LinkedIn Shelved Planned Move To Microsoft Azure, Opting To Keep Physical Data Centers (cnbc.com) 62
LinkedIn has set aside an effort to relocate its data center technology out of its physical facilities and into Microsoft's Azure cloud, CNBC reported Thursday, citing people familiar with the matter. From a report: The decision not to proceed with the project, code-named "Blueshift," marks a major reversal for LinkedIn, which announced its plan to move to Azure in 2019, three years after Microsoft acquired the company for $27 billion. LinkedIn had been using Azure for specific tasks.
The U-turn represents a setback for Microsoft, which is chasing Amazon Web Services in the lucrative cloud infrastructure market and has been counting on cloud technology and services to fuel much of its growth. Microsoft CEO Satya Nadella ran the cloud business before elevation to his current job in 2014. Mohak Shroff, LinkedIn's vice president of engineering, wrote in a 2019 blog post announcing Blueshift that "moving to Azure will give us access to a wide array of hardware and software innovations, and unprecedented global scale." Staffers started to learn of the decision not to follow through with the Azure migration last year, said the sources, who asked not to be named because of confidentiality. Executives stressed that the project was being put on hold, rather than getting canceled altogether, they said.
The U-turn represents a setback for Microsoft, which is chasing Amazon Web Services in the lucrative cloud infrastructure market and has been counting on cloud technology and services to fuel much of its growth. Microsoft CEO Satya Nadella ran the cloud business before elevation to his current job in 2014. Mohak Shroff, LinkedIn's vice president of engineering, wrote in a 2019 blog post announcing Blueshift that "moving to Azure will give us access to a wide array of hardware and software innovations, and unprecedented global scale." Staffers started to learn of the decision not to follow through with the Azure migration last year, said the sources, who asked not to be named because of confidentiality. Executives stressed that the project was being put on hold, rather than getting canceled altogether, they said.
This is especially funny (Score:5, Insightful)
The project should be named "Bullshit" (Score:1)
>> The decision not to proceed with the project, code-named "Blueshift
The project should rather be named "Bullshit"
Tells you what insiders think... (Score:1)
...of Azures reliability, performance and scalability.
Re: (Score:1)
And security. Although that Azure has no working security is now publicly known to anybody that wants to find out.
Re: (Score:2)
Re: (Score:2)
It is. And I am so incredibly pissed at MS. They give the whole CS/IT field a bad name by doing utter crap like this.
Re: Tells you what insiders think... (Score:2)
Re: (Score:2)
That is nice, but the crap that MS is doing can throw a whole national economy, or possible the whole world, into a deep economic crisis. They still do not understand that running a large cloud service does require a lot higher standards than their crappy OS or Office package has with regards to security.
Comment removed (Score:4, Informative)
Re: (Score:2)
It does apear that LinkedIn defied the old Douglas Adams trope about humans being remarkable for being one of the few species able to learn friom the mistakes of others, but also remarkable for being so disinclined as we are to actually do so.
Clearly, someone at LinkedIn remembered, and learned from, the Hotmail fiasco.
On hold = cancelled (Score:3)
Projects put "on hold" at mega corps and government are dead. It would cost more to restart later than just continue from where they are and fix whatever their problems are.
If they do eventually moved to Azure they'll start from scratch years from now. Probably won't have any of the same staff, management or goals.
Re:On hold = cancelled (Score:4, Interesting)
In this case, though, it probably means that the project is only "On Hold" until the executive(s) who were holding up the migration are fired and replaced with new Microsoft executives who are willing to be "team players" and finish the migration to Azure.
Now that Microsoft just had it's dirty laundry aired publicly, they'll want to fix this as quickly as possible. It doesn't really matter if moving to Azure increases operational costs to them or not, they'll want their "We eat our own dog food" rule enforced.
Re: (Score:3, Informative)
This is especially true if you are primarily a Linux operation.
Azure may be ok if you're primarily a windows operator, but geez, it was horrible trying to set up RHEL/Oracle there...especially with replication.
The Cloud is more expensive and support is worse (Score:2, Insightful)
For most companies the cloud is more expensive. If you're company is of any decent size its almost always better to do on-premise and/or hybrid than to go all in on cloud.
Remember boys and girls, cloud is just someone else's data center. And they have to add margin to cover the cost of it, and all of the additional services and features they support (and staff who gives you second class support when you need it).
Re:The Cloud is more expensive and support is wors (Score:5, Interesting)
Though in this case, the cloud *is* on-premise. It's not somebody else's data center, it's actually their own.
