-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
Linux or Solaris -- which to choose?
- Thread starter rif
- Start date
cannedcreamcorn
Member
I'll let others answer for Linux(which I run right now), but I have dealt with Solaris in big enterprise environments and have some distinct opinions after using both. With Solaris there are advantages and disadvantages:
Advantages:
1. Monoculture - It's all coming from 1 place (although more and more FOSS is creeping into Solaris).
2. Stability - This has as much to do with Sun hardware than the OS. On x86 it is probably similar to a good Linux distro
3. Scalability - Solaris is geared more toward Big Iron and has had several years head start on Linux scalability. Once again, Sun boxes take better advantage of this.
4. Development - If you develop large, complex apps, Solaris has many built-in features (kernel-level debuggers, trace utils) that are still more advanced than Linux. And of course Java runs better under Solaris (well, duh....)
Disadvantages:
1. Fewer Options for software - Most of the same software available for Linux is just a compile away, but if you expect a nice easy install with a package, you may not find it as easily as for Linux
2. Scott McNealy
3. Hardware Support - When I first installed Solaris 9 x86, the supported hardware list was so small that I had to go find 2-3 year old components just to get it to boot. Solaris 10 is supposed to be a bit better but YMMV. There's also Sun's waffling over whether Solaris x86 should even be around.
3. Scott McNealy
4. Sun - The wannabe Microsoft of the UNIX world. They will never get there though because the company is still run by.....
5. SCOTT MCNEALY!!!!
Advantages:
1. Monoculture - It's all coming from 1 place (although more and more FOSS is creeping into Solaris).
2. Stability - This has as much to do with Sun hardware than the OS. On x86 it is probably similar to a good Linux distro
3. Scalability - Solaris is geared more toward Big Iron and has had several years head start on Linux scalability. Once again, Sun boxes take better advantage of this.
4. Development - If you develop large, complex apps, Solaris has many built-in features (kernel-level debuggers, trace utils) that are still more advanced than Linux. And of course Java runs better under Solaris (well, duh....)
Disadvantages:
1. Fewer Options for software - Most of the same software available for Linux is just a compile away, but if you expect a nice easy install with a package, you may not find it as easily as for Linux
2. Scott McNealy
3. Hardware Support - When I first installed Solaris 9 x86, the supported hardware list was so small that I had to go find 2-3 year old components just to get it to boot. Solaris 10 is supposed to be a bit better but YMMV. There's also Sun's waffling over whether Solaris x86 should even be around.
3. Scott McNealy
4. Sun - The wannabe Microsoft of the UNIX world. They will never get there though because the company is still run by.....
5. SCOTT MCNEALY!!!!
Nothinman
Elite Member
So basically if he had a really big Sun box, Solaris would be a good idea? Who would have thunk it?
Frankly I never saw what was so great about Solaris, I'm not saying it's a bad OS but most of the benefits require big expensive hardware or you to sign your life away to pay for the additional features. Does it even come with a compiler out of the box (and a gcc package on the last CD doesn't count)? They're including more and more GNU tools with each release which is a huge step forward, the default userland most commercial unixes and BSDs come with is ass.
I would avoid Solaris unless you have a very specific reason to use it or until 2 or 3 releases after it's open sourced, because maybe by then it'll be really usable as a desktop machine and have a decent package manager.
Nothinman
Elite Member
Right, AFAIK there _still_ isn't an equivalent to the 'free' command found on every Linux system. How can you consider a system good if you can't even see what kind of memory usage you have?
I know you can use \ to prefix commands to have it skip aliases and such, but I'm not sure if it would help in your case.
Scalability, stability, and if you're using SPARC, I/O throughput.
Solaris is responsive with a load average of over 100 and will actually work through a load average of over 1000 and return back to you with every task completed.
While Linux has come a long way in these issues, especially scalability in the 2.6 kernel, Solaris still wins.
However, forget using Solaris on x86. Sun has never put the same effort into x86 as they do into SPARC. The hardware support is quite weak too. Perhaps Solaris 10 will change this, but they've claimed to be taking x86 seriously several times before and they haven't yet.
Nothinman
Elite Member
I have Linux on my two (well one's out of commission right now) UltraSparcs and it is a helluva lot more responsive than Solaris was.
Totally irrelevant to this thread. And if you're dealing with load averages of 1000 you would probably make things run a lot better and save a lot of money by not buying into Solaris and using multiple smaller machines.
