We (the Linux kernel community) put a lot of work into scalability back then. One time we delayed a Red Hat release several weeks due to a persistent bit flip on an NFS filesystem on one of our stress tests that we thought was a kernel bug. It ultimately turned out to be a broken Cisco switch that was happily fixing up the CRC and checksums on the NFS packets it was bit flipping.
Of course. Sun was ahead of the curve in OS scalability and SMP hardware departments and for some time Solaris was still the best OS to run databases and Java.
The Sparc2k was a beast. I got on in 1998 that (among other tasks) ran the IRC bot for our channel with a 4gb text quotes database to chain off of. ("Shut up, lucy"). It usually ran at a load average of 4 to 12 per CPU. I had just over 1TB of disk array on it; in 4 dozen, full height, 5 inch 23GB SCSI drives.
I sank about $10k into that system in ebay parts. at the time there just wasn't any way to get such bandwidth out of x86, but used gear was cheap. if you could afford the time, sweat and blood to keep it running.
I used to work at a shop that had hundreds of X4100's running x86 Solaris 10. It was a pleasure to work those hosts. Solaris had some really cool stuff like ZFS, dtrace, Zones/Containers, SMF, etc.
And now you read the release notes with dread as you witness more and more functionality being dropped because they choose not to maintain it. Luckily there is FreeBSD
Yeah it’s a shame. If Sun would have made Solaris (both SPARC/x86) free sooner, they might have prevented the mass exodus of companies moving off to save money.
In retrospect, they should have gone down the Sun386I path back in the day, instead of changing gears and spending $$$ building up the SPARC architecture and ecosystem.
