≡ Menu

Major Linux Vs UNIX Kernel Differences

UNIX has been regarded as the mother of most of the operating systems. Some of the popular members of this family Include :

  • System V Release 4(SVR4) developed by AT&T.
  • 4.4 BSD From university of California
  • AIX from IBM.
  • HP-UX from Hewlett-Packard.
  • Solaris from Sun Microsystems.


Linux is relatively a new member of this family. Linux was initially written by Linus Torvalds in 1991 for IBM compatible personal computers. As an OS, GNU Linux has experienced huge success and popularity in last 20 years with most of the commercial servers now using GNU Linux. To add to the popularity, end users have also started using Linux these days with most of the popular Laptop and PC manufactures giving GNU Linux as a per-installed OS.

For those who are still confused between Linux being an OS or kernel, Linux in true sense as written by Linus was a kernel that was written by referring to book on Unix internals (Though the Linux kernel has adopted good features from many other Unix like kernels too) while the commercially available distributions that contain utilities like graphical desktop, text editors, compilers etc on top of the Linux kernel are complete operating systems.

Though Linux Kernel borrows most of its features from Unix/Unix-Like kernels but still there are many points where the two type of kernels differ significantly. In this article, the main focus will be on these differences. The list is not exhaustive but contains the main differences.

1. Monolithic Vs Micro-kernel Approach

Monolithic kernels are those where all the kernel code runs as a single process while Micro-kernel kernels are those where the core of a kernel (that controls the different pieces of OS) runs in one process while other services like device drivers etc run as different processes. Linux follows monolithic approach while there are a couple of exceptions in Unix-Like kernels that follow Micro-kernel approach.

2. Adding/Removing features to kernel

While traditional Unix/Unix-like systems require static linking of new modules being added, Linux supports a powerful feature where-in kernel components like device drivers etc can be loaded and unloaded dynamically. This feature is known as Loadable kernel modules (LKM). Any new component can be added/removed as an LKM to the kernel. This means there is no need to compile the whole kernel again. Also, if a component is not needed, it can easily be unloaded. This feature makes Linux kernel very flexible.

3. Kernel Threading

Many Unix-Like kernels are organized as a set of kernel threads. A kernel thread can be thought of as an independent execution flow. A kernel thread can run user process or some kernel code. The basic Idea is to do context switches between the kernel threads which is less expensive than context switches between processes as threads operate in same address space. While many Unix-Like OS use kernel threads for process context switching, Linux uses kernel threads only for executing some kernel code periodically.

4. Multi-threaded application support

Almost all modern OS, be it Unix-Like or Linux distributions, support multi-threading. A multi-threaded application is one which creates more than one execution flows. These independent execution flows are known as threads. Threads are light weight processes. In most of the Unix-Like systems, light weight processes are based on kernel threads while in Linux these LWP are created by a call to function clone() which lets the application to create a separate process like fork() does but the difference being that with clone() the newly generated process can share its physical memory, opened files, address space etc. As these newly created process works in a shared environment, so they are given a different name ‘threads’. So we see that Linux and Unix/Unix-Like differ in the way multi-threaded environment is handled internally.

5. STREAMS

Streams I/O subsystem is included in most of the Unix kernels and has become a preferred interface for writing device drivers, terminal drivers etc. While on the other hand there is nothing like Streams in Linux.

6. Preemptive Vs Non-Preemptive Kernels

Preemptive kernels are the kernels which can preempt the currently executing process. It means that a process which is currently executing can be forcibly interrupted if a process with higher priority is ready for execution. On the other hand, Non preemptive kernels are those where a running process cannot be forcibly interrupted even if a higher priority process is ready for execution. Normally, Linux OS are Non preemptive while some of the Unix systems like Solaris 2.x etc are fully preemptive. Usually Real time OS have fully preemptive kernels. These days we have Linux Real time OS which have fully preemptive kernels.

So we see that though Linux was born out of the basic idea from Unix but still it differs from Unix/Unix-Like kernels in many ways. Despite of these differences Linux still inherits a lot from Unix and is still considered as a member of Unix family of kernels.

Add your comment

If you enjoyed this article, you might also like..

  1. 50 Linux Sysadmin Tutorials
  2. 50 Most Frequently Used Linux Commands (With Examples)
  3. Top 25 Best Linux Performance Monitoring and Debugging Tools
  4. Mommy, I found it! – 15 Practical Linux Find Command Examples
  5. Linux 101 Hacks 2nd Edition eBook Linux 101 Hacks Book

Bash 101 Hacks Book Sed and Awk 101 Hacks Book Nagios Core 3 Book Vim 101 Hacks Book

Comments on this entry are closed.

