Amazon Managed Service for Grafana | Hacker News Reader

Amazon Managed Service for Grafana

aws.amazon.com

204 pointsahachete66 comments

> With one click, upgrade to Grafana Enterprise, giving you access to a wide variety of third-party ISV plugins – AppDynamics, Datadog, Dynatrace, New Relic, Oracle Database, ServiceNow, Splunk, Wavefront, Snowflake, and MongoDB. You can also access consultation, support, and on-demand training content directly from Grafana Labs.

It's nice to see that AWS partnered with Grafana Labs on this. Both companies stand to benefit, and the success of Grafana Labs helps ensure the long-term success of Grafana.

Here's the Grafana Labs press release: https://grafana.com/blog/2020/12/15/announcing-amazon-manage...

Assuming they didnt make them an offer they couldn't refuse.

You know, work with us or we'll OpenDistro you.

This is how software is supposed to work. If your enterprise value-add can be replicated by a two pizza team and that team is cheaper than your licensing fees then it's barely even a question as to whether it's worth it. We do this all the time on a smaller scale at our company to avoid license fees. We need one small feature from the enterprise offering so we implement it ourselves to say with the free oss/community version.

> enterprise value-add can be replicated by a two pizza team and that team is cheaper than your licensing fees

This is always easier if the starting point of that two-pizza team is fairly close to "clone the version that already works, change the name, and run the deploy script".

> We do this all the time on a smaller scale at our company to avoid license fees

There is a huge difference between you doing something for your in-house use, and Amazon doing something in order to turn around and re-sell it.

> There is a huge difference between you doing something for your in-house use, and Amazon doing something in order to turn around and re-sell it.

Is there? Or is it your bias against commercialization or profit motive of modifications to free software? It seems to me that it's the same changes to the code.

I think they're talking about the scale of the tooling and integrations into existing systems required to roll out Grafana to hundreds of thousands of people.

> This is always easier if the starting point of that two-pizza team is fairly close to "clone the version that already works, change the name, and run the deploy script".

This model only works with the OSS/community versions of software. If the the only value of your hosted offering is the hosting I would expect you to get eaten by a company with cheaper hosting costs than you. I'm talking specifically about a clean-room reimplementation of proprietary add-ons that typically make up the "enterprise" version of lots of software.

> If the the only value of your hosted offering is the hosting I would expect you to get eaten by a company with cheaper hosting costs than you

By this logic Google and Amazon should just own-all-the-things because they have the cheapest hosting costs?

> I'm talking specifically about a clean-room reimplementation of proprietary add-ons that typically make up the "enterprise" version of lots of software.

I don't know what you mean here. As far as I know this is not what is being discussed. Amazon has done no "clean-room reimplementation" here.

That's not what I'm saying at all. I'm saying that if you write a piece of OSS software and your plan to monetize it is just hosting it then you're going to get eaten by companies who are better at hosting than you. Your valuable asset and your core business is your software not your ops.

Next, if the way you plan to monetize your OSS software is by selling a proprietary "enterprise" version but the features that version includes are easy to clone and your licensing costs are high then large companies will just hire people to develop their own in-house version and they might kill your business by open sourcing their version because they don't need to sell their in-house version to make it worth the cost.

Maybe they're referring to DocumentDB versus Mongodb? My understanding was documentdb was wire compatible but a clean room reimplementation of it?

It's always amusing when I see corporate engineers espouse some open source value and then follow it up with, "That's how it's supposed to be!" They totally ignore that if you entirely profit off of grafana and then destroy grafana labs in the process then they will be the maintainers of grafana and since that will go well beyond a "two pizza team" AWS will drop it.

tldr: corporate engineers have to be trained to respect and understand open source and it's just weird.

> There is a huge difference between you doing something for your in-house use, and Amazon doing something in order to turn around and re-sell it.

There really isn't. If you make your work available to the world so that everyone has the right to not only use it as they see fit but also extend it to fit their needs then you cannot complain that someone is using said work as they see fit and extending it to fit their needs.

I am not wholly convinced.

There is a considerable problem in enterprise software where trivial features are enormously expensive. If you hide all of the useful authentication features behind thousands (or tens, or hundreds of thousands) them don’t be surprised when somebody scoops you.

Value add needs to be more like added value than ransom.

But that's because the non-trivial stuff, the real meat of the core product, is free right? Almost every dual (or more) priced (core) source available (at least) is enterprises with the money subsidising it for people/companies for whom the alternative is not to use it, not to pay for it.

If you have any emotional response like that it's 'ransom', then you're probably not the target audience for the enterprise tier.

(And I don't mean that you don't need/can't make use of features hidden in that tier, just that if there are many in the same boat, then the provider misjudged the contents of that tier.)

I get the feeling you are talking about Elastic here, and I 100% agree that core authentication and security features (and sane defaults) being "Enterprise" is a bonkers decision.

Core authentication and security features in Elasticsearch have been free and not Enterprise features since last May: https://www.elastic.co/blog/security-for-elasticsearch-is-no...