Which makes this all the more strange, Microsoft is calling the shots at the provider and the "client" side and their business imperative is "everybody should use the cloud", so you'd think for sake of appearances they'd make it happen, no matter what.
Re: (Score:3)
Everybody ELSE should use the cloud. MS would like to keep operating, thank you very much.
Re: (Score:2)
Re:The Cloud is more expensive and support is wors (Score:5, Insightful)
Though in this case, the cloud *is* on-premise. It's not somebody else's data center, it's actually their own.
LinkedIn is owned by Microsoft, but it's not part of Microsoft. The distinction is blurry but still real. Microsoft might want to sell them someday, they have to be able to operate on their own.
Remember when Microsoft had to use half again more servers to move their mail services from FreeBSD to Windows, and then also had uptime problems? No doubt they are trying to escape the same sort of thing now, and will keep chipping away at it until it's actually ready... or they sell LinkedIn on again.
Re: (Score:2)
That's what I'm thinking too. Likely they did a small-scale pilot and it didn't go well, so now they're throwing the brakes on until Azure gets its act together.
Might be a while...
Re: (Score:2)
Knowing a bunch of people in the PGs, might be forever too. There were some bad decisions made when Azure was implemented that they just can't get around. The firewalling is one.
Re: The Cloud is more expensive and support is wor (Score:1)
Re: (Score:1)
Re: (Score:2)
Having it hosted on Azure would not preclude spinning off, they'd get capital + a nice chunk of MRC.
Back in the $employer hosted email lists for MSFT, some of the largest around at the time. On BSDi. MSFT had us spend a bunch of $ for a phase 2 scaleup, then announced that they were insourcing so they could run it on MSWNT, because they didn't want the headers showing someone else's OS. We tried the NT sendmail port at the time and it was less than impressive. I suggested that we just forge the headers
Re: (Score:2)
I was told long ago that MSFT at one point had an internal email system running on a "Xenix mainframe".
Well I'm sure they did, although the word "mainframe" was obviously grossly misapplied there since Xenix has always been an OS for PCs. You could maybe call a PC running Xenix a minicomputer. I ran SCO Xenix 2.3.2 on my 286-6 with 1 MB RAM/40 MB disk and used it as a UUCP node. It was quite a bit more graceful at the job than UUPC on DOS, or Waffle BBS for that matter (which was a reasonably good implementation, I found it to interoperate correctly with other UUCPs.) AmigaUUCP on a 2500 was quite a bit bett
Re:The Cloud is more expensive and support is wors (Score:5, Insightful)
Though in this case, the cloud *is* on-premise. It's not somebody else's data center, it's actually their own.
This isn't how big companies like Microsoft work. The LinkedIn part of the business is organizationally distant from the Azure part of the business. There may not be any management in common until you reach the CEO level. Azure mind as well be a different company.
You also have to consider that LinkedIn started as a separate business. They already had significant IT infrastructure in place when Microsoft acquired them. The decision they're facing is not "should we use Azure?". The decision is "should we fix something that isn't broken?"
Re: (Score:2)
Well they have to justify capital expenditure to maintain scale in their hosting vs moving it to Azure. The Azure leadership is constantly under pressure to develop scale to combat AWS. The belief is that if Azure were to somehow exceed AWS' size, it will have 'won', even if it is a technically inferior solution. Every expedient is used to increase the putative size of Azure and to move existing customers business there to increase that size, even temporarily.
It's true that it ends up in Satya's lap at s
Re: (Score:3)
I'm all too familiar with gigantic companies and dysfunction. The key thing was generally this level of dysfunction was kept internal. If this had played out at one particular very large company I used to work for, I'd expect the following progression:
-An initiative to move "as-is" just like a normal external customer would, except for accounting differences and a harder push on "make it happen"
-When that fails, a move to use this as an exercise in "identify pain points and adjust the offering to hypothet
Migration avoidance + cloud's dirty little secret (Score:5, Insightful)
Though in this case, the cloud *is* on-premise. It's not somebody else's data center, it's actually their own.
Which makes this all the more strange, Microsoft is calling the shots at the provider and the "client" side and their business imperative is "everybody should use the cloud", so you'd think for sake of appearances they'd make it happen, no matter what.