  • jalal hajigholamali January 9, 2012, 4:14 am

    Thanks..

  • Koutheir Attouchi January 9, 2012, 4:41 am

    Another clear distinction in kernels of Linux and UNIX: The kernel API.
    UNIX API is BSD, while Linux API is POSIX. Yes, BSD and POSIX share many functions, but not all. And even shared functions may differ in behavior and supported actions.

  • Himanshu January 9, 2012, 5:35 am

    @Koutheir Attouchi

    Thanks for the information.

  • rakesh kumar January 9, 2012, 6:05 am

    nice info ….

  • Salih Emin January 9, 2012, 6:50 am

    Nowadays, in distributions, linux kerne is by default in a “Voluntary – Preemptive” state. I belive there is an exception in Linux Arch were kernel is Preemptive.

  • Spliff January 9, 2012, 8:24 am

    Good article,

    Thanks ;D

  • Makus Mayer January 9, 2012, 9:48 am

    Unfortunately, I have to say this is the most misleading and uninformed article of thegeekstuff that I have seen to date.

    1. Monolithic Vs Micro-kernel Approach

    What you are writing is true, but it is misleading. Why? You say “while there are a couple of exceptions in Unix-Like kernels that follow Micro-kernel approach”, but you don’t give any examples of micro-kernel OSes (NeXTStep and Mac OS X would be such OSes, btw.). The OSes you mention (from Sys V to Solaris) all take the same Linux does: a monolithic kernel, so it’s hardly a distinctive feature of Linux.

    2. Adding/Removing features to kernel

    Again, what you are writing is true, but misleading. Yes, in the olden days one would have to re-link the kernel to add a new driver to it and reboot (even SunOS 4 still required this). However, most Unix-like OSes you mention have loadable kernel modules. It’s the standard nowadays and not something that is specific to Linux. It differentiates modern Unix-like OSes from their roots 30-40 years ago, but is not something that’s in any way specific to Linux.

    Sections 3-5, I agree with.

    Now for the last point — and the worst disinformation.

    6. Preemptive Vs Non-Preemptive Kernels

    What you write here is simply outright wrong. That’s all there’s to it. “Normally, Linux OS are Non preemptive while some of the Unix systems like Solaris 2.x etc are fully preemptive.” This is simply untrue (well, the Linux portion of the sentence is). I invite you to look through the kernel sources, specifically http://lxr.linux.no/linux/kernel/sched.c. You’ll find scheduling routines to interrupt processes and take the CPU away from them.

    It would look like you are mixing up two things here: real time scheduling and preemptive scheduling. Those two are not anywhere near the same. The only two OSes I know that used non-preemptive (i.e. co-operative) scheduling were Mac OS up to version 9 and Windows prior to NT. I don’t recall if the Windows 95 family had some preemptive capabilities or not, 3.1 definitely did not. No modern OS still uses non-preemptive scheduling!

    Yes, Linux is not a real-time OS by default. But no, the Linux kernel does not employ co-operative scheduling. Otherwise, say, a hung Apache process could hog the an entire web-server and bring it down, right? Nobody would want to run their corporate web-sites or databases off a Linux machine, but they clearly do, as you say yourself.

  • Karthik.P.R January 9, 2012, 10:22 am

    Very nice article.

    Thanks

  • Himanshu January 9, 2012, 8:35 pm

    @Makus Mayer

    Thanks Mr. Mayer for providing your comments. I appreciate the fact that you went through the article. Please read my view points below :

    1. Monolithic Vs Micro-kernel Approach

    ===================================================================
    @Makus Mayer
    >What you are writing is true, but it is misleading. Why? You say “while there are a >couple of exceptions in Unix-Like kernels that follow Micro-kernel approach”, but >you don’t give any examples of micro-kernel OSes (NeXTStep and Mac OS X >would be such OSes, btw.). The OSes you mention (from Sys V to Solaris) all take >the same Linux does: a monolithic kernel, so it’s hardly a distinctive feature of >Linux.
    ===================================================================

    @Himanshu
    I never mentioned anywhere in this article that only the Unix systems that I have listed (from Sys V to Solaris) are the ones I am comparing Linux with. I have clearly stated that these are some of the popular Unix-like systems.