Solaris knows that high end risc is a dead end nowadays. The big iron days are going the way of the mainframe.
(they are going to stuck in small relatively minor niche roles with few growth prospects).
Sparc is aiming for the datacenter with it's multiple threading core and they need to go to x86 to expand. That's were the future is for them.
If they don't make a big spash on the x86 then the server market is just going to get more and more and more linux-centric until it gets uncompatable with Unix stuff (that's one of Sun's big smears at Redhat, that they are using open source and propriatory standards to try to lock-in customers, which for the most part is BS.) and Solaris gets forced out of the market.
Otherwise Sun will end up like SCO and just survive making decent profits supporting legacy platforms and end up growing more and more obscure and unimportant as the RISC market dries up and they die a slow death 15-20 years from now. They need to get a large userbase now to try to prevent that from happening and ensure survival and actual growth in the forseeable future.
That and before Sun didn't take x86 seriously because x86 wasn't a threat, it wasn't compitition. Now you have Linux clustering technology + AMD64 + Linux 2.6 soon means that you can build a array of machines that will outperform and out-reliable the old big risc machines at a fraction of the price.
Say you have one risc machine you have one point of failure. Even though it's 99.99 percent reliable. Now take 95.0 percent reliable PC, not to impressive. Now take 12 95.0% reliable PC and have them mirror information and have them be used in a combo high-aviability and high-performance cluster. If done right you outperform and have better reliability then the big risc machines.
Look at how they do things like livejournal.org. That's pure Linux software, running a massive database and all of it's dynamicly created. They have 5,325,841 accounts and within the last 24 hours they have had 342,065 of them being updated by their users. They have it all running on rendundant machines, all the web server are completely diskless and run on RAM only. They need more performance? They plug another machine into the network and keep the old ones around until they aren't worth the electricity and room they are taking up. Multiple rendundant database backends, numerious machines in a distributed archatecture. All sorts of fun stuff and all custom designed and maintained by a mostly volenteer staff. All open source all using what would be considured "subpar" software by a lot of people... Lots of mysql and perl stuff. All very weird and relatively dirt cheap high performance and high aviability stuff. Users modify software on accounts, set things up to e-mail them or page them or call them on their cell phones to indicate changes and such. Having to deal with vindictive and immature user base.
Now it's not without it's issues, and is constantly evolving and changing, but you have to see the potential of this sort of thing. How much would it cost in Sparc hardware to get close to being able to do the same thing?
You have things like openssi and openmosix that are getting close to making and building clusters a almost brain-dead affar. Plug-n-play type stuff. In a few years it will be very mature. Your starting to see good distributed file systems for Linux for free and all sorts of fun stuff.
Redhat has it's stuff it's working on and is buying up software left and right and openning it up. Like GFS filing systems. Netscape's directory software, and other stuff. Also working on the "stateless Linux' type stuff were you plug a PC with a blank disk into a network and within a few minutes you have a fully functional workstation, a thick client with local disk for users that are backed up automaticly with a central server. Think Knoppix, but loading up over the network instead of the cdrom. That type of stuff. Not to also mention what Novell is doing and IBM is going to use Linux to push Power and PowerPC, and Sparc isn't close to being competative in raw performance with Power stuff right now.
All sorts of crazy stuff. People will need Sparc type stuff for a long long time and will pay big prices for it and keep Sun profitable, but it's a shrinking market. Each year that goes buy it's getting a smaller and smaller niche. Solaris desperetly needs to be part of the action. And if they pull it off it and do a good job with licensing it can be great. Linux has and will always have certain limitations in it's design and developer attitude, that rub people the wrong way. And monoculture sucks, even if it will be only Linux. Diversity = good.
But Sun says and does things very strangly...
(they are going to stuck in small relatively minor niche roles with few growth prospects).
Sparc is aiming for the datacenter with it's multiple threading core and they need to go to x86 to expand. That's were the future is for them.
If they don't make a big spash on the x86 then the server market is just going to get more and more and more linux-centric until it gets uncompatable with Unix stuff (that's one of Sun's big smears at Redhat, that they are using open source and propriatory standards to try to lock-in customers, which for the most part is BS.) and Solaris gets forced out of the market.
Otherwise Sun will end up like SCO and just survive making decent profits supporting legacy platforms and end up growing more and more obscure and unimportant as the RISC market dries up and they die a slow death 15-20 years from now. They need to get a large userbase now to try to prevent that from happening and ensure survival and actual growth in the forseeable future.