You assume:
If you gave my team infinite cloud for free...we'd still take years to move it. We're a large company with a mix of everything. We moved a portion from AWS to Azure and it was a 6 month slog. Getting from on-prem to cloud took a dedicated team nearly a year. Also, LinkedIn was already a big player and had substantial datacenter investments. Cloud, by definition, isn't cheaper when you get their size. It's cheaper if you're small. It's cheaper/easier to dynamically scale. Every company is lured with the sales pitch of elasticity...thinking you can provision resources when an app gets popular, but more importantly, reduce them for legacy apps and "right-size" everything...spend where you're making money, save where you're not
Here's the dirty little secret...barely anyone does. It's just like how every gym is packed in January and back to normal by March and fairly empty by Nov. Everyone who makes a New Year's Resolution stating "This is the year I'll get in shape and lose weight" never really does...if they had the motivation, they would already have a fitness routine...and like most life choices, there's no reason why something you thought of due to peer pressure on January 1st will stick with you more than any other day. If you're going to make life changes: lose weight, quit drinking, get in shape, you'll do it when you're ready....as frustrating as it is to watch obese alcoholic smoker family members slowly kill themselves for stupid reasons...no amount of logic/common-sense will make them change before they're ready
Everyone "says" they'll tune their instances for load...very few do. Most just keep more servers than they need "just in case". Few companies ACTUALLY save money going to the cloud. The fees are huge. They justify it based on peace of mind because you can click a button on a webpage and get a new server, whereas on-prem, you place an order and have an employee install it. Just like staying in shape, everyone says they will get their operations in order and reduce COGS (money spent), but they rarely do. Tuning instances is work and risk...it's easier just to keep cutting a large check every month than take some risks to optimize your spend.
Finally, if you're a serious professional, you take QA seriously. For every one-line change, we require months of verification from various sources, confirming no impacts downstream...seems wasteful and extreme? well...we do real work with complicated systems that people pay a lot of money for us to do. We messed up recently and just did some quick local tests that took a week...and missed an extreme edge case...but the customer didn't....and there were financial consequences to the very angry customer and us.
So if LinkedIn has all the resources they need and migration is expensive to verify (and I would imagine they do make good money)...what's the rush? They're doing the smart thing and migrating the pieces where it makes sense and keeping the rest on-prem until they have a need. This reduces the number of variables that can go wrong. Even if the cloud services from Azure are free, downtime and migration mistakes are far more expensive...so unless Azure is providing some tangible benefit...why bother? why not wait until your next major upgrade of hardware or re-architecture?
Re: (Score:3)
Re: (Score:1)
I don't doubt your claim ... but I'd also say most companies who "save significant money by going to the cloud" are really ONLY doing so because they're using it as a way (or an excuse?) to cut head-count/salaries and fees for outside partners to come in and support their infrastructure.
For most businesses today, they can make an investment in servers optimized to host multiple VMs and then those servers will get them by nicely for quite a few years. (I've worked for several smaller companies who kept serve
Re: (Score:2)
Well, I wouldn't doubt the technical challenges for a sincere migration, however:
-I don't know how much Microsoft might be willing to make life hellish for LinkedIn for the sake of Azure
-The possibility of an 'insincere' migration
To that second point, not exactly this but I have been at a large company with an acquisition where the imperative on high was to make the acquisition use some strategic offering. So that is what happened, except the acquisition didn't change a thing but things were rebranded enou
Re: (Score:3, Interesting)
Re: (Score:2)
In my experience, the only reason the cloud is more expensive than on-premise hosting, is that companies that do their own hosting skimp on things like proper security, OS and hardware upgrades, software component upgrades, maintenance, backups, and redundancy.
In (much of) the cloud, you get software and hardware upgrades automatically, you get backups automatically, you get geo-redundancy automatically, you get redundant power supplies and internet connections.
So yeah, it's more expensive in terms of the t
Re: (Score:2)
In my experience, the only reason the cloud is more expensive than on-premise hosting, is that companies that do their own hosting skimp on things like proper security, OS and hardware upgrades, software component upgrades, maintenance, backups, and redundancy.
In (much of) the cloud, you get software and hardware upgrades automatically, you get backups automatically, you get geo-redundancy automatically, you get redundant power supplies and internet connections.
So yeah, it's more expensive in terms of the top line, but the quality of your infrastructure is higher, because you don't have to fight the CFO for every single license renewal and upgrade and DR mitigation.
Sure. If you run your IT shop like a used car dealer that's true. But in apples to apples operations it is ALWAYS cheaper to run your own data center at scale. Again, for smaller shops that might not be the case.
Re: (Score:2)
Just about nobody does self-hosting like the cloud.
How many do geo-redundancy? How many implement auto-scaling based on demand? How many position self-hosted equipment around the country or world, to improve latency? How many do redundant utility power or internet connections? Want to scale down a database server that is under-utilized? Good luck with that on-premise, they'd rather keep the over-sized server running, than spend the time to migrate.