    2. Adding/Removing features to kernel

    ==================================================================
    @Makus Mayer
    >Again, what you are writing is true, but misleading. Yes, in the olden days one >would have to re-link the kernel to add a new driver to it and reboot (even SunOS >4 still required this). However, most Unix-like OSes you mention have loadable >kernel modules. It’s the standard nowadays and not something that is specific to >Linux. It differentiates modern Unix-like OSes from their roots 30-40 years ago, but >is not something that’s in any way specific to Linux.
    ==================================================================

    @Himanshu
    Firstly, I have mentioned that I am comparing Linux with traditional Unix systems in this point (“While traditional Unix/Unix-like systems require….” ), so I don’t think this confusion should ever arise.

    Secondly, you yourself gave an example of a Unix system (SunOS 4) which still requires static linking. So where does the confusion lie now???

    Lastly, I would re-iterate it here that the Unix-Like systems that I have mentioned in the beginning of article are just some examples ( “Some of the popular members of this family Include….” ) and not the ones I am directly comparing Linux with.

    6. Preemptive Vs Non-Preemptive Kernels

    ==================================================================
    @Makus Mayer
    >What you write here is simply outright wrong. That’s all there’s to it. “Normally, >Linux OS are Non preemptive while some of the Unix systems like Solaris 2.x etc >are fully preemptive.” This is simply untrue (well, the Linux portion of the sentence >is). I invite you to look through the kernel sources, specifically http://lxr.linux.no/linux>/kernel/sched.c. You’ll find scheduling routines to interrupt processes and >take the CPU away from them.
    ==================================================================

    @Himanshu
    As per your last line, there are two aspects to it. One is to preempt processes in user mode which is very well there in Linux while other is to preempt processes while in Kernel mode which can be done if kernel is preemptive and If you read the header of this point, it says “Preemptive Vs Non-Preemptive Kernels”. The support for kernel preemption has been added in Linux2.6 but that too is not complete. Prior to that Linux kernel was non preemptive.

    ==================================================================
    @Makus Mayer
    >It would look like you are mixing up two things here: real time scheduling and >preemptive scheduling. Those two are not anywhere near the same. The only two >OSes I know that used non-preemptive (i.e. co-operative) scheduling were Mac >OS up to version 9 and Windows prior to NT. I don’t recall if the Windows 95 family >had some preemptive capabilities or not, 3.1 definitely did not. No modern OS still >uses non-preemptive scheduling!
    ==================================================================

    @Himanshu
    I am not mixing anything here. I have clearly stated “Usually Real time OS have fully preemptive kernels” and there is nothing wrong in that. I never said real time scheduling is preemptive scheduling. I think you are getting confused between preemptive scheduling and kernel preemption. In this point I am talking about kernel preemption.

    Regarding any Modern OS that uses non-preemptive scheduling , please refer to TinyOS.

    =================================================================
    @Makus Mayer
    >Yes, Linux is not a real-time OS by default. But no, the Linux kernel does not >employ co-operative scheduling. Otherwise, say, a hung Apache process could >hog the an entire web-server and bring it down, right? Nobody would want to run >their corporate web-sites or databases off a Linux machine, but they clearly do, as >you say yourself.
    ==================================================================

    @Himanshu
    Linux kernel version 2.4 was non preemptive. Support for kernel preemption has been added in Version 2.6 and as I have already said, its still not complete.

    As for your example, if the hung process was in user mode then it would easily be preempted by Linux scheduler. If it got hung while in kernel mode then also it could be preempted with any other high priority process running in user mode. Starting from Linux 2.6, the same process which is hung in kernel mode can also be preempted by any other process running in kernel mode.

    I think the major point here that confused you was preemptive scheduling vs kernel preemption. I was talking of kernel preemption while you were thinking of preemptive scheduling. Both are slightly different and I hope you got my point.

  • vinay jhedu January 12, 2012, 12:15 pm

    seems like comments are as good as article

  • rakesh January 13, 2012, 12:17 pm

    Thanx its a amazing things for me…..

  • Simon January 18, 2012, 6:18 am

    Both the author and Makus make good points.
    Great article, thanks for writing!

  • Shakil June 15, 2012, 3:25 am

    I fully agree with Makus Mayer, and Himanshu’s rebuttal is again wrong atleast for the premptive kernel. When you are defining pre-emptiveness, you have to each time give context to either user space or kernel pre-emption. And I feel after reading the article that it has not been clearly written with concise words even though the author might be fully aware of preemption concepts.