Should Apple bring back the Xserve with Apple silicon?

[LeoI07]

In the 2000s, Apple sold the Xserve, a rack-mounted server system. In 2011, it was discontinued and replaced with server versions of the Mac mini and Mac Pro.

Over a decade after the Xserve was discontinued, Apple is about to transition the Mac Pro to Apple silicon, with M2 Ultra or M2 Extreme chips. I'm wondering if Apple should revive the Xserve with said M-series Extreme chips. I'm no server expert, so please correct me if I say anything wrong.

---

Hardware​

I think that the M2 Extreme chip that will end up in the next Mac Pro has remarkable potential as a server CPU due to its sheer performance.

According to my calculations, if you multiply the Geekbench 5 multi-core score of the M1 Ultra by the supposed CPU core increase going from M1 Ultra to M2 Ultra, then multiply that by the multi-core score increase going from M1 to M2 in the 13-inch MacBook Pro, you get a score of around 33,110 for the M2 Ultra. If we then multiply this by the performance increase going from M1 Max to M1 Ultra, you get a multi-core score of around 62,643 for the M2 Extreme. That's over 3 times more than the most powerful 2019 Mac Pro, almost 3 times more than Intel's Xeon W-3175X, over 3 times more than AMD's EPYC 7742, and over 15 times more than the last (2009) revision of the Xserve! I don't know about you, but I'm pretty sure servers need top-of-the-line hardware, and that's most certainly it. Note that the multiplication of existing scores probably isn't the best way to calculate the performance of unreleased CPUs.

Another advantage for Apple silicon is power efficiency. The RISC-based ARM architecture of Apple silicon creates much more efficient CPUs than the CISC-based x86 architecture of Intel and AMD CPUs. Data centers, which rack-mounted servers are found in, are notorious for using lots of electricity. Use of ARM-based CPUs that draw less power would not only save plenty of money on the electric bill, but would also help reduce carbon emissions, as less electricity use means less fossil fuels need to be burned to generate it.

Edit: A hardware disadvantage of an Apple silicon Xserve would be the potential lack of RAM expandability, I/O, and PCIe lanes. Credit to casperes1996 for pointing this out.

Software​

The operating system that the Xserve would run may be a downside. Most servers run either open-source operating systems or Windows Server, and macOS Server has been discontinued due to the ability of command-line software to substitute for it. Since modern servers commonly run UNIX-like operating systems, which macOS is one of, macOS could concievably substitute for them, but I doubt that all the existing server software for operating systems like Linux would "just work" on macOS, and x86-only server software would have to be run under Rosetta, which will take a performance penalty.

---

So, would an Xserve with Apple silicon be a good idea? I'd like to hear your thoughts.

 

[belvdr]

Not at all.

 

[casperes1996]

Servers have a wide range of use-cases and needs.

Some require tremendous amounts of memory, which is currently a point of uncertainty for Mac Pro-tier M chips, but I personally doubt it will go beyond 256GB, potentially 512.
Some servers need large amounts of I/O and PCIe lanes. Also not necessarily a strong point for M based computers relative to the competition, from both a hardware and software compatibility perspective.
Some servers may indeed by well served by an Apple Silicon Xserve. But I wager its a small niche, and from a software perspective a lot of them would probably rather get a Neoverse based machine with Linux anyway.

I think the potential market is too small for a dedicated Xserve product. However, the 2019 Mac Pro came with a rack mount option. Selling an Apple Silicon Mac Pro in a 1u rack mount option as well could be a good way of allowing setups like that for any potential customer base that may exist, no matter how small

 

[hobowankenobi]

All comes down to market. As much as I would like to to see it, I doubt there is a maket, at least by the tradional standards of what servers were. Video Render farm seems like the most likely option, if anything comes.

 

[jdb8167]

I think the potential market is too small for a dedicated Xserve product. However, the 2019 Mac Pro came with a rack mount option. Selling an Apple Silicon Mac Pro in a 1u rack mount option as well could be a good way of allowing setups like that for any potential customer base that may exist, no matter how small

Including Apple itself with their Xcode Cloud push. I've been curious how they are planning on handling that going forward. Obviously they can just use Linux servers if necessary but I'd imagine Apple could see some value in leveraging macOS directly.

 

[casperes1996]

Including Apple itself with their Xcode Cloud push. I've been curious how they are planning on handling that going forward. Obviously they can just use Linux servers if necessary but I'd imagine Apple could see some value in leveraging macOS directly.