Disclosure: I work for AWS

Free for some to use under certain restrictions. Therefore not Open Source (which is totally OK, I just want to make sure there's no confusion).

Sweet, I didn't know that had changed. Good decision.

Except that Amazon isn't "a two pizza team", it's a 1.5 trillion dollar company that could replicate any startup ever with 6-12 months and a hundred engineers.

I think in the business world this would be called "not having a moat." There's basically only one to proceed if your product can be semi-trivially copied by someone throwing money at the problem. You have to sell it to them for less than it would cost to build in-house. Sure they could throw money at the problem but they won't when it's cheaper to just buy it.

When you have that much money, moats don't matter.

Code commit.

It is so horrible compared to GitHub.

Don't forget their continued lack of anything to compete with Heroku (that is, hook up to GitHub, deploy automatically from master, with less than 20 clicks in the web UI).

EBS was never it, and Lambda is a different type of product (and also won't autodeploy from your GitHub master without custom CI/actions/whatever).

Then there's Lightsail, which is perhaps the lamest DO ripoff that could possibly be engineered. I'm sure there are other examples, too, but those are the ones with which I'm familiar.

It doesn’t guarantee that they can execute, only that they can easily do so with that much money. Not everything needs to be conquered nor makes sense to do so at that scale.

While I get the sentiment, this activity is what is killing open source. If you like a tool, contribute or license it to keep it a float. Don’t just rip it and do your own thing. There are maintenance costs that aren’t associated with the feature development. There are liabilities, legalese, indemnity, etc you should worry about.

Support open source.

Free software means freedom to modify it and build a paid service from your private fork.

If you can't do that, it's not free software, despite what the AGPL torch-carriers think. Freedom means free to do things that may be contrary to the desires and interests of the original authors: this is why software with "no military/no missiles/no bombs/etc" clauses in their licenses are not free software.

And why the JSON license isn't Free or Open Source since you're not allowed to use it for evil.

> We are also collaborating with AWS on its new Amazon Managed Service for Prometheus, built on the Cortex project for running Prometheus at scale.

That's even more exciting news.

But this also must have major implications for Grafana Labs, since selling hosted prometheus is also a part of their business.

Edit: Found the announcement at https://aws.amazon.com/blogs/aws/join-the-preview-amazon-man...

> Both companies stand to benefit,

I wouldn't be so sure.

This feels like a new level of gross for Amazon in terms of their previous "this open-source thing is mine now" strategy.

Not only is this taking-and-running an open source project, but on top of that, the pricing is per user? This feels like a step beyond "hosting software for you", and towards "building a business with someone else's product"

I see that this is in partnership with Grafana Labs, which I guess is a good thing, but I can't comprehend why Grafana would have gone along with this (short of some sort of "well we're doing it either way" threat from Amazon).

I don't see anything new here that Amazon is providing on top of what Grafana Labs already did. What am I missing?

Grafana has been fine with this for many years, as far as I can tell. They take a "big tent" approach to this.

Plenty of companies incorporate a hosted Grafana in their product.

A few monitoring platforms too: https://HostedGraphite.com for many years now, and my own https://HostedMetrics.com in the last few years.

I see your point, but there are 2 key differences here that I see.

1. Your product is integrating multiple tools and data sources to provide value beyond "just grafana".

2. You're are not Amazon, and I think that is actually a big point here. It's unlikely that someone will choose your product over Grafana if they are just looking for a hosted Grafana. I don't think that's really the case with an AWS alternative.

What if some of that pricing is passing though to Grafana Labs?

Do you have some special knowledge from a different source that indicates this would be the case?

If not you're just speculating that Amazon is acting in an very un-historical-Amazon way.

https://aws.amazon.com/blogs/opensource/how-aws-and-grafana-...

Down in the Conclusion section at the bottom:

"AWS, working with Grafana Labs, will be contributing licensing revenue and code to help make Grafana even better, not just for the AWS service, but also for open source users and Grafana Cloud customers."

Well that is good then.

The initial content that I had read indicated that it was Grafana Enterprise licences that would be contributed back.

This is a new step for Amazon, which is good. In this case this feels closer to a first-party AWS Marketplace app for Grafana.

Embrace (the tech), Extend (their revenues), Extinguish (the company). The financial connection of the "partnership" is now a leash and collar.

What's the survival rate of Amazon partners?

The only reason amazon does this is because their third party provider requires it.

Same with book prices. Lots of cheap e-books, and then I see "price set by publisher" and its $29.99.

While you are probably right, as far as ebooks go, Amazon keeps a higher percentage of the ebook price than a lot of their competitors do, so I would not be surprised if some publishers charge more on Amazon to cover the Amazon take.

> An Amazon Managed Service for Grafana Editor license costs $9 per active user per workspace... For customers who need only view access, an Amazon Managed Service for Grafana Viewer license costs $5 per active user per workspace

Yikes. Per-user pricing is tough for data visualization tools, since the audience for a dashboard could be really broad.

QuickSight and Tableau are also per-user pricing, seems to be the norm.

