I'm starting to say that everyone should use the tool he's more comfortable with. So if you're not comfortable with Linux, that's fine. But bashing Linux by saying something not true is not a right approach.
Solaris is dead. Please, stop talking about zombies. It sad to see it gone, really. But that's how it is.
Oracle fired the majority of SPARC and Solaris teams and they are not working on any release anymore, with the silly excuse of releasing it with a Continuous Delivery model, that actually didn't produce any real improvement but bug fixes.
Oracle is just keeping it commercially alive enough to milk their customers until 2034.
There is no reason to run Solaris anymore, unless you just want it. Even from a technical point of view, since Linux reached and now is even superior in performance.
And I'm not the one who assert that. Brendan Gregg said it, and he's one of the most prominent DTrace and Performance engineers in the world, who worked at Sun and later at Joyent, before start working with Linux at Netflix and now Intel.
Even Oracle now is pushing their Linux distro more than they are pushing Solaris. They even ported DTrace to it.
Yes, you're right here. FreeBSD is just nowadays as good as Solaris. And it is another reason for not deploying Solaris anymore.
What is really missing is a more comfortable jail management tool like zoneadm/zonecfg. But I think it will be there when PkgBase will be completed. After all, Solaris zone management became really good only after the introduction of IPS.
I think it's also missing something like FMA, but maybe here I'm wrong and developers can correct my statement.
OpenIndiana? Seriously? Come on...
There is still someone who believes in the dream of illumos?
OpenIndiana was never near to what Project Indiana really was. It's just OpenSolaris 2010.05 with MATE instead of GNOME 2.x, with most consolidations ported from the open source parts of Oracle Solaris, but with an illumos kernel under the hood since b151.
Alasdair Lumsden, the founder of OpenIndiana, abandoned the project, and now there are few guys who are running it in an eternal beta stage called Hipster. SFE is dead and the main repository lacked of a lot of useful desktop applications.
And illumos is just what the ONNV_b146 (the last public build of OpenSolaris) was. There is no real innovation, no real improvements neither from performance point of view nor from technical point of view.
Even Solaris 11 went beyond that, and ZFS developers had to created OpenZFS to save it from the vegetative state of illumos.
There is no real commercial backing behind it. Delphix migrated to Linux. Joyent is no more. OmniTI stopped being involved with illumos almost 10 years ago. Nexenta stopped being publicly involved with illumos because it sells its own closed source solution.
There is only Oxide nowadays.
Oracle introduced ZFS encryption in Solaris 11 Express, in late 2010. illumos couldn't even develop something like that, even when at that time commercial entities were "founding" the project. ZFS encryption was developed in ZFSonLinux, that became the reference implementation of OpenZFS.
Linux is way better than this, to be honest.
About NetBSD, I really like it. It was my favourite BSD for years (now I prefer OpenBSD), but I'm failing to see how it can be better than Linux.
Well, it's technically better designed from a portability point of view (I seriously love build.sh), since this was not Linux initial goal, but nowadays there are more embedded Linux out there than embedded NetBSD installations.
That said, I think NetBSD released some really interesting projects, like RUMP. And I'm a big fan of unikernels.
Yes, that's true.
On the desktop side Windows can do even more than Linux, since desktop Linux is a mess. And it will always be like that for multiple reasons.
On the server side, though, this is a joke that is not funny.
Is there seriously anyone in 2024 who is still debating about KVM not being a "pure" type 1 hypervisor, even when there are WAY MORE mission critical KVM deployments than Hyper-V out there?
What are you losing with KVM from it being a type1-2 hybrid? Features? No. Security? No. Perfomance? Hell, no.
Then again, if you want an enterprise-grade hypervisor for your workloads and you're not a hobbyist, your best bet would be still vSphere, even after VMware being bought by Broadcom. It's just the best hypervisor out there, and everything else in the x86 space can only running behind it for the foreseeable future.
ReFS is just an another half-assed alternative of ZFS. And no one who is sane enough to choose a storage solution would choose it as such.
It's just a traditional file system with CoW and snapshots capabilities. It's even inferior than BTRFS, which pales as well in comparison of ZFS.
