Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Cloud Technology

Rideshare Giant Grab Moves 200 Macs Out of the Cloud, Expects To Save $2.4 Million (theregister.com) 82

Singaporean super-app company Grab has dumped 200 cloudy Mac Minis and replaced them with physical machines, a move it expects will save $2.4 million over three years. From a report: Grab is Southeast Asia's leading rideshare and food delivery outfit and therefore needs to build apps for iOS to connect with customers. In a Thursday post, the company explains it builds those apps using Continuous Integration and Continuous Delivery/Deployment (CI/CD) infrastructure that runs on Apple Mac computers.

The company started with a single on-prem Mac Pro -- its post shows 2013's cylindrical model based around an Intel Xeon processor -- but eventually reached over 200 Macs, running in the cloud at an unnamed US cloud provider. "At the beginning, it was a no-brainer to rent when our demand for macOS hardware increased from 1 Mac Pro to 20 times that size," Grab's post explains. "However, when that grew to over 200 machines, the total cost became significant."

This discussion has been archived. No new comments can be posted.

Rideshare Giant Grab Moves 200 Macs Out of the Cloud, Expects To Save $2.4 Million

Comments Filter:
  • Those ridiculous Mac Mini servers were the worst idea ever.

    • by registrations_suck ( 1075251 ) on Friday November 07, 2025 @11:10AM (#65780354)

      I dunno about the servers, but as a desktop machine, they're pretty awesome for their intended use case.

      I replaced my 2014 unit this year, after 10 years of continuous use. I could have kept using it too. But I'm expecting to be dead soon and I don't want to leave my wife with an old machine, so I replaced it this year with a new one. This would will easily keep her "in business" for another ten years.

      I still have its predecessor. I'll probably set it up for our 9 year old to use.

      • by pcaylor ( 648195 )

        > But I'm expecting to be dead soon

        I hope you are wrong about that.

      • Random guy on the internet here is always glad to see your posts.

      • In 2021, I chucked my old desktop for a Mac Mini, and earlier this year, went to a M4 Mac Mini due to concern about embargoes/tariffs. No complaints.

        My recommendation is to definitely max out RAM and CPU, and get the 10gig Ethernet option. You can boot the Mini from a TB5 SSD, which is about 10% slower than the internal and use that, and Apple Intelligence can't be used (which is something I don't bother with anyway.)

        Overall, zero complaints. Nothing is perfect, and you do trade one platform's annoyances

    • by Z00L00K ( 682162 )

      Even those mini computers often have better performance than laptops.

      • For many home users, "performance" doesn't even matter. I mean, how much "performance" does someone need for a web browser, email, Word and maybe Excel? Anything on the market will do.

        My Mini has an "M2" processor, 32gb of memory, 1TB of internal storage, and two 32" screens. It's capable of doing way more than anything I need it to do. I guess time flies, as I see I got it in Dec 2023. It was my 2012 laptop that I replaced this year. My bad memory (;

    • by PuddleBoy ( 544111 ) on Friday November 07, 2025 @11:33AM (#65780420)

      I've run many Mac servers over the years and the hardware just keeps on chugging away. Web, email, even DNS (at one time) I typically don't retire a Mac until it's 8-10 years old. I have a couple at home right now running SurgeMail. Never give me a lick of trouble.

      • Enough to be statistically significant? My own experience isn't but it wasn't good. For work we had two mac mini build machines, and they both died within three years, granted this was in the intel era where heat management needed to be better, but still sad.
        • I've run minis continuously for years, with no issues of any kind.

          Maybe you got unlucky.

        • --Enough to be statistically significant? --

          No, not really. Maybe 20 servers over the course of time. I have tended to keep them in a cool, dark place. ;)

      • by Gilmoure ( 18428 )

        I thought Macs were just toys running a toy OS and only useful for folks looking to enterain themselves instead of real tough serious business machines that are designed for work and not games?

        • by cusco ( 717999 )

          So long as you don't need any sort of security or need to integrate into any real security backend like Active Directory they're probably adequate. They're in no way an enterprise server, maybe for a small business that doesn't run any enterprise apps. I've never worked anywhere that they would have been acceptable, though.

    • Those ridiculous Mac Mini servers were the worst idea ever.

      In this story, Grab is replacing 200 Mac Mini servers they are renting on the cloud with 200+ Mac Mini servers they have on premsises:

      “We have got four 42RU (600x1200x42RU) racks housing 200+ Mac minis, plus some spare racks to house upcoming scheduled capacity upgrades,” Grab’s post states

      Obviously Grab does not think Mac Mini servers as "the worst idea ever".

