive, but when you look at the TCO it's less obscene than if you just look at EC2 pricing alone.
1 reply →
Do you really think that all it takes to operate an elasticsearch instance is to white-label an equivalent ec2 instance and execute some shell scripts on it?
If that were true, people wouldn't pay for it...
1 reply →
In the world of OSS, code speaks. A small team may not help the cause of Open Search. This article nails how code influences leadership in an OSS project: http://hypercritical.co/2013/04/12/code-hard-or-go-home
they have as many people dedicated to the capability as needed. at least the capabilities have owners, which from my experience is not trivial for an organization to achieve.
curious if others have noticed this as well (capabilities without clear owners)? what does this depend on? time? company size? both?
Could you please shed some light on how many people would be behind a product like AWS Lambda or AWS CloudWatch?
As an outsider, I would guess huge swaths of developers with a massive hierarchy. Buildings full of folks working on AWS services. I have no idea and extremely curious.
AWS Lambda team 3 years ago on Reddit: https://www.reddit.com/r/aws/comments/7a5anz/we_are_the_aws_...
at least 12 people. Of course they are in turn supported by many different teams, which also consist of who knows how many people.
Usually services like this start with one 2-pizza team.
Then, as the scale increases, and the number of features increase, so do the number of teams.
Lambda and CloudWatch are both a couple hundred people which wouldn't be surprising once you consider all the things that they do.
But I would venture a good half of all AWS services are built and run by one team of 10-12.
Hundreds. Each team owns a different part of the service.
Interesting. I wonder if the market can only support managed services when it’s provided directly by the cloud provider like AWS. I assume Elastic has to add margins to cover for their existence, which make them less competitive with AWS.
This sounds a little unfair, even if I agree with the argument that they’re free to fork OSS software and do whatever the fuck they want.
It sounds trivial to "wrap open source software", but surprisingly it is big value-add to thousands of companies. We can't just look at successful companies like Netflix to downplay the challenges of operating a service. Not every company knows how to operate complex systems under manageable cost. How many companies can really manage a Kafka cluster, let alone scaling it, for instance? Indeed, even companies that people deem powerful may screw up, if they don't get their culture or process right. Take Uber for example, for god damn five years, they still couldn't offer a service like EC2, let alone supporting persistent volumes. They still couldn't make their database provisioning on demand via an API. Their MySQL-based NoSQL solution was still based on FriendFeed's architecture and the APIs were hard to use. Yet they spent millions building a k8s replacement, building a GPU database, switching from mysql to postgres and back to mysql, etc and etc. So, yes, cloud companies like AWS buildd mere control planes to wrap open-source software, yet such seemingly mundane offering does bring values to many customers.
A key reason for Netflix to have an easy-to-operate infrastructure is that Netflix prioritizes productivity and scalability. They specifically did three things:
1. No fixed deadline, with a few exceptions of course, for platform-related projects.
2. Promotion/salary negotiation was not tied directly to release of external features.
3. A single engineer could be responsible for more than one service for the entire company, with 24x7 oncall.
With Netflix establishing such incentives, engineers naturally focus on getting infrastructure right, to the point that oncall 24x7 is a non-issue.
So, yeah, culture matters, big time.
Edit: another incentive was that a service was measured by its adoption. The more people praised it, the more successful the service would be. Requiring meetings to get buy-in for a new service was considered a sign of potential failure. As a result, every single team focused on making the value proposition of their services obvious. Path of least resistance was a given instead of a debated topic.
This is the most important point, IMO. Amazon's value add is not the software itself, it's the operation of the software. That includes a LOT of stuff, not just making sure it's running. It's security modeling and patching, compliance, DDOS protection, etc. Amazon's product is an army of ops engineers working 24/7 to keep your stuff secure and online.
With that in mind, their behavior here makes a lot more sense, and comparing it with companies who have dramatically different products, like Facebook and Google, takes a lot of effort to understand the differences and what impact they have.
You can rent a car but nobody would suggest the car manufacturer be paid nothing for the privilege just cause the rental company cracked car maintenance and how to fill the tank.
The AWS "value add" is only value add in the context of being locked into AWS in the first place.
3 replies →
AWS just has engineers full stop. No specific ops engineers.
90% of the companies do not know how to manage software. They got weird dogmas, no KPIs, no ability to measure performance or debug problems. This is why they got external consultants and cloud vendors. What is really funny how they think internally about these issues. If Netflix and Amazon was publishing efficiency numbers and we were able to compare with the bottom 95% of tech users people would be shocked. The difference between the numbers I am aware of (number of computers / engineer for example) is 100x.