There *is* no apples-to-apples comparison, unless you are tr
Re: (Score:2)
Just about nobody does self-hosting like the cloud.
How many do geo-redundancy? How many implement auto-scaling based on demand? How many position self-hosted equipment around the country or world, to improve latency? How many do redundant utility power or internet connections? Want to scale down a database server that is under-utilized? Good luck with that on-premise, they'd rather keep the over-sized server running, than spend the time to migrate.
There *is* no apples-to-apples comparison, unless you are truly large like Netflix. And guess what, even they use AWS. If they could really save that much money, they would self-host.
You seem to be unable to read. I said CHEAPER - three times. For fucks sake.
Re: (Score:2)
I know what you said, you said "it is ALWAYS cheaper to run your own data center at scale".
And I'm saying, it's not really cheaper, it just seems cheaper because those who run their own data centers fail to do things they ought to do. Your position is like saying that it's "cheaper" not to hire a competent IT security team. Yeah, sure, it's cheaper because you're not paying those pricey salaries. But you'll pay for it when you have a breach! So it's not really cheaper, it's just cheaper in the short term, m
Re: (Score:1)
Project Bullshit (Score:2)
There, I've corrected it for you. ;-)
Re: (Score:2)
See in TFS where it says Microsoft bought LinkedIn in 2016?
So yeah, it probably is just you.
Re: (Score:3)
Wow. Someone should warn the owners [microsoft.com] of LinkedIn that Microsoft may steal their data!
The cloud is about flexibility, not cost savings (Score:3)
Using Azure, AWS, or any other cloud service isn't about saving money. At least not once you get to any sort of scale.You can almost always host it yourself for less.
What the cloud provides is flexibility. You can spin up additional resources on demand, and shut them down when no longer needed. You can get these resources in data centers all over the world.
Whether that flexibility is important enough to justify the added expense will depend on what you're doing.
Re: (Score:1)
>> What the cloud provides is flexibility.
Cool story bro.
But that does also not work at huge scale: the cloud provider needs to plan datacentres for your big inrush.
Re: (Score:1)
You can probably host yourself cheaper (and in the case of Azure with better security) from larger SME level upwards.
The problem with flexibility is that basically no enterprise really needs it. Growing? Sure. But the shrinking part is where the cloud is better than your own infrastructure and that rarely happens.
Re: (Score:2)
I'd argue that you do (or can) save money in the cloud. This is because most on-premise systems skimp on DR, redundancy, backups, and upgrades. They keep that old Windows 95 "server" in the back room because nobody wants to spend the time and effort to update it. In the cloud, you generally get all that stuff built in.
Not eating own dogfood (Score:2)
Microsoft has form here. Back when they bought hotmail they wanted to switch the servers from *BSD to Windows. The latter at the time (arguably even now without huge amounts of hardware) wasn't even close to being up to the job so *BSD remained behind the scenes for a very long time.
Re: (Score:1)
Might help if MS finally got a security clue... (Score:1)
I mean, getting your whole cloud compromised and every customer exposed due to abject stupidity on a level that can only be classified as complete gross organizational failure will not endear you with customers that need some real level of security...
skill sets to migrate between cloud and datacenter (Score:2)
"external Azure customers" is the real story (Score:3)
The key piece from this article is:
“With the incredible demand Azure is seeing and the growth of our platform, we’ve decided to pause our planned migration of LinkedIn to allocate resources to external Azure customers,” Hiremagalur wrote in his memo.
At least a couple of ways to interpret this comment. Azure either doesn't have the capacity required to support LinkedIn or the cost of running it "on-prem" is so much better than in Azure vs paying customers it doesn't make sense...or a combination of the two.
So they decided (Score:2)
to keep their LinkedIn servers in a Microsoft data center, instead of moving them to another Microsoft data center. Good move.
Planned usage (Score:2)
LinkedIn has an almost predictable load, so why use "expensive" ondemand services for that? Even if your owner is the provider, it becomes expensive in terms of overhead at this scale as there is more virtual infrastructure. Just my tuppence.
Wild guess (Score:2)
(I don't have any connection to Linkedin other than as a customer.)
I wonder if Microsoft asked them to hold off for capacity reasons? They are growing absurdly fast.
Do whatever you want (Score:2)
Re: (Score:2)
Oh c'mon, this one's easy! (Score:2)
...change the sign on the door from "LinkedIn Datacenter" to "Azure Datacenter" and make a press release, then go to accounting and shift around the balance sheets.
Migration complete!