Also, integrity streams feature is not about self-healing. It is just on the fly scrubbing. And it is a mostly useless feature on a server/workstation, since scrubbing greatly impacts I/O performance, and it is something usually done at specific intervals at storage-level.
There is no something like self-healing in file system, unless you're deploying it on top of a RAID with its parity and by using ECC memories to avoid errors in the I/O pipeline.
So ReFS is not doing any black magic that both ZFS or BTRFS can't do. In fact, it is doing less, since ZFS and BTRFS have RAID capabilities built-in and therefore scrubbing can actually fix corruption, while ReFS without Storage Spaces would just calculate the checksum and block the access to the file, just like BTRFS would do. Heck, even APFS in modern Apple devices can do that much.
Storage Spaces also is a way inferior solution than LVM+RAID. Do you need to partition your Volume Group, to make use of LUNs aggregation without losing flexibility? Well, bad news for you, pal. You can't. Just format it with NTFS/ReFS and use it as a single volume. Useful, really useful, if all you want to do is playing. Then again, it's also true that if you would run something serious, you wouldn't deploy Windows Server in the first place.
Do you know why Storage Spaces were developed? Because in Windows there was no capability of LUNs aggregation.
Every LUNs had to be considered as a different volume. Did you need more space? You had to enlarge the LUN(s) and prey God that the multipath service wouldn't complain about the change of the volume(s) it was managing.
In Linux/AIX/HP-UX? You could just create another LUN in the same masking group and you were good to go. You just had to enlarge your VG and forget about anything else.
VSS was never a good snapshot solution when not backed by a real storage, which provides snapshots with volume replication capabilities.
And when you're backing it with a storage, VSS is just a graphical frontend with no real advantages, since all the dirty work is done by the latter under the hood.
The usefulness of VSS is more related to how Windows can help you managing your files client-side. And that's why I agree with you that Windows is a better desktop OS than Linux.
This is not something you should do server/workstation-side, since dedup is I/O intensive and it can (and it does) compromise the performance. Especially since the majority of instances of Windows are virtualized.
And even when they would be physical they wouldn't have enough I/O bandwith capabilities to perform deduplication without impacts.
Even BTRFS can do dedup. And since both of ReFS and BTRFS are NOT SAN-based storage solutions that feature is useless.
Deduplication shall ALWAYS be done at storage-level.
Telling that Windows is more customizable than Linux is outragerous at best. Even for a server configuration.
Stop with the FUD, please.
Again, ReFS and Storage Spaces are useful only when scaling your local storage. They are not a NAS/SAN alternatives.
You shouldn't care about parity at server/workstation-level, because that kind of performance tuning should be done at storage-level. Your NAS/SAN should care about parity, not Windows or Linux servers.
That said, what you're calling Mirror Allocation Parity is just RAID50 (but with the name not decided by the marketing department), something that you could do in Linux even 25 years ago.
And Storage Tiers (another marketable name) are nothing special. You can mix SSDs and HDDs in RAID in Linux. You could even make asymmetric arrays if you want.
But, again, this is not something you should care at server/workstaton-level. Let the storage do this kind of stuff.
Linux is not a disaster at all. You're not harsh, you're just misinformed.
And there are no "better" alternatives. There are alternatives. What makes them better is the kind of workload you're going to deploy on your server.
Don't get me wrong, Linux has its own share of problems and quirks, but it is still miles ahead of Windows Server, which was always a sub-optimal solution, forcefully deployed in data centers mostly due to Active Directory, Sharepoint and custom ASP.NET websites.
Windows Server always suffered of scalability and performance issues. It was always beaten by commercial UNIXes first and Open Source UNIXes (*BSD and Linux) later. Even Microsoft replaced it in most of its mission-critical parts of Azure, with its own Linux-based OS.
And look, I'm not a Linux guy. My daily driver is a MacBook Air M2, my Lenovo x270 runs OpenBSD-current, my Pinebook Pro runs NetBSD 10-STABLE and my NAS runs FreeBSD 14.1. I also bought a server yesterday, which will run FreeBSD.
Also, I'm a Solaris guy. I always considered Solaris as the best thing since sliced bread. But still your statements are just plain false.