As a longtime ElasticSearch cluster admin/developer and Elastic Cloud customer, I don't feel bad for Elastic in the slightest and I'm psyched about this fork.
The way they operate their cloud service leaves a lot to be desired and encourages maximum spend if you end up wanting to use it for anything demanding in production.
Not to mention the support is pretty slow to respond and rarely respond adequately.
At my current client we started with their cloud and in a few months deployed our own in Kubernetes using the official operator.
It's a lot cheaper and gives us fewer headaches this way.
My favorite moment as a customer dealing with all of the random times _they_ would take us down (even with triple redundancy) was their suggestion of operating a duplicate 3 datacenter cluster in production as a hot spare.
They must think their customers are cash machines.
7 replies →
This is really a shame to hear. There was once a an Elastic SaaS company from Norway called found.io that were pretty sharp and customer-centric. They were acquired be Elastic pretty early on[1]. I believe Elastic Cloud was built from this. I guess found.io's culture of delivering a good product didn't survive?
https://www.elastic.co/about/press/elastic-acquires-elastics...
Perhaps telling that none of the Found team are still at Elastic, from what I can tell on LinkedIn. I pay attention to that kind of stuff because (full disclosure) I operate the _other_ small customer-focused managed Elasticsearch company (bonsai.io) that _didn't_ get acquired by Elastic back in 2015.
2 replies →
Honestly the amount of negativity towards Elastic in this thread is jarring. I’ve been an ES customer (self hosted, some cloud) for years and have only good things to say about them. Maybe I’m less a fan of the number of features they try to bake in and the direction of the company towards using logs for metrics (which causes heavy disk load instead of just storing metrics in time series) but. Yeah.
I truly believe the negative voices are coming to the fore in this thread or they are paid Amazon folks (or they have a vested interest in AWS succeeding here).
6 replies →
> The way they operate their cloud service leaves a lot to be desired and encourages maximum spend
Pretty sure this is the definition of SAAS and IAAS
They all give you levers. I'm more referring to being required to overspend to overcome Elastic's incompetence (i.e., you're already aware of the levers, they're maxed out and the provider doesn't have a way forward).
Referring to my other example, in more detail: Elastic Cloud suggests operating your cluster across 3 datacenters for redundancy. This is a good idea. Then Elastic does maintenance in all three datacenters that your infrastructure is in at once and takes your clusters hard down. This is fucking stupid. Elastic's suggested solution to the problem: Operate a _duplicate_ cluster in 3 datacenters in a different region/provider. No guarantees that they won't do the same there either so it's not actually a solution.
This! The company I work for spending over $750,000 with Elastic Cloud every year and the quality of service we get from them leave a lot of be desired. I don't feel bad for Elasticsearch corp, they have done this to them selfs.
> In fact, Amazon contributes very little to open source in general, considering how much they take from it
I don’t think this is a fact. Amazon seems to contribute pretty significantly according to the pages [0,1] they put out that describes their contributions. Not to mention their membership in OSS foundations like Linux Foundation. [2]
You have the caveat about in relation to benefit they gain, but that’s pretty hard to measure. And I think isn’t really a good measure.
I’d like to learn more about why you make such an absolute claim and maybe you have some better measure.
I remember back in the 90s when big orgs (Microsoft, IBM) didn’t contribute to open source and can’t even think of any big orgs today that don’t contribute to open source. Even Oracle has big open source projects.
[0] https://amzn.github.io/ [1] https://aws.amazon.com/opensource/ [2] https://www.linuxfoundation.org/en/join/members/
The absolute claim, aside from being at home in a rant on HN, comes from a cursory glance at https://github.com/amzn, weighted by contributors and popularity, and compared to companies of similar size. Google, Microsoft and Facebook all build and maintain multiple open source projects that are hugely popular with people who use them outside of the company sandboxes. For example, people benefit from React without Facebook gaining much directly. (Facebook! If Facebook has any redeeming qualities, it's their open source contributions to the frontend ecosystem, although I promise you I could ascribe malicious intent to those as well...) Contrast that with Amazon. On their GitHub page, I see a few obscure projects amongst a bulbous array of AWS SDKs.
