On Tue, Jul 16, 2013 at 11:03 PM, shiek kaleem <shiek.kaleem@gmail.com>wrote:
Hi Guys ,
Just the same question from my side is there any guidelines i can follow if I have created my own OS.(LFS)
On Mon, Jul 15, 2013 at 12:26 PM, bill4carson <bill4carson@gmail.com>wrote:
Hello Steven
On 2013年07月15日 10:56, Steven Zhou wrote:
Hi,
We have developed our private OS and we want to benchmark the performance of it, including time of context switch, interrupt latency, IO output and so on ... ^^^^^^^^^^^^^^^ To what degree does "private" goes? Is your private OS an newly created one from scratch or based on one of open source version?
We want to study from Linux firstly, so could you guys give me some
guide of Linux benchmark testing, including test methodologies, test code and so on.
Thanks in advance.
-- Best Regards.
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
-- 八百里秦川尘土飞扬,三千万老陕齐吼秦腔。
--bill
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
_______________________________________________ Kernelnewbies mailing list Kernelnewbies@kernelnewbies.org http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
Hi, Benchmarking is a difficult ask. You need to define the matrix of what all things you would be measure and for what kind of workload. This data should be directly influenced by the application of the OS, where and how your OS would be used, what are critical performance areas you need to focus on. First of all whats the type of kernel you guys have implemented. The main thing is, what were the design decisions taken while implementing the policies, how should they perform in theory. After listing all these things down, you need to design benchmarking software modules for that though you can have a look at some existing once for the other OSes. For linux, 1. systemtap. 2. perf. 3. Vtunes from intel. 4. ftrace. Also, you would be required to emulate the workload to the real load that would be on your OS and do benchmarking. - Rohan