Not even LinkedIn is that keen on Microsoft's cloud: Shift to Azure abandoned LinkedIn has abandoned its efforts to migrate its datacenter infrastructure to Microsoft Azure four years after announcing the planned move. Citing sources familiar with the matter, CNBC reports the effort, codenamed "Blueshift," had run up against numerous challenges in the years since Microsoft acquired the professional …
Would have been fun if someone in that meeting had said "it would work if we went to AWS/GCP/... instead of Azure".
Fun, in a "there isn't enough popcorn for everybody else to enjoy that show" kind of way.
That said, it is a bit weird: lift'n'shit generally always works, but it's the one where you absolutely bankrupt yourself, as it's like buying a new data centre every year without depreciation. Makes you wonder what an utter shitshow the LinkedIn codebase is.
Maybe Azure couldn't deliver .. er .. the deliverables /s
@Michael Hoffmann: Quote: ‘Would have been fun if someone in that meeting had said "it would work if we went to AWS/GCP/... instead of Azure".
Fun, in a "there isn't enough popcorn for everybody else to enjoy that show" kind of way.
That said, it is a bit weird: lift'n'shit generally always works, but it's the one where you absolutely bankrupt yourself, as it's like buying a new data centre every year without depreciation. Makes you wonder what an utter shitshow the LinkedIn codebase is.’: Unquote
I don't recall them trying even once. They tried a lot in the early days to get to windows off of freebsd i think it was. I assume that "windowa core" was the result of those efforts trying to make a smaller OS.
But good to see yet another example of people admitting failure in public cloud instead of just doubling down and paying more just to show they can do it. Not as if MS has shallow pockets.
>> How many times did they try to move Hotmail to Exchange before giving up?
> I don't recall them trying even once.
“This white paper discusses the approach used to convert the Hotmail web server farm from UNIX to Windows 2000 .. We will discuss the techniques from the viewpoint of human engineering as well as software engineering.”
That article is very detailed and doesn't contain the word "exchange" once.
You don't know what you're posting about. By the sounds of it Hotmail was written in C++, running it on Windows 2000 instead of Unix meant it performed better.
That sort of article doesn't go down well around here!! We all know Linux is the one true OS!!!!
Well if you depend on Fork - BSD has (had)? a faster implementation than Linux at one point in time.
I remember the BSD hacker gleefully pointing out the results to me. Sadly I don't remember why it was, something along the lines of the BSD fork implementation being O(1),
I don't know if it's true anymore.
As a C++ person who writes networking software - I want to like BSD, but the antiquated userland irks me. (same as when forced to use a mac).
I'll happily deploy my stuff to (real)BSD not (Darwin) but I don't want it as a daily driver. I want Linux syscalls, I want a GNU userland. I want a package manager, but I could live with BSD ports.
yes that looks like a more detailed version of what I read back then. Specifically cites FreeBSD. I recall something along the lines of the FreeBSD install was maybe 50MB or something vs windows was several hundred megs or gigs at the time, which made it impossible(or at least real hard) to deploy the systems over the network at the scale they were at.
I would think the large reputation hit of failing to migrate a web service they'd bought onto their own azure, which they claim is the best target for migrating, would be big enough nadella and the board would insist on a deviation from their standard in house pricing. I bet they just weren't making enough headway, kind of like every other Windows release until they went to continuous...
"ongoing work to consolidate our datacenter locations that are currently spread across multiple buildings under a single roof"
Errmm... are you sure that's a good idea??
Mind you, if what they've got is tightly-coupled services spread across multiple locations which can't run independently, then they might as well lump them all together.
Why weren't they able to switch to Azure? I mean, you can just rent a server with Azure and put any stuff on it you want, just like with any VPS provider. You should be able to get anything working that way, unless they used a hardware and software platform not supported by Azure.