To the sibling comment that asked about Firecracker -- I think Firecracker is awesome, and I did mention that in my original complaint. They even created it themselves! Well, a team of amazing engineers in Romania did. I have no personal insight into the matter, but it seems like they operate relatively independently from the AWS profit machine. Good for them too, it's incredible software. But I'm sure if they were to tell the story of how they got buy-in at Amazon to open source it, the same themes would come up -- how does Amazon benefit from this? In the case of Firecracker, the more people test it / harden it / run Doom on it, the more value Amazon can provide on its serverless platform. So again, unlikely to be purely altruistic intentions... but that's not to say there's anything wrong with that. I just find it all a bit distasteful in aggregate.
When I think of Google I think of Go, Angular, protobuf, Bazel, Dart, Flutter, Android, Chromium, Kubernetes, Tensorflow, etc.
When I think of MS I think of C#, TypeScript, VSCode, .NET
When I think of Facebook I think of React, Flux, Jest, PyTorch, GraphQL, Haxl
When I think of Apple I think of Swift, WebKit
When I think of Amazon...I can't really think of anything
11 replies →
> Well, a team of amazing engineers in Romania did. I have no personal insight into the matter, but it seems like they operate relatively independently from the AWS profit machine.
Amazon and AWS is a massive multinational corporation with development teams around the world. Including Romania, where we had a dev center for a very long time: http://romania.amazon.com/#/
There is no such thing as the AWS profit machine. All dev teams around the world operate with similar levels of autonomy and responsibility. It's just that some of them are working on super internal systems, some on super external, and some open source. Some make a ton of money, and some don't make any, but are beneficial to the overall developer experience/ecosystem, and so make sense.
They didn't even create firecracker themselves. It's a fork of Googles crossvm.
A better URL to look at is https://github.com/enterprises/amazon.
2 replies →
https://github.com/aws
https://github.com/awslabs
To me it looks like they mostly contribute to projects that are SDKs to use AWS.
Yeah and also what about projects like Firecracker?
https://github.com/firecracker-microvm/firecracker/
I have no idea if Amazon does this or not, but maintaining forks of projects is no fun, so it's in a company's best interest to contribute bug fixes and improvements that aren't part of their secret sauce.
Did the BSDs do it wrong, too? Apple uses a lot of FreeBSD software that they turn around and sell for profit. How about PostgreSQL, didn't Amazon fork that as well? My point is, there's nothing wrong with forks nor companies forking if the license allows for it. It's up to the developers to choose an appropriate license to not be forked/ripped off, if they so desire.
I personally am against modern day corporate America, but I can't blame them for this. The software is given away free/libre/gratis to be forked by whomever.
Perhaps to combat this, one should choose a non "Open Source(TM)" license, but a source availbe license. E.g https://mariadb.com/bsl11/ (not my personal favorite, just an example).
Also, I definitely agree with/do the same:
> Personally I'm doing my part by not building anything with vendor lock-in. It's great to be able to deploy to any cloud, if you value either robustness or flexibility.
My favorite "conspiracy theory" is that AWS intentionally creates stupidly verbose and numerous headers in all of their APIs just to up the bandwidth usage a few bytes per request at a time.
You don’t pay for the bandwidth of API calls to AWS
It it quite funny, elasticsearch is also kind of wrapping an open source library (Apache Lucene) and selling it as their own product.
Very good point.
Could Elastic's business model survive if lucene adopted the SSPL license that Elastic has, saying it's the "spirit" of open source?
That is an interesting question.
The problem is, that AWS is not really sharing the internals of their fork. Big corps always say that there's no point in open sourcing their fork, because it depends so much on internal systems. Which might be true, but in case of AWS and infrastructure the possibility of an AWS competitor similarly implementing those internal services would probably help the market...
MySQL for example took a proprietary API (the mSQL API) and implemented an open software for it. Then Oracle bought MySQL. MariaDB started and it quickly dropped full compatibility. Options for the developers/users. Great.
AWS/GCP/Azure on the other hand distorts the market. Just as Google/Apple/MS fund Chrome/Safari/Edge development from other sources, thereby distorting the market, forcing browsers to be free as in beer. (Which arguably hurts the market.)
1 reply →
People want to pay for services, not software and licenses. They want turn-key solutions that are available via API and GUI, instantly and on-demand. This is the fundamental reason why AWS is so successful and the demand is constantly proven with every new product launch.
Elastic (and other vendors) complaining about this instead of using it for their own success is a problem of their own making. At least a few companies are finally learning.
What ES wants of course is for Amazon to give them a cut of revenue from hosting ES.
We already know Amazon isn't interested in doing that (either at all, or at whatever price ES wanted, we don't know that).
They had no legal requirement to when ES was open source. So ES changed the licensing to no longer be open source.
