> 
> I have tried to run libmicro-0.4.0 in XVM Guest,
> but it takes very very long time to finish some
> workloads, like following:
>
> Running:           pthread_8 for      0.52335 seconds
> Running:          pthread_32 for     99.42079 seconds
> Running:         pthread_128 for    709.45210 seconds
> Running:         pthread_512 for   2173.69252 seconds
> Running:             fork_10 for    450.65546 seconds
> Running:            fork_100 for   2422.17060 seconds
> Running:           fork_1000 for  12677.38952 seconds
> Running:             exit_10 for    479.75481 seconds
> Running:            exit_100 for   5319.76236 seconds
> Running:           exit_1000 for  11960.19162 seconds

Yep, these number look really bad...

> This performance is unacceptable. Does
> anyone run libmicro in XVM guest and meet
> similar performance issue?


Not yet.

I just tested...


Box #1, AMD Athlon(tm) 64 X2 Dual Core Processor 6400+, 8GB memory

Metal, 64-bit snv_89_jk:

Running:           pthread_8 for        0.03313 seconds
Running:          pthread_32 for        0.16763 seconds
Running:         pthread_128 for        0.85011 seconds
Running:         pthread_512 for        3.54108 seconds
Running:                fork_10 for     0.96588 seconds
Running:            fork_100 for        4.64858 seconds
Running:           fork_1000 for        26.34765 seconds
Running:                exit_10 for     0.95890 seconds
Running:            exit_100 for        9.20730 seconds
Running:           exit_1000 for        24.42260 seconds
Running:        exit_10_nolibc for      0.87233 seconds
Running:                exec for        1.64558 seconds
Running:                system for      0.29083 seconds


64-bit xen 3.1.2, 64-bit snv_89_jk dom0:

Running:           pthread_8 for        0.07088 seconds
Running:          pthread_32 for        0.37521 seconds
Running:         pthread_128 for        1.77242 seconds
Running:         pthread_512 for        7.45081 seconds
Running:                fork_10 for     2.34170 seconds
Running:            fork_100 for        11.30274 seconds
Running:           fork_1000 for        59.89658 seconds
Running:                exit_10 for     2.37156 seconds
Running:            exit_100 for        22.60985 seconds
Running:           exit_1000 for        59.26271 seconds
Running:        exit_10_nolibc for      2.20487 seconds
Running:                exec for        3.74100 seconds
Running:                system for      0.58775 seconds

64-bit xen 3.1.2, 64-bit snv_89_jk dom0, 64-bit snv_79_jk PV domU:

Running:           pthread_8 for        0.06508 seconds
Running:          pthread_32 for        0.38477 seconds
Running:         pthread_128 for        2.37908 seconds
Running:         pthread_512 for        10.55652 seconds
Running:                fork_10 for     2.35764 seconds
Running:            fork_100 for        12.30732 seconds
Running:           fork_1000 for        65.83688 seconds
Running:                exit_10 for     2.43565 seconds
Running:            exit_100 for        24.66313 seconds
Running:           exit_1000 for        65.25513 seconds
Running:        exit_10_nolibc for      2.20230 seconds
Running:                exec for        2.96338 seconds
Running:                system for      0.37897 seconds

32-bit xen 3.1.2, 32-bit snv_89_jk dom0:

Running:           pthread_8 for        0.08169 seconds
Running:          pthread_32 for        0.40919 seconds
Running:         pthread_128 for        1.93781 seconds
Running:         pthread_512 for        7.91462 seconds
Running:                fork_10 for     5.93390 seconds
Running:            fork_100 for        29.17303 seconds
Running:           fork_1000 for    150.23701 seconds
Running:                exit_10 for     5.93749 seconds
Running:            exit_100 for        57.74059 seconds
Running:           exit_1000 for    150.18952 seconds
Running:        exit_10_nolibc for      6.03767 seconds
Running:                exec for        9.41334 seconds
Running:                system for      1.18549 seconds

32-bit xen 3.1.2, 32-bit snv_89_jk dom0, 32-bit snv_79_jk PV domU:

Running:           pthread_8 for        0.04871 seconds
Running:          pthread_32 for        0.34750 seconds
Running:         pthread_128 for        1.95965 seconds
Running:         pthread_512 for        8.54885 seconds
Running:                fork_10 for     10.86304 seconds
Running:            fork_100 for        49.61916 seconds
Running:           fork_1000 for    268.37711 seconds
Running:                exit_10 for     12.75652 seconds
Running:            exit_100 for    103.24694 seconds
Running:           exit_1000 for    265.21563 seconds
Running:        exit_10_nolibc for      11.75072 seconds
Running:                exec for        9.49524 seconds
Running:                system for      0.75156 seconds


- xVM / xen + dom0 is slower than metal
- 32-bit xVM is slower than 64-bit xVM
- 32-bit domU is slower than dom0
   (but that does not happen in the 64-bit case)


Box #2, Intel(R) Core(TM)2 Quad CPU    Q6600  @ 2.40GHz, 8GB memory
is running xen 3.2.0 and a 32-bit gentoo linux dom0

32-bit xen 3.2, PV solaris-domU

Running:           pthread_8 for        0.07845 seconds
Running:          pthread_32 for        0.56077 seconds
Running:         pthread_128 for        2.83015 seconds
Running:         pthread_512 for        13.30005 seconds
Running:                fork_10 for     3.59635 seconds
Running:            fork_100 for        17.53722 seconds
Running:           fork_1000 for        92.09840 seconds
Running:                exit_10 for     3.63996 seconds
Running:            exit_100 for        34.96151 seconds
Running:           exit_1000 for        92.49952 seconds
Running:        exit_10_nolibc for      3.27794 seconds
Running:                exec for        6.22332 seconds
Running:                system for      0.74665 seconds

32-bit xen 3.2, PV gentoo domU

Running:           pthread_8 for        0.04229 seconds
Running:          pthread_32 for        0.20588 seconds
Running:         pthread_128 for        0.90911 seconds
Running:         pthread_512pthread_create: Cannot allocate memory
                            ...
                              for       3.35929 seconds
Running:                fork_10 for     0.66098 seconds
Running:            fork_100 for        3.59144 seconds
Running:           fork_1000 for        21.62967 seconds
Running:                exit_10 for     0.64355 seconds
Running:            exit_100 for        6.92005 seconds
Running:           exit_1000 for        20.78521 seconds
Running:        exit_10_nolibc for      0.53173 seconds
Running:                exec for        1.18908 seconds
Running:                system for      0.34060 seconds


linux domU fork / exec performancs is better than Solaris
domU fork / exec performance
 
 
This message posted from opensolaris.org
_______________________________________________
xen-discuss mailing list
[email protected]

Reply via email to