      • imagine how many more thousands they could save by replacing 200k in toy computers with a real datacenter platform.

        • 1) You do know the M series chips offer the best performance per watt right now compared to Intel and AMD. 2) Saving costs means little if your hardware accomplishes zero goals. Grab develops iOS apps on these machines; what "real data platform" do you recommend to replace Macs?
  • There are some interesting development environment decisions going on over at "Grab".

  • You mean if you admin your own machines on your own premise you can save a metric shitton of money at scale by not paying a markup on absolutely every part of your operation? You don't say!? Shocking.

    Of all the foolishness I've seen to kick ourselves in the shins, cost way too much, and result in much less control and stability: "Cloud" computing definitely takes the cake.
    • Repatriation makes sense for expensive edge cases, like running Apple hardware (which is a questionable architectural decision). Short of that the average cloud environment is significantly more secure, and much easier to scale in than self hosted infrastructure...until you get a very large footprint. The generation of cloud-phobic infrastructure people from around the early 2010s largely either adopted the cloud or found other work within that decade. Same was true for the prior generation of anti-vir
      • Repatriation makes sense for expensive edge cases, like running Apple hardware (which is a questionable architectural decision).

        And why is it a questionable decision? Currently, the M series chips are the best in performance per watt and can beat Ryzen chips in some benchmarks. For Grab, since they need Macs for iOS development, what alternative do you propose?

        • by tepples ( 727027 )

          For Grab, since they need Macs for iOS development, what alternative do you propose?

          The alternative is developing a progressive web application (PWA) that runs in Safari instead of a native iOS application.

          • The alternative is developing a progressive web application (PWA) that runs in Safari instead of a native iOS application.

            Considering that Grab has building iOS apps for over 12 years instead of developing PWAs, I would guess they know more about their business than you. Have you presented your ideas to Grab?

            • by tepples ( 727027 )

              I was expecting someone who has used the product to help others in this discussion understand why Grab probably chose and continues to choose to develop iOS apps instead of PWAs. The answers might have taken the form:

              A. PWAs weren't capable enough 12 years ago for X, Y, and Z reasons, are now, and the engineering resources to port the native app to a web app would exceed the cost of acquiring and maintaining Macs capable of running the latest macOS
              B. PWAs still aren't capable for X, Y, and Z reasons

              • I was expecting someone who has used the product to help others in this discussion understand why Grab probably chose and continues to choose to develop iOS apps instead of PWAs. The answers might have taken the form:

                Probably because in 12 years of developing iOS applications, developing PWAs does not meet their needs. Bear in mind, Grab is spend lots of money for their infrastructure. If PWAs was the best option, they would have chosen it.

                A. PWAs weren't capable enough 12 years ago for X, Y, and Z reasons, are now, and the engineering resources to port the native app to a web app would exceed the cost of acquiring and maintaining Macs capable of running the latest macOS

                Why are you assuming Grab is still using 12 year old technology? They are developing iOS which means tablets and phones. I seriously doubt anyone is still using 12 year old phones or tablets to use Grab apps.

                B. PWAs still aren't capable for X, Y, and Z reasons

                Maybe you should do some research [hostinger.com]. I would venture the #1 reason PWAs are not

                • by tepples ( 727027 )

                  I would venture the #1 reason PWAs are not used is they require a constant internet connection.

                  The service worker API is explicitly designed to avoid downasaurs in "offline-first" use cases. It acts as a proxy to serve the shell document, style sheet, scripts, and stale data, even without an Internet connection. That's why I asked what obstacles there are other than a downasaur.

                  Again, have you presented your ideas to Grab?

                  I have not presented my ideas to Grab because I am not a user of Grab. I would imagine that most readers of Slashdot are likewise not users of Grab.

                  • The service worker API is explicitly designed to avoid downasaurs in "offline-first" use cases. It acts as a proxy to serve the shell document, style sheet, scripts, and stale data, even without an Internet connection. That's why I asked what obstacles there are other than a downasaur.

                    And you have not considered to enter offline-first, the service worker API has to load? Again, Grab has been doing this for 12 years. PWA is not new and they have chosen native apps.

                    I have not presented my ideas to Grab because I am not a user of Grab. I would imagine that most readers of Slashdot are likewise not users of Grab.

                    But most of us did not assume to know better than Grab unlike you.

                    • by tepples ( 727027 )

                      And you have not considered to enter offline-first, the service worker API has to load?