So, Amazon could... a) decide to give ES a cut after all, b) decide to stop hosting ES, or c) fork the last open source version.
I don't think anyone is surprised they chose c? Presumably ES isn't either? Maybe ES thinks this will be good for them/bad for Amazon anyway, because they are hoping potential customers will abandon the Amazon fork and stay on the original ES fork?
Not sure why they'd be confident in that exactly. Maybe they know what they're doing.
As users/customers, we would rather have a choice of hosted vendors/platforms, and that it remain un-forked (so we can use/write software compatible with either vendor/platform). Competition is good for us as users/customers, that's in fact one of the reasons we choose open source, so no one vendor can set the hosting price all on their own without competition. We want to be able to choose among competitors for hosting, based on price, customer service, performance, uptime, whatever.
But ES didn't want that, they didn't want hosting competition to exist -- at least not without permission and agreed upon cut for them -- because, I guess, hosting was how they planned to make money as a company to fund development as well as profits for investors etc. So they changed their license to no longer allow it. So of the possible outcomes remaining... this one seems as good as any for the user/customer, I guess?
So, when you say "I'm doing my part by not building anything with vendor lock-in" -- I'm not sure which course you are suggesting. In fact, between ElasticSearch and new OpenSearch fork.. it's OpenSearch that is the one without vendor lock-in, right? OpenSearch is Apache licensed, and can be hosted by any vendor and still forked by anyone . It's ElasticSearch that has a license limiting what vendors can host it (without permission of ES), it's the one with vendor lock-in, right? So not building anything with vendor lock-in means... ?
Good interesting points. Now Amazon will be the good guy because they will run open-source version, whereas ElasticSearch is not, if I understand you correctly.
No single capitalist wants competitive markets. They want monopoly, for themselves. It is only when they don't have the monopoly or an easy way to get it that they cry for competitive markets. And that is good of course.
How is it price gouging if the price is on the tin? It isn't like there's a "surprise" as to how much they charge, and it isn't like there aren't a dozen alternatives including DIY.
I'm the first to say that AWS is too expensive, and I vote with my wallet (and the company I work for by proxy). But I'll never claim that there's any gouging involved.
Price gouging is the practice of using outsized leverage in a particular market to charge excessive prices. Like snow shovels doubling in price after a snow storm. Or $10 water bottles after a hurricane.
So for AWS the term is arguably correctly applied.
But I'd be more worried about the market if AWS was artificially undercutting pricing because it would kill the incentive to create competitors or innovation in the space.
> Price gouging is the practice of using outsized leverage in a particular market to charge excessive prices. Like snow shovels doubling in price after a snow storm. Or $10 water bottles after a hurricane. > So for AWS the term is arguably correctly applied
What outsized leverage have AWS had for a decade? There are multiple competitors at different levels, AWS are just better in terms of coverage/redundancy and amount of services.
1 reply →
> So for AWS the term is arguably correctly applied.
Can you elicit the argument by which this is correctly applied?
1 reply →
"without violating the license a small startup put in place to stop them?"
Elasticsearch was first released over a decade ago. ElasticSearch, now just Elastic, the company was founded over 9 years ago and now is public. Are they still a "small startup"? If so when does a company graduate from that status?
> Amazon is not doing it for any altruistic reason
The beauty of OSS is that motives don't matter. If Amazon contributes and it's not detrimental in someway to the code, then it's a plus for anyone else who wants to use it.
Precisely! While their business itself may need to be broken up, a community governed OSS project isn't bad for OSS when the alternative is a proprietary license that gives a single corp the ability to not contribute back or be exposed to virality.
All this being said, progressive corporate taxes seem more enticing year after year.
When a product that was previously Open Source changes to a non-open license, it's not uncommon for someone to pick up the last Open Source version and fork it, and release that for others to use and collaborate on. That's always going to be a good thing; that means people who care about Open Source licensing continue to have a version to use and collaborate on.
Is it really price gouging for bandwidth? Or is bandwidth just really expensive in general? I honestly don't know. I would assume if it was actually much cheaper one of the cloud's would undercut the other to get customers.
It's absolutely price gouging. I'm not going to rant about this for the 100th time, but at least I'm in good company [0]. Do the math on the cost you pay if you saturate 1gbps for a month vs. the cost you pay for 1gbps IP transit at basically any colocation provider.
