Hacker News Reader

Comment by cyberpunk | Hacker News Reader

Im not going to name names, but i know of a system still using sunos 4 in prod...

I really loved solaris. Its such a shame how sun worked out. Oh well. FreeBSD has some of the features we had in 2005 on solaris 10...

View on HN

Sun's attempt to build a massively multi-thread system failed at that time, it's interesting to see the direction we have moved, almost 20 years later again (I say this whilst building a "new" 2nd hand epyc system with 32 cores...)

Would there have ever been a pivot to the consumer market, or would it be stuck behind big iron like with IBM and whatever they have hiding under the hood?

Sun did this because they had a lot of issue building faster cores. The were very late on out of order processing and generally their CPU were never all that good.

And massive multi-thread system that doesn't even have FPU and is a slow in-order core isn't actually useful. Sun CPU deserved to fail.

> FreeBSD has some of the features we had in 2005 on solaris 10...

Yeah, it really does feel like FreeBSD is the best successor:

* ZFS built in

* boot environments

* dtrace

* zone^wjails

(I feel like there was more that I'm forgetting?)

Although I do feel like it's less polished and the desktop support is comparatively lackluster. Oh well.

+ A clean, stable base system

Which meant that on this Unix system you had to compile a completely new Unix system (with the GNU tools and libraries) to be able to use any software that wasn’t hopelessly outdated.

Nothing will be as good as Solaris because the codebase was proven with years upon years of enterprise use at scale with a support organization behind it. Big-ass paying customers will find more/less all of the edge cases and want their systems fixed with a patchset now.

One example of how things from Solaris aren't as good: tldr: ZFS. Solaris' was good, all others are fragile and unproven.

FreeBSD ported ZFS first from OpenIndiana. They threw that away and adopted ZoL (OpenZFS). ZoL morphed from the OpenIndiana code into an largely untested, hacky manner. Each OS should've thrown it away a long time ago and started with a clean room implementation for each OS, sharing only comprehensive disk specification, user documentation, unit and integration tests, and fuzzing. The OpenZFS initiative tries to serve too many masters, has too many cooks, and no one's validating code adequately.

On Linux, for anything real, I'll take XFS on mdadm. At least I won't lose an array because of shitty ZoL code and lack of support for said shitty code.

https://utcc.utoronto.ca/~cks/space/blog/solaris/Solaris10X8...

Here's someone comparing Solaris 10 to Linux:

> the sysadmin experience on a modern Linux is much better than on Solaris 10. There really is no comparison between, say, apt-get and the official Solaris patch management tools.

> there is much more useful software conveniently available for your average Linux distribution (either as prepackaged binaries or as easy to build source code).

> Solaris's driver support seems lacking. Some things (eg, jumbo frame capable Ethernet drivers) are backwards compared to Linux; others that I would like are just missing, such as drivers for 3Ware Escalade controllers.

> (Note: 'supported in Nevada' doesn't count, except as an indication that it might show up in Solaris 10 at some indeterminate time.)

> I don't know how well Solaris 10 x86 runs on non-Sun x86 hardware, especially recent hardware.

> (Of course, it runs fine on Sun x86 hardware and Sun has competitively priced x86 stuff these days, so this may not matter to you.)

The blog has a whole section for solaris, but the past few years it's just ZFS stuff because Solaris is no longer interesting to them compared to modern Linux.

Yeah well Solaris died. Of course it can’t compete with Linux anymore.

I’m not buying the network complaints though. I did some really weird networking stuff with Solaris boxes that I still can’t easily do again today. I mean, can you control bandwidth of individual processes from right within systemd? With a single command? We had that in 2011… can you apply vlan tags to a single docker container based on which user started it? Jumbo frames when you talk to the emc array for iscsi but Normal for your db traffic?

Nope. Well, maybe. But nowhere near as easy [0]

0: https://en.wikipedia.org/wiki/Solaris_network_virtualization...

Thouroughly enjoyed working with Sun and Solaris back in the day, but I don't follow your logic.

> years upon years of enterprise use at scale with a support organization behind it

Isn't this just as true of enterprise Linux, for coming on twenty years now?

Linux evolves really fast and in an unplanned chaotic way. So it partly applies to linux for sure but there are also forces in other directions.

Also, linux isn't an OS, its a kernel. The OS is built out of bits and pieces of open source software flying around.

With Solaris you get a kernel and a user-land portion of the OS in one package designed together and for each other.

I’ve heard this argument for years but come on how many Linux systems do you have in prod where awk isn’t gawk, sed isn’t gsed, libc isn’t glibc (okay, alpine with musl has a bit of market for containers). I think the Linux userland is pretty much identical across the distros and tbh I only deploy Ubuntu server when we need Linux vms anymore.

Yeah but for a long time there were many different init systems. There are different packaging systems. There are a huge amount of difference above the very basic tools.

And I didn't say Linux was bad or anything, its just that there are a lot of hands in the pudding, both in the kernel and in the users space.

> how many Linux systems do you have in prod where awk isn’t gawk

All default Ubuntu installs? Maybe also Debian, I’m not sure.

What features are missing from freebsd?

Kern features were integrated in unexpected places. Like, for instance, I could run some command on my Solaris server and create a resource pool of processors and memory; and it was so integrated that this contract could be applied to individual network connections. I could allocate half a processor or some memory to, say, an oracle process for licensing reasons right within the smf, one command. Sure, I can do these things on modern Linux and bsd, and I make my living doing complex things with k8s these days, but it all seems like such a hacked together mess compared to what we had.

Reply on news.ycombinator.com