                      The first time you add a website to your home screen, it installs the website's service worker. You have to use the Internet for that, just as you have to use the Internet to download an application from Apple's App Store.

                      Again, Grab has been doing this for 12 years.

                      And I'm curious about what the blockers for even a partial PWA implementation have been during each of these 12 years.

                      PWA is not new and they have chosen native apps.

                      All I've been asking is what features of Grab combined with missing features of PWA likely led to their continuing to choose native apps.

                      But most of us did not assume to know better than Grab unlike you.

                      I don't see where I "assume[d] to k

                    • The first time you add a website to your home screen, it installs the website's service worker. You have to use the Internet for that, just as you have to use the Internet to download an application from Apple's App Store.

                      So the service worker installs the entire Grab site to you phone? Grab handles food delivery, grocery delivery, package delivery, ride sharing, financial services, etc.. That seem extremely inefficient to load every single function to your phone just because you visited their website. Also I would imagine that working for Grab requires different functions than consumers. But according to you, every time someone visits Grab, it should install all these functions to your browser. I doubt it.

                      And I'm curious about what the blockers for even a partial PWA implementation have been during each of these 12 years.

                      Maybe you should

                    • by tepples ( 727027 )

                      So the service worker installs the entire Grab site to you phone? Grab handles food delivery, grocery delivery, package delivery, ride sharing, financial services, etc.. That seem extremely inefficient to load every single function to your phone just because you visited their website.

                      Each function could be loaded the first time the user uses it. The device has to be online to query what is in stock at any given moment anyway. And I'd be interested in others' speculation about why the client side of the most widely used functions can't all fit in (say) 5 MB, which is twice the size of Doom.

                      You suggested a solution that Grab, Doordash, Uber Eats, Instacart, Favor, Grubhub, Postmates, etc. do not use. I pointed out maybe these companies know way more about their needs and solutions than you. Do you accept that?

                      I accept that, adding a clarification that I suggested the solution for the purpose of asking other people what these companies might know that I don't.

      • These sound like they're being used to build and test software. That's an important part of development, but it's not critical infrastructure that the business absolutely needs to be up 24/7. If they buy the cheapest Mini, it's $600 each and that's $120,000 in hardware costs. They may have some replacements over the life of the hardware, but they'll be able to use those for another five years at the least. Someone needs to wrangle those machines so add another $120,000 for that. Apple's chips are quite effi
      • by Bongo ( 13261 )

        Generally agree, I mean, companies don't need to make their own steel beams, cars, and teacups, Cloud gives the lower parts of the stack over to the specialists, who can industrialise their skill with a massive production line.

        But what's kinda interesting is that there's still industries where lots of small players are needed, like housing construction and maintenance. We don't all live in an IKEA like mass produced kit house. There's huge variety of small custom house designs and arrangements, ad-hoc piece

        • Henry Ford knew something about running production lines. He said "If you need a machine and don't buy it, then you will ultimately find that you have paid for it and don't have it." Relying on rentals, outsourcing, or borrowing, can accumulate expenses without building equity and this is generally a bad business decision. That is especially true in light of the fact that you can rent talent to fix/maintain the machines as needed. This is the critical omission when arguing about the merits of owning your ow
      • by sjames ( 1099 )

        Actually, the cloud remains more expensive and less secure. Remember all that meltdown, spectre, row hammering, etc? All largely irrelevant to people who use their own servers in their own environment.

        You still need an ISP with the cloud. Somehow, you have to be able to launch and monitor, do updates, etc. Smoke signals won't work for that application. You still need IT guys for the office LAN, server admins for your office infrastructure, etc.

        If you decide to go with anything but very vanilla virtual hosti

        • Have you been in a modern tech office lately? With Zero trust principles, there is no office LAN. The office is a glorified wifi access point you can get an MSP to install and support. Rarely do companies that would use AWS/Azure (without a massive footprint as I discussed) have in-office servers etc anymore. Offices with server rooms built out are using them as sparse network racks and gyms. And no, with the cloud you don't get an ISP in the traditional datacenter sense where you sign contracts with t
          • by sjames ( 1099 )

            So those funny things that look like desktop machines are not? and there's no LDAP or domain controllers involved?

            That's funny because the places I'm familiar with have desktop machines, domain controllers, often a NAS or two, and a router with a firewall.