Really this is the secret sauce of the cloud. Create new abstraction layers where you can charge for logical separation on a physical basis. First VMs, then containers, then serverless... Would be cool if somebody did it with bandwidth (looking at you, Cloudflare). Why can't I buy an elastically sized pipe? Why do I need to pay for the stuff I put through it instead of reserving a size for the time I'll need it?
[0] https://twitter.com/eastdakota/status/1371252709836263425
That Twitter thread is only including pure bandwidth. What about all the highly redundant networking equipment ( firewalls, routers, switches, Nitro offloading, DDoS protection, attack detection), software for all those abstractions you get ( VPCs, subnets, security groups, vpc peerings, Elastic IPs etc. ) and engineers? None of what i listed you pay for directly, and bandwidth seems to be the most reasonable product to lump it all in.
It's like going to a restaurant and complaining about the price of steaks because beef should cost a lot less. There's a ton of other things involved, and yes, they probably have a decent margin, but not as much as you initially imply.
1 reply →
boy howdy, with all the flak i hear about this and the awesome talent in tech, you'd figure an entrepreneur or 1000 would take a stab at this, make it better, charge less. apparently there's gazillions to be made by even charging 50 percent of what AWS does.
so, when should we expect this gloriously efficient competitive market to kick in to action?
my guess is that the AWS ecosystem, despite "price gouging", is simply the best and will be because this is really hard, non-glorious engineering, where solid reliability actually matters. anyone who wants to can go ahead and co-lo, so, whatever. people who want cloud will pay, and those who can't or won't, will not.
8 replies →
You're free to go build your own cloud and sell it for cheaper. I'll wait.
I'm kind of surprised that people are this upset about how much AWS charges for bandwidth. They may charge more for bandwidth than a colo would but they're not a colo. A colo you get a network port and -thats it- you provide everything else yourself, with its attendant cost, and you roll that up into your total bill.
If a colo provides you a 1 Gbps connection if you use less, you don't get a refund. And most of the time you don't get 24/7 saturation, or you get charged on some 95th percentile billing, and their networks are almost always oversubscribed anyways.
AWS is trying to disincentivize using it as a dumb pipe. They want you to use it smartly and if you just want to push static data there are much more cost effective ways to do it, such as CDNs, which are more cost effective for both you AND AWS.
Comparing AWS bandwidth costs and Colo and even other clouds like Oracle isn't fair because different things are associated with that cost.
It really is price gouging - bandwidth is actually cheap.
A couple of comparisons:
Oracle Cloud give you 10TB of bandwidth for free, with overage charged at around €7.5/TB.
You can rent a VPS from the likes of Hetzner, and they will throw in 2-20TB of bandwidth for free, with overage charged at something like €1/TB - AWS charge an eye-watering €125 for each TB!
I think the reason the big 3 (AWS, Azure, GCP) still charge such huge amounts is that they profit so greatly from it, and there is more than enough business to go round.
You know you're greedy when Oracle is cheaper
> I think the reason the big 3 (AWS, Azure, GCP) still charge such huge amounts is that they profit so greatly from it, and there is more than enough business to go round
Or that their networking does vastly more and is of better quality? Hetzner can't provide you with everything around a VPC that AWS do. Just a tiny example - peering VPCs across regions, which is free.
I would say that it's two simple factors:
1. High bandwidth charges are an effective form of lock-in. Once your data is there, it's prohibitive to move it out again.
2. Bandwidth use is very poorly understood in many businesses, compared with simpler metrics like storage, memory, and CPU. AWS can run razor-thin margins on things that people easily compare to on-prem or VPS-style offerings, and then make the money back in areas like network traffic, fine-grained monitoring, and other items that as less obvious.
I totally agree, but was stunned to see DigitalOcean charging $0.10/GB for outbound transfer for their new (quite cool imo) apps service. You do get 40GB-100GB included, but it means it's unusable for bandwidth heavy apps. They include 1TB (which is pooled across all VMs) and 0.01/GB on their standard platform.
2 replies →
Why would players in an oligopoly undercut each other when their implicit agreement around pricing makes all of them richer? Also, second tier cloud providers like Oracle give deep discounts and still can't compete with AWAZGO so pricing isn't necessarily a main competitive advantage.
Oracle's problem is that nobody wants to work with Oracle. If I was managing a high-bandwidth service I would avoid Oracle Cloud purely to avoid legal risk to my clients.
1 reply →
Keep in mind bandwidth gets cheaper as AWS gets bigger. If you are some random tiny colo provider, people don't necessarily care to peer with you unless you pay them for the privilege. If you are originating 20% of internet traffic, now people need to peer with you or their customers won't have a great experience.
