search for: usecs

Hi, another issue I''m seeing with 3.0-rc2 and Xen is that there is an unexpectedly high amount of kernel reserved memory. I suspect that Linux allocates page table entries and corresponding data structures for the whole 6GB areas of the provided ''physical RAM map'' even though it has rather big unusable holes in it. [ 0.000000] BIOS-provided physical RAM map: [

Hi folks, The following are initial virtio-scsi + target vhost benchmark results using multiple target LUNs per vhost and multiple virtio PCI adapters to scale the total number of virtio-scsi LUNs into a single KVM guest. The test setup is currently using 4x SCSI LUNs per vhost WWPN, with 8x virtio PCI adapters for a total of 32x 500MB ramdisk LUNs into a single guest, along with each backend

Hi folks, The following are initial virtio-scsi + target vhost benchmark results using multiple target LUNs per vhost and multiple virtio PCI adapters to scale the total number of virtio-scsi LUNs into a single KVM guest. The test setup is currently using 4x SCSI LUNs per vhost WWPN, with 8x virtio PCI adapters for a total of 32x 500MB ramdisk LUNs into a single guest, along with each backend

...rmance Events: unsupported p6 CPU model 44 no PMU driver, software events only. [ 33.870316] SMP alternatives: switching to UP code [ 33.905773] ACPI: Core revision 20090903 [ 33.914571] calling migration_init+0x0/0x43 @ 1 [ 33.919130] initcall migration_init+0x0/0x43 returned 0 after 3906 usecs [ 33.925856] calling spawn_ksoftirqd+0x0/0x43 @ 1 [ 33.930645] initcall spawn_ksoftirqd+0x0/0x43 returned 0 after 3906 usecs [ 33.937469] calling init_call_single_data+0x0/0x64 @ 1 [ 33.942755] initcall init_call_single_data+0x0/0x64 returned 0 after 0 usecs [ 33.949862] calling spawn_...

- Use more friendly output with GiB and MiB/s. - Measure time per operation, providing finer grain stats - Add missing stats for flush I hope that these changes will help to understand and imporve virt-v2v performance. Nir Soffer (3): filters: stats: Show size in GiB, rate in MiB/s filters: stats: Measure time per operation filters: stats: Add flush stats filters/stats/stats.c | 117

...ZER; -static uint64_t pread_ops, pread_bytes; -static uint64_t pwrite_ops, pwrite_bytes; -static uint64_t trim_ops, trim_bytes; -static uint64_t zero_ops, zero_bytes; -static uint64_t extents_ops, extents_bytes; -static uint64_t cache_ops, cache_bytes; +static uint64_t pread_ops, pread_bytes, pread_usecs; +static uint64_t pwrite_ops, pwrite_bytes, pwrite_usecs; +static uint64_t trim_ops, trim_bytes, trim_usecs; +static uint64_t zero_ops, zero_bytes, zero_usecs; +static uint64_t extents_ops, extents_bytes, extents_usecs; +static uint64_t cache_ops, cache_bytes, cache_usecs; static inline double...

- Use more friendly output with GiB and MiB/s - Measure time per operation, providing finer grain stats - Add total stats for understanding system throughput - Add missing stats for flush I hope that these changes will help to understand and improve virt-v2v performance. Changes since v1: - Keep bytes values - Increase precision to 0.001 GiB and 0.001 MiB/s - Add total stats - Show time before

...define MiB 1048576.0 +#define GiB 1073741824.0 +#define USEC 1000000.0 + static char *filename; static bool append; static FILE *fp; @@ -64,34 +68,34 @@ static uint64_t extents_ops, extents_bytes; static uint64_t cache_ops, cache_bytes; static inline double -calc_bps (uint64_t bytes, int64_t usecs) +calc_mibps (uint64_t bytes, int64_t usecs) { - return 8.0 * bytes / usecs * 1000000.; + return bytes / MiB / usecs * USEC; } static inline void print_stats (int64_t usecs) { - fprintf (fp, "elapsed time: %g s\n", usecs / 1000000.); + fprintf (fp, "elapsed time: %.3f s\n&...

...8576.0 +#define GiB 1073741824.0 +#define USEC 1000000.0 + static char *filename; static bool append; static FILE *fp; @@ -70,29 +74,35 @@ static stat extents_st = { "extents" }; static stat cache_st = { "cache" }; static inline double -calc_bps (uint64_t bytes, int64_t usecs) +calc_rate (uint64_t bytes, int64_t usecs) { - return 8.0 * bytes / usecs * 1000000.; + return bytes / MiB / usecs * USEC; } static inline void print_stat (const stat *st, int64_t usecs) { if (st->ops > 0) - fprintf (fp, "%s: %" PRIu64 " ops, %" PRIu64 &quo...