            • Dude, no one who isn't legacy or huge legacy enterprise has a domain controller or LDAP anymore. Domain controller lives in Azure/Microsoft Entra -- workforces are mostly laptops now and you need to VPN in to get to a traditional DC, or it caches passwords locally for some time (not the best security). LDAP is called Okta, Azure Identity services etc; because no SaaS apps callback to your LDAP. I haven't seen a DC in the wild since 2016 or so, and that was to decommission it...at a library.
              • by sjames ( 1099 )

                I'll be sure to tell my very real clients their infrastructure doesn't exist.

        • Also please explain how meltdown/spectre, which are notoriously difficult to actually exploit, didn't affect self hosted setups? It's a vulnerability in process architectures, not cloud infrastructure.
          • by sjames ( 1099 )

            Cloud servers may have more than one user running things on the same CPU. God only knows who the other users actually are. In a corporate environment, everyone running jobs on the server works for the company. It doesn't reduce the risks to zero, but it does reduce them a lot.

      • You're simply ignoring inconvenient use cases. You assume everyone needs a hyperscaler on the back end with BGP peering etc... That's simply not the case. First off, the "stack" is mostly free if you're running say FreeBSD or Ubuntu Server. Upgrading is built into the OS and you can do so for decades. Second, if folks don't care if their site/services drop from time to time, simple one-ISP peering and backups to a cheap USB disk might be all they need. They generally have contracts with consultants to help
    • by sjames ( 1099 )

      But you can "Rent to own, Right on the phone!". Well, minus the "to own" part, that is.

      How is it that so many fall for a deal WORSE than what most people knew was a deal for suckers in the '80s.

    • by Gilmoure ( 18428 )

      Boss, Boss, duh cloud, duh cloud!

  • If cost savings is the goal, they could save a lot more by switching to Windows, or better yet, Linux.

    This seems weird, I've never heard of a company using Macs as servers. Desktops, yes, but servers?

    • by Kitkoan ( 1719118 ) on Friday November 07, 2025 @11:12AM (#65780356)
      They were being used just for IOS app development. You can't build for IOS without using Macs.
      • That makes sense.

      • by Bahbus ( 1180627 )

        You can't build for IOS without using Macs.

        Yes, you can. It is fairly easy to develop for Apple without using Apple products. You must *pay* Apple for the privilege to develop for them (which is a ridiculous concept) and you can't really do any testing without the Apple hardware, but it can be done.

    • Apple used to make dedicated server hardware, as well as a server version of the OS. It's all been around as a long time.

      From the article, it sounds like they were using Mac Mini Servers, a sort of light weight server version of the regular Mac Mini. There wasn't too much service specific about them though, unlike the real hardware they used to make with redundant power supplies, Ethernet cards, drives, etc.

    • If cost savings is the goal, they could save a lot more by switching to Windows, or better yet, Linux.

      Only if saving money was the only goal and nothing else. In this case, not being able to develop iOS apps using Windows or Linux is a significant obstacle over any cost savings.

      This seems weird, I've never heard of a company using Macs as servers. Desktops, yes, but servers?

      Then you haven't been paying attention. Since the Mac Mini launched many, many companies have used them as servers. The main reason is they offer high performance per watt in a small package. The demand is high enough that companies offer custom racks for Mac Minis [racksolutions.com] for higher density in cabinets.

  • by Hentes ( 2461350 ) on Friday November 07, 2025 @11:13AM (#65780364)

    Less than a million a year is nothing for a company at that scale, the move was probably motivated by control/data security reasons.

    • Never underestimate how dumb small departments are when they don't look into the big picture. I was recently denied a business trip (one of two this year) that involved a drive and a hotel for 3 nights for a total cost of $350 because we are "lowering costs". For the record we just announced a $4bn profit for Q3.

      When individuals are given a cost reduction mandate and don't generate many costs to begin with, you can end up with some mindbogglingly stupid and meaningless actions.

  • "Singaporean super-app company Grab"

    So yhey do rideshare and food delivery. How the fuck does that make them a "super-app" company ?

    Ah, the Register. Full of shit they never correct.

    Link to news sites for news, not joke sites.
    • I see that you've never been to that part of the world. Grab is ubiquitous in Indonesia, to the point that if you're on the street you can't go more than a minute without seeing their brand.
  • Wow, I'm in the wrong business. Buy 200 mini-macs for less than $200k and charge $2.4 million for 3 years? This would be about $4 million for a 5-year hardware lifetime. I realize there are additional costs, such as power and network costs, but they said they are "saving" this much money, so doesn't enter the cost. I'm also wondering why developers would be using mini-macs in the clouds and not a desktop? You can have Mac desktop development, but still use Linux machines for things like git, email, sto
    • but they said they are "saving" this much money, so doesn't enter the cost.