If you think AWS is expensive, give AZURE a go and be appalled.
IP transit costs something like $350-$700/mo for a Gbit. Amazon are certainly getting better rates, so even with equipment costs I doubt they're spending much more than $0.005/GiB. Their pricing starts at $0.15/GiB. (Not to single out AWS, the other big providers are much the same.)
I'm below $100/1Gbps 95% with more than one provider as a small player
4 replies →
There are also ways out of vendor lock-in. Alternator comes to mind as a way of migrating DynamoDB workloads out to other cloud vendors or your own servers: https://docs.scylladb.com/using-scylla/alternator/
The main thing I look at in this situation is their approaches to security. ES decided authentication is a paid-only feature implemented via closed source proprietary code, and the result has been countless PII leaks; now sure you could say that’s the developers’ fault for making the endpoints internet-accessible, but when the system has been designed from the start to both be insecure and hold PII, you have to place some blame on the provider as well.
Amazon on the other hand developed a free and open source auth plugin and anyone is able to deploy it no problem.
There is absolutely nothing “cloud agnostic” about using Terraform. Every provisioner is specific to a cloud provider. If you are at any scale, moving a k8s cluster is the least of your issues.
It seems reductionist to say Amazon primarily wraps around open source. What about EC2? S3? Glue? DynamoDB? Many of the services that provide the most value are services Amazon has built out.
Om top of these, many of the core services that AWS themselves rely on, like SQS, SNS, Kinesis, Lambda, Cloudfront, ECS, Fargate, Elastic Beanstalk are mostly homemade
EC2 uses the KVM hypervisor.
This is Amazon's playbook. Make a direct competitor and squeeze the originals out. They did that with jewelry early on and then anyone they couldn't buy out they would under cut until they capitulated like diapers.com The Everything Book by Brad Stone goes over this in detail. Clearly anti-competitive monopolistic actions are taken constantly by Amazon. The only reason they aren't trust busted is because the common line of reasoning is that consumers pay less for goods, but this is being looked at because shouldn't competition be lowering prices. IE if Amazon hadn't killed diapers.com, wouldn't diapers be cheaper overall? And the answer is they should be, but the government hasn't caught up. Once they start getting into the weeds, they'll see example after example of monopoly behavior destroying competitors and ultimately raising prices on consumers.
Here are the main reasons to make open source software:
To provide something for free, public use
To get the world to help you maintain your software
A reason to not make open source software:
You want the exclusive right to offer that software as part of a paid service
This is not the first instance of a company not understanding the last point there, and it won't be the last.
You described the cloud lock in model very well, especially the bandwidth part. What they charge for outbound is nuts, and the other clouds are not much better. Inbound is free of course to make it easy to send your data in but costly to get it out.
As someone who tried buying services from ES and had to deal with their smug sales people that had a total disdain to those who wanted to give ES money, I am happy I will never need to deal with them in future.
Coincidentally, AWS hasn't open-sourced anything that they use internally. Zilch. Nada. And yet they are using 'open source' developed by another firm (smaller is inconsequential here) to market themselves.
IMO, FOSS licensing is completely broken. Its definitions (of what is free/open) are from a boomer's era that is no longer sustainable. At least I wouldn’t want any of my FOSS projects become “corporate strategy” of any particular proto-nation.
That made me wonder which megacorp did? Google published a lot of papers about their architecture, but no code. They have many open source projects modeled after internal tools (eg. bazel), but there's no search/GFS/mapreduce/monorepo OSS project from them. MS has VSCode, that's great, and they even open sourced .NET. But Azure is a full black box, just like (almost) every Windows component. (Finally calc.exe is OSS!)
> Coincidentally, AWS hasn't open-sourced anything that they use internally. Zilch. Nada.
Firecracker. s2n.
Not much, I agree
gplv3 was supposed to fix this though no?
What's worse is the second order repercussions. Future open core/open source SaaSes will go straight to something like the Business Source License or the MongoDB license instead of traditional libre licenses. Amazon has done an incredible amount of damage to the open ecosystem.
Good. So the VCs will go back to playing like in the old days of shareware software licenses. It'll be good.