This is the third attempt to make the stats filter better. This time I kept the changes minimal to make it eaier to make progress. I tried the idea of showing global stats in separate "total" section, but it became messy and hard to implemnt, so I tried the simpler solution of showing both operation rate and total rate in the operation line. Nir Soffer (5): filters: stats: Add

...add = bucket->rate * usec / 1000000; add = MIN (add, bucket->capacity - bucket->level); diff --git a/filters/stats/stats.c b/filters/stats/stats.c index 639ceacf..3b633cdf 100644 --- a/filters/stats/stats.c +++ b/filters/stats/stats.c @@ -164,7 +164,7 @@ stats_unload (void) int64_t usecs; gettimeofday (&now, NULL); - usecs = tvdiff_usec (&start_t, &now); + tvdiff_usec (&start_t, &now, &usecs); if (fp && usecs > 0) { ACQUIRE_LOCK_FOR_CURRENT_SCOPE (&lock); print_stats (usecs); @@ -247,10 +247,10 @@ static inline void record...

...1 file changed, 52 insertions(+), 18 deletions(-) diff --git a/filters/stats/stats.c b/filters/stats/stats.c index 6bef65c..2c92c65 100644 --- a/filters/stats/stats.c +++ b/filters/stats/stats.c @@ -62,6 +62,7 @@ typedef struct { const char *name; uint64_t ops; uint64_t bytes; + uint64_t usecs; } stat; /* This lock protects all the stats. */ @@ -80,28 +81,42 @@ calc_rate (uint64_t bytes, int64_t usecs) } static inline void -print_stat (const stat *st, int64_t usecs) +print_stat (const stat *st) { if (st->ops > 0) - fprintf (fp, "%s: %" PRIu64 " ops, %&...

...queue_mask 0x1 --coalesce tx-packets 10 > Would set max_packets=10 to VQ 1. > $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10 > Would set max_packets=10 to VQ 0. > $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce > Queue: 0 > Adaptive RX: off TX: off > stats-block-usecs: 0 > sample-interval: 0 > pkt-rate-low: 0 > pkt-rate-high: 0 > > rx-usecs: 222 > rx-frames: 0 > rx-usecs-irq: 0 > rx-frames-irq: 256 > > tx-usecs: 222 > tx-frames: 0 > tx-usecs-irq: 0 > tx-frames-irq: 256 > > rx-usecs-low: 0 > rx-frame...

>Search took -125.376129 seconds I double checked with a handheld stopwatch, and at no point did the hands spin backwards. Known problem?

...s 10 > > Would set max_packets=10 to VQ 1. > > $ ethtool -Q eth5 queue_mask 0x1 --coalesce rx-packets 10 > > Would set max_packets=10 to VQ 0. > > $ ethtool -Q eth5 queue_mask 0x1 --show-coalesce > > Queue: 0 > > Adaptive RX: off TX: off > > stats-block-usecs: 0 > > sample-interval: 0 > > pkt-rate-low: 0 > > pkt-rate-high: 0 > > > > rx-usecs: 222 > > rx-frames: 0 > > rx-usecs-irq: 0 > > rx-frames-irq: 256 > > > > tx-usecs: 222 > > tx-frames: 0 > > tx-usecs-irq: 0 > >...

2017-10-10 8:25 GMT+02:00 Karan Sandha <ksandha at redhat.com>: > Hi Gandalf, > > We have multiple tuning to do for small-files which decrease the time for > negative lookups , meta-data caching, parallel readdir. Bumping the server > and client event threads will help you out in increasing the small file > performance. > > gluster v set <vol-name> group

I am trying to get the handytone 286 to make a very simple call to * and having problems. It registers with * just fine, but when I place a call (to echo test, for example), the RTP stream seems to have problems opening. Here is there error I get in *: WARNING[98311]: File chan_sip.c, Line 464 (retrans_pkt): Maximum retries exceeded on call 20d1c411-e210-5f3d-3f88-19035c8fcb26@192.168.2.6 for

Holding the vblk->lock across kick causes poor scalability in SMP guests. If one CPU is doing virtqueue kick and another CPU touches the vblk->lock it will have to spin until virtqueue kick completes. This patch reduces system% CPU utilization in SMP guests that are running multithreaded I/O-bound workloads. The improvements are small but show as iops and SMP are increased. Khoa Huynh

Holding the vblk->lock across kick causes poor scalability in SMP guests. If one CPU is doing virtqueue kick and another CPU touches the vblk->lock it will have to spin until virtqueue kick completes. This patch reduces system% CPU utilization in SMP guests that are running multithreaded I/O-bound workloads. The improvements are small but show as iops and SMP are increased. Khoa Huynh

I noticed that R was burning up 100% of a CPU when a call to Sys.sleep() was made. Upon investigation, I discovered that R_checkActivityEx() in src/unix/sys-std.c was putting the entire timeout (in usec) into the struct timeval tv_usec member, leaving tv_sec set to 0. I don't know about other unix variants, but Solaris requires that the timeout value be normalized (i.e. 0 <= tv_usec