+1 to everyone who choose to discuss technical merits and not to choose to engage in os wars in the comments.
Announcement
Collapse
No announcement yet.
DragonFlyBSD 5.2, TrueOS 18.03, FreeBSD 11.1, Ubuntu 16.04/18.04 & Clear Linux Tests
Collapse
X
-
Originally posted by F.Ultra View PostWell it's your guess as well, funnily however you are the only one that combines your guesses with insults. Anyway the main gist of my original comment was that they didn't choose BSD due to it's network stack which is not a guess but a fact since networking on those types of hardware is done outside the control of the main CPU.
What does Unix-like have to do with scaling to many CPU:s and "nanosecond-scale-delay processor bus"? A commodity x86 server in 2018 have far more cpu cores than the biggest system you ran traditional Unix on back in the day.
There are 6 other contemporary Unices that conform to The Open Group's UNIX 03 standard. Your subjective perspective that last UNIX died in the nineties is somewhat inaccurate.
Because there are some tasks where delays inherent to network-interconnected nodes (clusters are networked) are not acceptable. It's too big of a bottleneck. Mostly heavy-duty database applications.
Mainframe hardware especially and to an extent everything else with a big socket (8S,16S+) count is next to unobtainable for Linux kernel developers. It's just too flippin' expensive or covered by NDA's. So, if you cannot develop for it, it usually also won't run on it. Or at best, runs on reduced performance. Vendors of some of this hardware at the same time produce their own commercial Unix that is optimized to run on this hardware. Sounds logical?
Originally posted by Pawlerson View Post...
Overall, "bla-bla" from a known Linux-nazi in a BSD thread. Again. Get a life retard. You are worth neither further attention nor reply. Want to insult, be insulted.
Originally posted by edenist View PostIt's interesting that the Poisson Pressure Solver has FreeBSD 11.1 at the bottom and TrueOS at the top [Effectively 12-CURRENT]. Either there are some big performance improvements in AVX in the 12.x kernel, or the TrueOS guys have figured out some mean compiler options to achieve such huge gains. As someone who runs some AVX-heavy workloads on 11.1, I'm going to seriously look into what is giving such a big boost.
Thanks for continuing with the BSD tests, Michael!
Comment
-
Originally posted by aht0 View Post
Enterprise doesn't give a fuck about performance alone. In fact they rather give up performance in exchange for stability. Even computer noobs like you should know it.
Overall, "bla-bla" from a known Linux-nazi in a BSD thread. Again. Get a life retard. You are worth neither further attention nor reply. Want to insult, be insulted.Last edited by Guest; 23 April 2018, 11:00 AM.
- Likes 2
Comment
-
Originally posted by aht0 View PostMainframe hardware especially and to an extent everything else with a big socket (8S,16S+) count is next to unobtainable for Linux kernel developers. It's just too flippin' expensive or covered by NDA's. So, if you cannot develop for it, it usually also won't run on it. Or at best, runs on reduced performance. Vendors of some of this hardware at the same time produce their own commercial Unix that is optimized to run on this hardware. Sounds logical?Last edited by Guest; 23 April 2018, 11:34 AM.
- Likes 2
Comment
-
Originally posted by aht0 View Post"bla-bla" from a known Linux-nazi in a BSD thread. Again. Get a life retard. You are worth neither further attention nor reply. Want to insult, be insulted.
Comment
-
Originally posted by aht0 View Post
Uhm.. Where did I insulted you?
There are 6 other contemporary Unices that conform to The Open Group's UNIX 03 standard. Your subjective perspective that last UNIX died in the nineties is somewhat inaccurate.
Because there are some tasks where delays inherent to network-interconnected nodes (clusters are networked) are not acceptable. It's too big of a bottleneck. Mostly heavy-duty database applications.