Only good one I can think of is Amazon Java, not sure how essential it is though
I am with you. The release could read:
"having subsumed the opensource version of ES, we are now relicensing, calling it our own, and would be really happy if the opensource community would lie to contribute, because actually we don't totally understand how this product works. Many thanks to all who help us"
The issue with AWS version are many fold, but the main one is that it forces extra usage of expensive EC2 units, for the following reasons:
1. Blue/Green updates -> Start a new cluster with new version, lock current cluster, copy over all the data (can take over a day), at the same time write to both clusters, when finished, lock both clusters while endpoints are swapped over, unlock new cluster, trash old cluster. During this process Customer pays for both. - Solution, done properly, fire up new node with new version, swap it in, wait for 'green', take out old... wait for 'green'.. rinse and repeat. Result.. system never goes down, endpoints remain the same, less cost to customer.
2.There are a minimum of the folloowing node types: - Master (small) x3 - Data (heavy on storage, medium on memory) - Hot Data (very heavy on memory as shards have to be held in memory) - Coordinating (query) nodes, heavy on memory, light on storage (cos there is no real storage) - Ingest (same as Query). - voting only.. tiny - AI/ML heavy on memory and storage.. cos they do real work In the AS world they have: - Masters - Data - Hot Data (cos they are really pricey) The Data nodes do all teh functionality of the data, ingest and query nodes. Th emain query always gores to a data node, while it passes to other data nodes to get the data out, then aggregates locally to it... so its incorrect usage of the system. A data node should only ever deal with its own data, and pass the results it finds back to a node away from the data.
As your system is squeezed.. you add data nodes... not coordinating/Query or Ingest nodes, which you probably need. But thats more money into the coffers.
3. Their userhasging function is also old (2011 vesrion of bcrypt) , and fails. Any static password produces different results everytime you use it.. at least on the current opendistro. So you are forced into either not using security, or using proper security, which can be cumbersome across the cluster (if rolling your own). however.. on the cloud version they have base level security working, so thats Ok.. they arent using their own 'open' software.
I could go on.. but its all flaky, and misunderstood at the core developer level. Toput on record, I have spoken twice to teh core development team to describe how updates should happen.. the last time 18 months ago (by 'spoken' I mean a face to face video call) .. so they know hoe it should be done.. but dont do it.
4. I have noticed that there are some functions/methods available on the setup in Amazon Linux, that dont exist on other Linux versions (centOs/Debian) that are security related. AWS Linux is 'lifted' from Redhat.. so another piece of software that they didnt write.. but obviously Redhat are happy with this. Maybe they got the licensing deal sorted .. who knows.
basically gores like this:
"Who would actually install, in production, what essentially is very close to pirated software?"
Roll your own.. its a bit of grunt work up front, then 50% the cost of the cloud version.
So VSCode, which is basically slomo Sublime Text for retards and Typescript which is usable only for prototyping are a good enough compensation for having most of their income from Linux (on Win and Azure) after all the almost criminal FUD they poured on Linux[0]? Come on, get real.
0. https://en.wikipedia.org/wiki/Halloween_documents
No need to feel bad. The CEO sold a lot of his shares prior to the license change.
> Is Amazon too big and a bully?
Is Starbucks too big and a bully? Sure, they force Mom and Pop shops to close by out-competing them, and that sucks. But bullying? I believe that's just Capitalism.
Or say you're 6'6 and weigh 230LBS and you join a football team full of people who are 5'9 and 175lbs. Are you a bully just because you're bigger?
ES basically handed them a platter with a goose laying golden eggs and a sign that read "Free Goose" and hoped they wouldn't try to make money off the eggs.
> Is Starbucks too big and a bully? Sure, they force Mom and Pop shops to close by out-competing them, and that sucks. But bullying? I believe that's just Capitalism.
I'd say it's more like a mom-and-pop coffee shop giving free coffee to patrons hoping to make money on cookies, and Starbucks coming in, taking the free coffee, opening a nice stand right next to the shop and selling the coffee they got for free.
I'm having trouble justifying this ethically.
How is it unethical if ES chose to explicitly allowed it with the license they picked? The analogy would have to include a sign under the free coffee that literally said "Feel free to sell this free coffee for profit!", for it to be accurate.
I don't see how it's unethical if you do something that someone said was perfectly fine to do. "Obvious chosen outcome" comes to my mind long before "unethical".
You could always look at teh Oracle version (wait.. please... ) They started selling consultancy.. the solution was an Oracel DB...
To expand across the world they franchisedit.Th eagreeement was, you take te hfranchise, if yoou hit x target in 4 years, we will pay you y amount (large) ...and take it over.
That worked well. DOnt see any reason why Starbucks, or anyone else could nottake that on board, then everyone is a winner.
If they're giving you free coffee, and you sell the coffee you got for free... where's the harm? Starbucks is giving it away. Why would anyone buy your coffee rather than get it for free from the source?
