Atlassian twice shunned AWS Graviton CPUs, but now runs Jira and Confluence on them

Atlassian twice marked Amazon Web Services’ Graviton CPUs off-limits for production purposes, but recently relented and now uses the processors to power thousands of server instances that run its Jira and Confluence products. So what changed? A recent post by Atlassian principal site reliability engineers Paulo Almeida and …

[elsergiovolador]

Gasps

I thought Jira runs on morning gasps:

"Where is my ticket?"

"Which ticket you are working on?"

"Can you move your ticket, please?"

"Is this a child ticket?"

"Which board are you on?"

[FirstTangoInParis]

Re: Gasps

My recent experience of Jira is that there is lots of stuff on the screen but it tells me nothing I actually need to know. Pretty useless really.

[Pascal Monett]

"the invisible costs of implementation"

Aka, the costs that the customer supports.

[Charlie Clark]

They call themselves engineers

“In the tech industry there are only a couple of things that can make engineers try something new. Either it’s because we can do something faster or because we can do something cheaper,”

I've never met an engineer that would make that kind of statement. There interested in solving problems and making new things and making existing ones better. This is, after all, what led to Google's work on CPUs, which got companies like Amazon interested.

I'd expect engineers to understand the technical trade-offs inherent in the chips: single-thread performance, precision, parallelism and memory, etc. Anything built for x86's great single-threaded performance and massive set of caches is going to need some work on it to get the most out of ARM-based design.

[Anonymous Coward]

Re: They call themselves engineers

Then you've never met any truly awful ones....I have.

Anonymous because otherwise they will know who they are.

[Charlie Clark]

Re: They call themselves engineers

I'm sure there are but I still find it strange that an "engineering" discussion would start with money.

[spuck]

Re: They call themselves engineers

Cost is always a factor, if you're engineering for anyone making a product that has to be profitable. Apparently only engineers working with someone else's money (i.e., the government) have the luxury of not caring what something costs.

Engineering is not science: it is the art of reasoned guesswork, trying to solve a problem by using available resources. Money is just one of those resources.

[Charlie Clark]

Re: They call themselves engineers

I'm not saying it's not a factor, I'm just saying you don't start there.

In fact, I'm a fan of the kind of "lean engineering" that used to be popular but has gone out of fashion as everybody chases "scale". In Atlassian's example, it's also not really their money, but part of the calculation they use to set prices, ie. they're able to pass the price on.

In the case it sounds like they looked at the money first and decided it wasn't worth doing the work to see what needed to change to get the most out of the cheaper architecture.

[FirstTangoInParis]

Re: They call themselves engineers

Charlie Clark is correct. Out and out engineers don’t care about money. In fact I’ve heard project managers tell engineers to just do the work and the PM will find the money.

Engineers who are dyed in the art of communicating with manglement sadly have to learn to talk money. But big handfuls like a make / buy decision, or how many months of effort.

[Anonymous Coward]

Re: They call themselves engineers

"An engineer is a man that can do for 10 bob, what any fool could do for a pound"

I am that kind of engineer. My role models are Clive Sinclair and Earl Muntz, with a nod to Hyman Rickover.

[PRR]

Re: They call themselves engineers

> I still find it strange that an "engineering" discussion would start with money.

Engineering is mostly about money.

"It would be well if engineering were less generally thought of, and even defined, as the art of constructing. In a certain important sense it is rather the art of not constructing; or, to define it rudely but not inaptly, it is the art of doing that well with one dollar, which any bungler can do with two after a fashion." - https://en.wikipedia.org/wiki/Arthur_M._Wellington

https://www.quora.com/Who-originally-said-What-any-damn-fool-can-do-for-a-dollar-an-engineer-can-do-for-a-nickel

[Charlie Clark]

Re: They call themselves engineers

In a sense, this encapsulates the two dominant chip architectures we have at the moment: x86 has long been a series of improvements on an unsound base; this made it inefficient and expensive. For various reasons, but mainly compatiblity, the market was essentially captive so prices could be kept high. ARM was designed to be efficient, and thus cheap, from the start.

In another part of the same ecosystem: look at how much ASML is charging for the new generation of lithography machines…

[MatthewSt]

Re: They call themselves engineers

Faster and Cheaper are both subsets of better. Your example about Google working on CPUs was an effort to make their infrastructure faster and cheaper.

[Anonymous Coward]

Did their costs drop by 9.8%, 25%, or 34.8%? If this is on top of their October price hikes, are they now able to make a profit?

[spuck]

The article's datapoint of "over 3,000" customer workloads moved to Graviton processors means they must have plenty more than that... and yet they still think they're saving money using AWS instead of owning their servers.

Serious question: Given that they chose to migrate away from their own hosting towards AWS a couple of years ago, is this now strictly about accounting and managing cash flow? I could see how a subscription-based service would prefer the pay-as-you-go option of Operational Expenses rather than the Capital Expenses of buying their own servers.