Mainframe hardware especially and to an extent everything else with a big socket (8S,16S+) count is next to unobtainable for Linux kernel developers. It's just too flippin' expensive or covered by NDA's. So, if you cannot develop for it, it usually also won't run on it. Or at best, runs on reduced performance. Vendors of some of this hardware at the same time produce their own commercial Unix that is optimized to run on this hardware. Sounds logical?
Enterprise doesn't give a fuck about performance alone. In fact they rather give up performance in exchange for stability. Even computer noobs like you should know it.
Overall, "bla-bla" from a known Linux-nazi in a BSD thread. Again. Get a life retard. You are worth neither further attention nor reply. Want to insult, be insulted.
At a guess.. some API changes between v.11 and 12-CURRENT and/or changes in compiler (Clang version tends to be different)
Big socket hardware runs mainly AIX and Solaris and the reason IMHO is that #1 the market is very small (because most people have switched over to cheap commodity servers in a HPC cluster so the only enterprises left with the big socket hardware are the enterprises that work with data sets that does not scale well to HPC) when compared to the server market as a whole and #2 IBM and Oracle who are the main players in this market serves mostly legacy customers and these legacy customers does not want to have to switch to something else due to all the recoding and recertification that would require so they keep on selling AIX and Solaris as the main solution for these systems, not to mention the vast amount of fanboism inside these enterprises (which I well know since I have worked in both AIX and Solaris heavy places in my 20+ years in the computer industry).
- Likes 1
Comment
-
Originally posted by F.Ultra View Post
I have not written that the last Unix died in the nineties. Your claim was that "Unix-like" was a requirement to scale to many cores which I refuted by pointing out that you today can buy cheap commodity x86 hardware with more cores (you can e.g get 144 physical cores and 288 with HT in a x86 Intel server today [don't know if AMD have any systems with more cores]) than you could run with any Unix just a few years ago. This is not to say that AIX cannot run with massive amounts of cores today, but "Unix-like" is a philosophy since many decades (that was the point).
Big socket hardware runs mainly AIX and Solaris and the reason IMHO is that #1 the market is very small (because most people have switched over to cheap commodity servers in a HPC cluster so the only enterprises left with the big socket hardware are the enterprises that work with data sets that does not scale well to HPC) when compared to the server market as a whole and #2 IBM and Oracle who are the main players in this market serves mostly legacy customers and these legacy customers does not want to have to switch to something else due to all the recoding and recertification that would require so they keep on selling AIX and Solaris as the main solution for these systems, not to mention the vast amount of fanboism inside these enterprises (which I well know since I have worked in both AIX and Solaris heavy places in my 20+ years in the computer industry).
which means single physical mainframe with shitload of processors and gargantuan computing ability on it's nanosecond-scale-delay processor bus - which still takes real old-fashioned Unix.
You are implying that architecture in question is meaningless. By following your logic: ARM octacore SoC should be equally performant compared to x86 octacores running on similar frequency . Do you realize now what a bullshit your claim is?
Comment
-
Originally posted by aht0 View Post
Do not twist or change my words. It's the problem with bigots. Whatever you say, it's going to be twisted any which way fitting their point.
You are implying that architecture in question is meaningless. By following your logic: ARM octacore SoC should be equally performant compared to x86 octacores running on similar frequency . Do you realize now what a bullshit your claim is?
Architecture in the way that you are reasoning is not only meaningless, it's also work against you since (once again) the Unix of yesterday where the Unix "architecture" was designed scales horrendously as compare with any modern OS (which includes modern versions of Unix). The fact is that there is nothing in the Unix architecture that promotes scaling of any sort. Do you even know what this Unix architecture is that you speak of?
So now I'm both a "Phoronix Linux-Supremacist/Apologist", a "greedy son of a bitch" and a "Bigot" and you have the nerve to write "Uhm.. Where did I insulted you?". You don't really put the BSD community in a good light (if you even are a BSD user in reality) almost makes me feel ashamed as a FreeBSD developer (I maintain some ports packages on FreeBSD).
- Likes 1
Comment
Comment