Switching to Cloud Native (i.e. Azure Functions, CosmosDB) you obviously need to rewrite parts of your code. I understand they weren't willing or able to do so in the time-span of 7 years. Makes me question the skill level of their engineers.
It just means they made a convoluted mess which is impossible to migrate without a complete rewrite or re-engineering.
I hate it when people make simple things difficult and complex for no reason. To me LinkedIn is a relatively simple web application which should easily be hosted on a variety of platforms, including cloud services.
On Hacker News it's claimed that LinkedIn engineeered their own freaking cloud making it difficult or impossible to move to any other cloud provider.
LinkedIn may have been started on a Beowulf cluster in Reid Hoffman's living room and they just moved it and expanded it.
Since nobody has ever said 'let's make a Beowulf Cluster out of Azure VMs' I'll add another commonly stated phrase possibly heard within the walls of LinkedIn,
'you just can't get there from here'.
They'll probably just do what they did with hotmail and put a front end load distribution configuration on LinkedIn, running in Azure, and claim its migrated to Azure is complete.
Damn, this is telling. Consider the fact that the costs would obviously be much, much less than what any other non-Microsoft owned company would pay.
It makes me think of when Google cloud "showed off" their cloud's prowess by calculating 100 trillion digits of π using their "high performance" cloud offerings. They, conveniently, never mentioned price once. I roughly calculated that I could buy all the hardware I'd need (high end Epyc hardware, 1/2 petabyte of storage), pay for an expensive hotel for three months, run the calculations, pay myself handsomely, then keep the hardware when I'm done, and it'd still have been significantly cheaper than if a customer had to pay Google for what they ran.
If Linkedin can't use Azure at steeply discounted / possibly free pricing, then what does that say to anyone else considering using Azure?
"If Linkedin can't use Azure at steeply discounted / possibly free pricing, then what does that say to anyone else considering using Azure?"
It was stated that Microsoft does NOT give discounts to other Microsoft companies and they must pay full retail prices for Azure.
So your one assumption you based your whole comment on isn't valid, ie they'd get such cheap hosting it doesn't make sense not to do it.
And remember, LinkedIn was created in 2002 and went public in 2003 and Azure first showed up FIVE years later. AWS was launched in 2006 so just maybe LinkedIn developers built their platform to scale off a clustering technology which is close to the metal and doesn't work well or at all in the virtualized environment cloud environments which came years later.
Even if they don't get mate's rates, they would get a significant commitment discount the same as any other (very) large customer could negotiate. Which begs the question then about the benefit of any company using Azure?
As someone else already mentioned, MS would have done a hell of a lot to make this work given the reputational fallout of failure. MS has also been known to significantly sweeten (I.e. buy) business, so there must have been some serious problems here.
I have no inside information about LinkedIn or Microsoft, but big companies typically can't get anything done. Too many bad people have touched the product then left. Current employees shield themselves behind work ticket queues that add days or weeks to tiny but important tasks. Problems cascade out of control.
I do have the experience and it is a bit more complicated. There are multiple reasons and one of them is that the incentives are aligned to favour new features and new products. Also, if a paying customer complains then mountains will be moved. Otherwise the amount of work you need to do and various checkboxes one needs to tick to fix some minor issue and roll it out globally is quite painful. “Bad people” exist but these days it is hard to slip something stupid into production due to many changes that happened in the past decade. To summarise it is a bit hard to iterate on a product in the big tech without substantial investment of energy, hence nobody fixes that annoying bug like a dropping bluetooth connection in Teams.
>Sources told CNBC that issues arose when LinkedIn attempted to lift and shift
There it is. Lift and shift is one of the worst ways to utilize cloud tech. It's the most expensive and most unstable method. I've been saying this for years. I know I'm not the only one who's realized this.
We keep hearing that return-to-office mandates are a way to justify real estate decisions. Isn't on-prem hosting the same thing in a way? So why isn't it getting the same treatment?
Boggles the friggin' mind it do.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