Now, assuming Starbucks charged you for the coffee, and you then re-sold it, this should also be fine. Starbucks is charging you presumably a rate with which they can recoup their costs. But if they are selling it below cost, they are clearly putting themselves at risk. A lot of businesses take this kind of risk as a strategic part of their business, like with making the coffee free. But you have to do it in a way that a competitor can't turn around to their advantage, or you're screwing yourself.
Enter the concept of "not for resale". If a seller enters into a contract with a buyer, that contract can stipulate that the buyer can't resell the goods. Starbucks could theoretically require you to sign a contract saying you will not resell their coffee. That's pretty standard with licenses, even software licenses.
ES must have known that their license did not forbid reselling. Yet they based their business model on this resellable coffee that they were giving away in order to make money on cookies.
Is it Amazon's fault that ES chose a business model where they were selling coffee at a loss? Does Amazon have an ethical responsibility to keep ES's business afloat? Should we find any business unethical that tries to undercut the customer base of a rival, or take advantage of a rival's shaky business model?
I think you have to come up with a whole framework for ethical competition, because one rule at a time isn't gonna capture it. (But I also think Capitalism is inherently unethical)
I think the claim that Amazon is winning through "vendor lock-in" is pretty silly. Honestly anyone who can't quickly migrate the stuff they're hosting on AWS onto one of the many other cloud platforms is pretty bad at DevOps. If you're using K8S/Docker/etc it should be trivial. But even if you're not, the vast majority of AWS offerings were either built to be API-compatible with other existing tools (e.g. postgres-compat Aurora RDS), are literally identical to other services you can self-host (e.g. ElasticSearch) or others have built services compatible with AWS services (e.g. DigitalOcean's "Spaces" aim to be API identical to AWS S3 – you can literally use Amazon's S3 client libs to interact with various S3-compat services from other clouds).
It's not "lock-in", it's providing a great all-in-one solution. You can host everything you want to host on AWS, which has good stability, good latency, etc. People are locked in because the DX around using AWS for everything your platform needs is just better than other platforms / having different services on different cloud providers (at least for many people).
If you're not using any of the AWS services, that might be true but then you're also leaving a lot of potential on the table.
If you're "cloud-agnostic" and could migrate away from AWS in the blink of an eye then you're paying for an overpriced VM offering and should probably migrate to a cheaper hosting provider immediately.
> If you're "cloud-agnostic" and could migrate away from AWS in the blink of an eye then you're paying for an overpriced VM offering and should probably migrate to a cheaper hosting provider immediately.
No this is what everyone suggesting this does not get. The offerings are not equivalent.
There are a baseline of services that the cloud providers offer that can be made functionally-equivalent. It's not just EC2 but more like EC2,S3,Lambda,RDS,DynamoDB,ECS,EKS (plus some others and of course the other cloud's equivalents). The secret sauce is in the APIs and permisioning and all of these available within the same VPC (talking to each other without paying bandwidth costs).
"Cloud-agnostic" has _never_ meant "just VMs". Some of these services are majorly hard to duplicate on your own VMs as well. Feel free to implement "cheap VM hosting + S3" and burn cash on transit costs.
Cheaper hosting providers do not give you this by miles.
3 replies →
I buy the main reason being the all-in-one solution; the comprehensiveness is attractive. However, I think you're underplaying the lock-in: migrating clouds is non-trivial - mostly due to stuff that's not running in k8s/docker/etc; stateful apps (Postgres, etc), and or just static data like s3. This takes time, and careful planning and sometimes downtime - and is mostly avoided due to it being hard.
I specifically talked about state and storage.
You're conflating the difficulty of migrating a complex system to anywhere with "vendor lock-in". It would be harder to migrate an Aurora RDS Postgresql instance to an Aurora RDS MySQL instance than it would be to migrate from Aurora RDS (postgres) to a hosted postgres anywhere else.
That to me screams "not vendor lock-in".
Do you know can I migrate my sophisticated security setup easily over? Including users, groups, roles, policies and instance policies.
Everything on AWS is accessible via API, which means you can easily automate the migration process. So do that?
Obviously the more complex a system is the harder it will be to migrate, but that has nothing to do with Amazon trying to "lock you in" and everything to do with it just being a complex system that there is no industry standard solution for.
1 reply →
The lock in is real, but they are also the best cloud option and its not even close in my opinion.
What would you say they lock you into? What is AWS doing specifically that makes it harder to move between providers than if they were not doing that thing?
3 replies →