That and before Sun didn't take x86 seriously because x86 wasn't a threat, it wasn't compitition. Now you have Linux clustering technology + AMD64 + Linux 2.6 soon means that you can build a array of machines that will outperform and out-reliable the old big risc machines at a fraction of the price.
Say you have one risc machine you have one point of failure. Even though it's 99.99 percent reliable. Now take 95.0 percent reliable PC, not to impressive. Now take 12 95.0% reliable PC and have them mirror information and have them be used in a combo high-aviability and high-performance cluster. If done right you outperform and have better reliability then the big risc machines.
Look at how they do things like livejournal.org. That's pure Linux software, running a massive database and all of it's dynamicly created. They have 5,325,841 accounts and within the last 24 hours they have had 342,065 of them being updated by their users. They have it all running on rendundant machines, all the web server are completely diskless and run on RAM only. They need more performance? They plug another machine into the network and keep the old ones around until they aren't worth the electricity and room they are taking up. Multiple rendundant database backends, numerious machines in a distributed archatecture. All sorts of fun stuff and all custom designed and maintained by a mostly volenteer staff. All open source all using what would be considured "subpar" software by a lot of people... Lots of mysql and perl stuff. All very weird and relatively dirt cheap high performance and high aviability stuff. Users modify software on accounts, set things up to e-mail them or page them or call them on their cell phones to indicate changes and such. Having to deal with vindictive and immature user base.
Now it's not without it's issues, and is constantly evolving and changing, but you have to see the potential of this sort of thing. How much would it cost in Sparc hardware to get close to being able to do the same thing?
You have things like openssi and openmosix that are getting close to making and building clusters a almost brain-dead affar. Plug-n-play type stuff. In a few years it will be very mature. Your starting to see good distributed file systems for Linux for free and all sorts of fun stuff.
Redhat has it's stuff it's working on and is buying up software left and right and openning it up. Like GFS filing systems. Netscape's directory software, and other stuff. Also working on the "stateless Linux' type stuff were you plug a PC with a blank disk into a network and within a few minutes you have a fully functional workstation, a thick client with local disk for users that are backed up automaticly with a central server. Think Knoppix, but loading up over the network instead of the cdrom. That type of stuff. Not to also mention what Novell is doing and IBM is going to use Linux to push Power and PowerPC, and Sparc isn't close to being competative in raw performance with Power stuff right now.
All sorts of crazy stuff. People will need Sparc type stuff for a long long time and will pay big prices for it and keep Sun profitable, but it's a shrinking market. Each year that goes buy it's getting a smaller and smaller niche. Solaris desperetly needs to be part of the action. And if they pull it off it and do a good job with licensing it can be great. Linux has and will always have certain limitations in it's design and developer attitude, that rub people the wrong way. And monoculture sucks, even if it will be only Linux. Diversity = good.
But Sun says and does things very strangly...
For desktop use, Linux may win, but for server use, Solaris SPARC will win.
It's relevant to this thread if you want your machine to handle load spikes. If it was a normal load, yes, you obviously don't have enough hardware, but this case was a rare spike of a high enough load to cause Linux to fail, while Solaris pulled through.
Yes, it's too difficult to complete with the gigantic PC market in hardware, and microrprocessors alone are too expensive for anyone but a few companies like Intel and IBM to develop and manufacture. Sun's best hope for SPARC would be to farm it out to IBM like Apple did with PowerPC, but I don't think they're likely to try or succeed at that, hence their Opteron move.
However, moving to x86 puts them in the classic Innovator's Dilemma position a la Christensen's book, where the low end stuff doesn't make enough profit to sustain the company *and* cannabalizes the market for their high end products. That's why no mainframe maker succeeded in the minicomputer market and no minicomputer maker succeeded in the PC/workstation market. IBM was big enough to survive losing the minicomputer market, and was smart enough to spin off the PC effort into its own geographically separated group, though even that wasn't enough in the long run.
Nothinman
Elite Member
Tell that to our users complaining about their Oracle database speed on every platform except Linux. They were having the same problems on Linux, but we found a fix for it on Linux but AFAIK the same thing still running on other commercial unixes is still having problems.
The chances of getting a load spike to 1000 or even 100 is pretty low and can be avoided if you setup proper limits on things.
I've had my load up that high and Linux has never failed, sure it slows down and becomes a PITA to login and sh!t, but it keeps working.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