Can you share more detail? in this pricing model, you only pay if you use the service in the last 30 days. for a company with 100 engineers most of which don't author dashboards this will be a few hundred USD a month for HA, Security, Ptaching, scaling - IMHO self managed engineering time will cost more

HA on Grafana isn't that difficult. The majority of Grafana is browser-side rendered data of queries which are just passed through Grafana to the data source. It's fairly light weight.

Security wise, as far as I know AWS isn't offering anything additional here.

The argument of "but you're paying to not have to worry about it!" that is a fair argument. But comments like this seem to purposely over-state the difficulty of running a not-super-complicated tool like Grafana in order to make "only $600/month, or $7.2k/year", sound more appealing.

I agree with this. I've been running a mission critical Grafana dashboard for a long time now. It took about 15 minutes to setup, and not once in 3+ years has there ever been any maintenance required.

Moving to a different server (or cloning for HA redundancy) takes maybe 5 minutes to copy the JSON and make a crontab entry. It's pretty much the lowest headache software imaginable.

Grafana is one of the easiest internal services to host.

They follow the 12-factor app principles, provide a ready-to-use docker image, support various auth providers ( my company uses GitHub), integrates well with AWS IAM, etc.

As someone else said here, you just have to use RDS to store the state.

Its Grafana, HA grafana isn't very hard. Just store state in RDS and you're done.

In my understanding, AWS is selling the service and the hosting while Grafana is selling the software. AWS is the distributore of the software and it is paying a reasonable fee to Grafana.

IF this is the case, this is a HUGE step forward for OSS.

It is a clear and simple way for OS to monetize software, while AWS keep focusing on what it does best, operations.

This ensure long term sustainability of software and hefty profit margin for the operators.

Win-Win!

My hope is that major cloud providers standardize the price around these collaborations with open source. If Azure and GCP did the same thing, why pay them differently?

Does that not risk Amazon becoming an "app store" for OSS and wielding quite a bit of influence over it?

I don't think Amazon would mind being in that position.

Realistically though, with 3 major cloud vendors to choose from, I don't think the danger of app-store style control holds in the same way.

The App Store is fundamentally a problem because there exists only one (eg apple) or one is inherently privileged (eg android); on a PC/servers/cloud, this isn’t likely to ever occur.

Although, I suppose steam is an argument to the alternative, though they haven’t successfully abused their position afaik (like their attempt to monetize general modding communities across non-valve games)

Even if it does, I don't see how they can lock them in just their version of "app store". If it takes off, I can see GCP, Azure and others offering similar deals too.

I must say it's quite an expensive jump for a medium size company (50-100 users) given Grafana requires very little maintenance and runs smoothly in low budget instances

Right. Grafana is a low maintenance application to run -- its just a web interface in front of various databases. More interesting is managed Prometheus, which it looks like they've also announced today.

I agree, I have used Grafana for a few years and it requires very little effort. However, the dashboards I use are static and I don't back it with a RDBMS because of this. I suppose there is more value for users who plan to have more complicated configurations or want to give others the ability to build ad-hoc dashboards.

But it looks cool.

How is that to generate the same kind of situation as the one we have seen regarding MongoDB, I mean, "giving the middle finger to open source" [1]?

I know one relates to API compatibility, but those companies never end to surprise me.

[1] https://techcrunch.com/2019/01/09/aws-gives-open-source-the-...

Disclaimer: I work at AWS but not on anything related to this.

I don't think this announcement is as juicy or controversial as you were hoping. It appears this was developed jointly with Grafana Labs as a part of a longer term partnership: https://grafana.com/blog/2020/12/15/announcing-amazon-manage...

This is a middle finger to open source the business model.

This is a thumb up for open source the development model.

IMHO I will be more worried that a good piece of open source software is only commercialized by one company, by all accounts, that leaves a lot of risks unaddressed.

Lastly, to anyone claim that this hurts their feeling, I cannot help feel very surprised that one with some ideas of the business cycle and Amazon's (or any big corporations) track record, would not anticipate this. Personally, I am surprised that AWS have not started this in 2014, I guess they have been so successful that they probably cannot hire fast enough to get started.

If I want to create completely open-data workspaces, does that mean I need a $5 per viewer license for everyone that accesses it? Or would it not be possible to allow public read-only access?

Something like this: https://grafana.wikimedia.org/

What would be the drop-in metrics store on AWS to use alongside Grafana? I wasn't aware of an infludb/prometheus/etc type data store on AWS.

Edit: ha, nevermind I see they just released AWS Prometheus today as well.

Does anybody know if the managed Grafana will live within a VPC? Would love to be able to query resources without public IPs using this service.

so amazon is waiting for open source projects to mature and large have adoption so they can easily sell them to their existing enterprise customers. A lot of open source projects like tensorflow, kubernetes (which amazon is making millions from) where contribution from google's years development and RD. I know amazon keeps saying they're contributing to open source but it's easier to contribute when you're already getting revenue from that product

inhouse we use a few different data viz tools and Grafana is in my opinion subpar.

Which of the other data viz tools do you like more?

And what do you recommend?

Reply on news.ycombinator.com