      Like many things in the cloud, the flexibility was probably they could add/remove more machines easily with no long term commitment. The problem with cloud has always been at some point, the cost of leasing is greater than building on premises as the number of machines/data/bandwidth reached a certain threshold.

      .

  • I'm guessing they've been using MacMiniColo. Super convenient to set up and run, but kinda pricey. According to the article they'll still be using Mac minis, but buying them and then running them in a Malaysian data center.
  • Just the savings...

    Fire all those "managers" for more savings.

  • by darkain ( 749283 ) on Friday November 07, 2025 @12:18PM (#65780544) Homepage

    There are very VERY few things that are cheaper in the cloud. You're just trading one expense for another, not eliminating it.

    To date, the most cost-effective thing in the cloud is still DNS hosting, because who the hell wants to run thousands of DNS hosts globally distributed, with all the maintenance that comes with it?

    But for compute, or storage, or bandwidth: on-prem will always win in cost.

    • by tepples ( 727027 )

      But for compute, or storage, or bandwidth: on-prem will always win in cost.

      With two exceptions I can think of. Correct me if I'm wrong, but as I understand it:

      1. For lightweight web hosting, a low-end VPS from a company like DigitalOcean is likely to be less expensive than upgrading a home office from home-class home Internet to business-class home Internet to unblock inbound ports 80 and 443.
      2. SMTP is still an old boys' club, with major mailbox providers (such as Gmail and Outlook) blocking connections on port 25 from on-premise IP addresses as likely sources of spam.

    • There are very VERY few things that are cheaper in the cloud. You're just trading one expense for another, not eliminating it.

      It's often not about expense. It's also about externalising the management, the capability. I mean say what you want about the cloud from my end user perspective the options and capabilities for me personally working from home are significantly expanded now that everything is sitting in a cloud somewhere. Especially when that cloud is fully integrated into your groupware.

  • by ctilsie242 ( 4841247 ) on Friday November 07, 2025 @02:49PM (#65780894)

    I wish Apple could bring back the XServe. For a while, Apple was actually the #1 storage vendor with their repurposed Promise arrays and XSan over fiber channel.

    Apple sells AI servers. Maybe they should sell virtualization farms with macOS?

    Then, there is OS X Server. At the time, it was one of the best business framework servers for SMBs out there. Everything from LDAP, to Wiki pages, to web pages, to network booting, to program caching. Apple could definitely make some money from something like that, be it on an XServe. For a SMB, combining this with something like a Time Capsule would make a mint.

    Apple could easily win big if they get back into the enterprise, and hopefully with their AI servers, they can get that framework back. They have money to throw at things, and I'm sure if they have something that can replace AD/Entra, people will flock to them in droves, just to be MS free.

  • Uh, yeah (Score:5, Interesting)

    by Local ID10T ( 790134 ) <ID10T.L.USER@gmail.com> on Friday November 07, 2025 @03:30PM (#65780980) Homepage

    When you start out with a couple boxes and experience sudden growth it is a no-brainer to just rent cloud hosting. Intermittent need, and uncertain volume? -yep, it's cloud time.

    When you grow up a bit and have a stable, predictable volume of usage? Bring it in-house.

    You might go back to the cloud when you get big enough to need reliable global access. And then bring it back in-house again when you are big enough to have your own infrastructure around the world..

    Business needs change over time.

  • by Fallen Kell ( 165468 ) on Friday November 07, 2025 @04:28PM (#65781098)
    Cloud is useful for a few very specific things. Specifically, one off requirements (i.e. I need to temporarily use a few thousand CPUs/GPUs to run this study/calculation and never need to do it again once it is done), and initial ramp up/expansion (i.e. we need to start working now and we can't wait for our datacenter to finish building before we start work, but have a specific timeline for when we use the "cloud" and bring the work back home as the datacenter(s) come online).
  • Let's just say they're amortising the purchase price of the 200 Mac minis over 3 years.
    How much were they spending on cloud-hosted machines? For them to be not just breaking even, but saving $2.4M, they must have previously been spending more than $2.4M over 3 years.
    Split across 200 machines, $2.4M is $12,000 per machine. Over 36 months this is more than $330 per machine per month.
    When the cost of a base model Mac min is $599, this doesn't add up. Even the most expensive M4 Pro Mac mini with a 14-core CPU,

The finest eloquence is that which gets things done.

Working...