Maybe. But Apple can afford specialised hardware and software solutions for setups like these. iCloud is partially on AWS and other cloud providers already so it's also not necessarily something they see as their ballgame either way.
And if they want macOS, they can make macOS run on a Neoverse server easily enough. But frankly I think they just use Linux servers. With CoreFoundation for Linux even things that are programmed against a Mac target may easily enough translate to Linux, which usually isn't the hardest process in the first place

 

[jdb8167]

Maybe. But Apple can afford specialised hardware and software solutions for setups like these. iCloud is partially on AWS and other cloud providers already so it's also not necessarily something they see as their ballgame either way.
And if they want macOS, they can make macOS run on a Neoverse server easily enough. But frankly I think they just use Linux servers. With CoreFoundation for Linux even things that are programmed against a Mac target may easily enough translate to Linux, which usually isn't the hardest process in the first place

I'm not sure what they are doing but they do support Homebrew on their servers. I'd be very curious to find out how well that works. But not curious enough to sign up for the beta.

 

[casperes1996]

I'm not sure what they are doing but they do support Homebrew on their servers. I'd be very curious to find out how well that works. But not curious enough to sign up for the beta.

Huh; That's interesting. I didn't actually know that. It's especially cool to me given Apple's relationship with Homebrew. I know a lot of people at Apple use it internally too (along with MacPorts), so it's almost the official macOS package manager in an unofficial capacity, haha. - But home-brew does also exist for Linux as a Linux package manager, it's just less common and of less use there, so that might be what's happening? To make it familiar/portable with compile/CI actions between macOS and Linux, I dunno. For all I know it could also just be a bunch of rack mount Mac Pros or arbitrary AWS instances running a macOS based VM instance

 

[jdb8167]

Huh; That's interesting. I didn't actually know that. It's especially cool to me given Apple's relationship with Homebrew. I know a lot of people at Apple use it internally too (along with MacPorts), so it's almost the official macOS package manager in an unofficial capacity, haha. - But home-brew does also exist for Linux as a Linux package manager, it's just less common and of less use there, so that might be what's happening? To make it familiar/portable with compile/CI actions between macOS and Linux, I dunno. For all I know it could also just be a bunch of rack mount Mac Pros or arbitrary AWS instances running a macOS based VM instance

I always thought Homebrew’s Linux support was half-baked. Maybe it has gotten better since I last looked.

 

[casperes1996]

I always thought Homebrew’s Linux support was half-baked. Maybe it has gotten better since I last looked.

It's clearly not the main platform, and it's also mostly unnecessary and redundant on Linux where the preferred package manager is your distro's default anyway, so yeah

 

[MacsRgr8]

To me? Nah...

Back in the day I was an IT Admin managing a few hundred Macs.
Started off back in the Mac OS 8 days (couldn't call that actually "managing Macs" really.. but had some serious fun with Mac OS X Server 1.2 "Rhapsody") and went fully blown from Mac OS X Panther onwards, and yes using Apple Xserve (so Mac OS X Server 10.3) with Xserve RAIDs for the Open Directory stuff, syncing of home-folders, NetBoot / NetInstall, et all.

With every new Mac OS X / Server and new Xserve CPU (G4 -> G5 -> Intel) we were excited and updated regularly.

• Then Apple discontinued Xserve RAID.... okay... well, then use the Promise VTrak RAIDs instead
• Then Apple discontinued Xserve... say what..? Damn.. quick get some Xserves now we still can!
• Then Apple introduced OS Lion and Server.app.... duh... hmm.. Apple is starting to dumb down their own Server system now
• Then, only half a year ago.... Apple discontinued Server.app completely

With the introduction of Profile Manager, which was to be installed on a desktop Mac..., I never really understood Apple's intention of Mac / iOS management.
Do they seriously want us to use that? Get a Mac mini fire up Server.app, create an OD Master and enable Profile Manager? It was a single point of failure, which inevitably will go wrong (seen some serious OD databases gone wrong).

It was obvious that SaaS was the way to go for managing Apple devices, and Apple did not want to do this themselves. Leave it to those who want to specialise in that, and want to give 24/7 support on it
Enter Jamf Software, Cisco Meraki and other 3rd party MDMs, but also AD integration, etc.

A few years on, VPP and DEP was introduced to help in Mobile Device Management. So... Apple offering those services "in addition to" 3rd party MDMs.
Later VPP and DEP combined with Managed Apple ID became Apple School Manager and Apple Business Manager.... still "in addition to" 3rd party MDM solutions.

And now Apple Business Essentials... and I wonder when that will get fully a blown MDM, or enable granular user management through Managed Apple ID, so.... no need for 3rd party MDM anymore?

So, after that little story, this is the reason why I say no to Apple Xserve with Apple Silicon:
No "on premise" services or management of Apple devices by Apple. SaaS only.

 

[hobowankenobi]

To me? Nah...

Back in the day I was an IT Admin managing a few hundred Macs.
Started off back in the Mac OS 8 days (couldn't call that actually "managing Macs" really.. but had some serious fun with Mac OS X Server 1.2 "Rhapsody") and went fully blown from Mac OS X Panther onwards, and yes using Apple Xserve (so Mac OS X Server 10.3) with Xserve RAIDs for the Open Directory stuff, syncing of home-folders, NetBoot / NetInstall, et all.

With every new Mac OS X / Server and new Xserve CPU (G4 -> G5 -> Intel) we were excited and updated regularly.

• Then Apple discontinued Xserve RAID.... okay... well, then use the Promise VTrak RAIDs instead
• Then Apple discontinued Xserve... say what..? Damn.. quick get some Xserves now we still can!
• Then Apple introduced OS Lion and Server.app.... duh... hmm.. Apple is starting to dumb down their own Server system now
• Then, only half a year ago.... Apple discontinued Server.app completely

With the introduction of Profile Manager, which was to be installed on a desktop Mac..., I never really understood Apple's intention of Mac / iOS management.
Do they seriously want us to use that? Get a Mac mini fire up Server.app, create an OD Master and enable Profile Manager? It was a single point of failure, which inevitably will go wrong (seen some serious OD databases gone wrong).

It was obvious that SaaS was the way to go for managing Apple devices, and Apple did not want to do this themselves. Leave it to those who want to specialise in that, and want to give 24/7 support on it
Enter Jamf Software, Cisco Meraki and other 3rd party MDMs, but also AD integration, etc.

A few years on, VPP and DEP was introduced to help in Mobile Device Management. So... Apple offering those services "in addition to" 3rd party MDMs.
Later VPP and DEP combined with Managed Apple ID became Apple School Manager and Apple Business Manager.... still "in addition to" 3rd party MDM solutions.

And now Apple Business Essentials... and I wonder when that will get fully a blown MDM, or enable granular user management through Managed Apple ID, so.... no need for 3rd party MDM anymore?

So, after that little story, this is the reason why I say no to Apple Xserve with Apple Silicon:
No "on premise" services or management of Apple devices by Apple. SaaS only.

Right there with ya bro. Appleshare IP forever! 🙂

I would only add a small sidebar: before Casper server (now Jamf), some of us used Radmind, Bombich's NetRestore, not to mention NetBoot and NetInstall, and for sure Deploy Studio.

Never fear! Jamf is not the only game in town....we still have open source stuff like Munki and MDS.

Back on topic: I love the notion of Apple leverainge their own silicon for some fabulous, feature rich, more secure, more private, more cost effecitve cloud services. But I don't see it happening. If they could deliver a better SaaS product, with better value, they would. But that seems unlikely to me. They would have to have a such a compelling density/heat/power advantage to justify the work of developing and supporting enterprise servers that the bottom line would get folks to consider jumping ship (from the likes of AWS).

My take is that Apple jumped into the server world to try and make a run at IT and the enterprise world...because there was a big Apple-shaped hole to be filled. But after less than a decade, there were quite a few new storage and server options available (once Macs played nice with SMB back about 10.6 or so) IT shops could run to supprt Macs and/or iOS. And SMBs could get along just fine with NAS boxes for traditional services...so the hole Apple filled with server gear basically ceased to exist.

Throw the cloudification of nearly everything else on top (...remember running email servers?! That was a hoot! And the spam filters...oh the joy of fighting spam 24/7!) and it is very understandable how Server ceased to be a thing.

 

[MacsRgr8]

Ooh.. yes!
AppleShare IP... Retrospect as backup software...

Strolling down memory lane! 😁

 

[hobowankenobi]

Ooh.. yes!
AppleShare IP... Retrospect as backup software...

Strolling down memory lane! 😁

Oh yes.

I occassionally still wake up in cold sweat having a nightmare about having to feed about 16 DAT tapes into the drive, over and over...to recover some senior VP's data. What a disaster that was. Hours and hours to restore the data set to grab a few files. That part I don't miss.

 

[A.Goldberg]

Ooh.. yes!
AppleShare IP... Retrospect as backup software...

Strolling down memory lane! 😁

Wow. I’m in my 30’s, but it takes me back to elementary school when we had to login to the server to get our documents. This was in the days of Performa AIO’s and later the original iMac and iBook G3’s.

I still remember the instructions… “Go under the Apple, click on the chooser, click on AppleShare…”

I guess decades later I can now share that my assigned password in 3rd grade was “Oliver”… the computer teacher said the theme of our passwords were “dog names”. I guess she failed to realize there was a kid in the class name Oliver.

Strolling down memory lane indeed.