Re: [PATCH v3 2/4] ceph: track average/stdev r/w/m latency

[Venky Shankar <vshankar@redhat.com>]

On Wed, Sep 22, 2021 at 5:47 PM Jeff Layton <jlayton@redhat.com> wrote:
>
> On Tue, 2021-09-21 at 18:37 +0530, Venky Shankar wrote:
> > Update the math involved to closely mimic how its done in
> > user land. This does not make a lot of difference to the
> > execution speed.
> >
> > Signed-off-by: Venky Shankar <vshankar@redhat.com>
> > ---
> >  fs/ceph/metric.c | 63 +++++++++++++++++++++---------------------------
> >  fs/ceph/metric.h |  3 +++
> >  2 files changed, 31 insertions(+), 35 deletions(-)
> >
> > diff --git a/fs/ceph/metric.c b/fs/ceph/metric.c
> > index 226dc38e2909..ca758bff69ca 100644
> > --- a/fs/ceph/metric.c
> > +++ b/fs/ceph/metric.c
> > @@ -245,6 +245,7 @@ int ceph_metric_init(struct ceph_client_metric *m)
> >
> >       spin_lock_init(&m->read_metric_lock);
> >       m->read_latency_sq_sum = 0;
> > +     m->avg_read_latency = 0;
> >       m->read_latency_min = KTIME_MAX;
> >       m->read_latency_max = 0;
> >       m->total_reads = 0;
> > @@ -255,6 +256,7 @@ int ceph_metric_init(struct ceph_client_metric *m)
> >
> >       spin_lock_init(&m->write_metric_lock);
> >       m->write_latency_sq_sum = 0;
> > +     m->avg_write_latency = 0;
> >       m->write_latency_min = KTIME_MAX;
> >       m->write_latency_max = 0;
> >       m->total_writes = 0;
> > @@ -265,6 +267,7 @@ int ceph_metric_init(struct ceph_client_metric *m)
> >
> >       spin_lock_init(&m->metadata_metric_lock);
> >       m->metadata_latency_sq_sum = 0;
> > +     m->avg_metadata_latency = 0;
> >       m->metadata_latency_min = KTIME_MAX;
> >       m->metadata_latency_max = 0;
> >       m->total_metadatas = 0;
> > @@ -322,20 +325,25 @@ void ceph_metric_destroy(struct ceph_client_metric *m)
> >               max = new;                      \
> >  }
> >
> > -static inline void __update_stdev(ktime_t total, ktime_t lsum,
> > -                               ktime_t *sq_sump, ktime_t lat)
> > +static inline void __update_latency(ktime_t *ctotal, ktime_t *lsum,
> > +                                 ktime_t *lavg, ktime_t *min, ktime_t *max,
> > +                                 ktime_t *sum_sq, ktime_t lat)
> >  {
> > -     ktime_t avg, sq;
> > +     ktime_t total, avg;
> >
> > -     if (unlikely(total == 1))
> > -             return;
> > +     total = ++(*ctotal);
> > +     *lsum += lat;
> > +
> > +     METRIC_UPDATE_MIN_MAX(*min, *max, lat);
> >
> > -     /* the sq is (lat - old_avg) * (lat - new_avg) */
> > -     avg = DIV64_U64_ROUND_CLOSEST((lsum - lat), (total - 1));
> > -     sq = lat - avg;
> > -     avg = DIV64_U64_ROUND_CLOSEST(lsum, total);
> > -     sq = sq * (lat - avg);
> > -     *sq_sump += sq;
> > +     if (unlikely(total == 1)) {
> > +             *lavg = lat;
> > +             *sum_sq = 0;
> > +     } else {
> > +             avg = *lavg + div64_s64(lat - *lavg, total);
> > +             *sum_sq += (lat - *lavg)*(lat - avg);
> > +             *lavg = avg;
> > +     }
> >  }
> >
> >  void ceph_update_read_metrics(struct ceph_client_metric *m,
> > @@ -343,23 +351,18 @@ void ceph_update_read_metrics(struct ceph_client_metric *m,
> >                             unsigned int size, int rc)
> >  {
> >       ktime_t lat = ktime_sub(r_end, r_start);
> > -     ktime_t total;
> >
> >       if (unlikely(rc < 0 && rc != -ENOENT && rc != -ETIMEDOUT))
> >               return;
> >
> >       spin_lock(&m->read_metric_lock);
> > -     total = ++m->total_reads;
> >       m->read_size_sum += size;
> > -     m->read_latency_sum += lat;
> >       METRIC_UPDATE_MIN_MAX(m->read_size_min,
> >                             m->read_size_max,
> >                             size);
> > -     METRIC_UPDATE_MIN_MAX(m->read_latency_min,
> > -                           m->read_latency_max,
> > -                           lat);
> > -     __update_stdev(total, m->read_latency_sum,
> > -                    &m->read_latency_sq_sum, lat);
> > +     __update_latency(&m->total_reads, &m->read_latency_sum,
> > +                      &m->avg_read_latency, &m->read_latency_min,
> > +                      &m->read_latency_max, &m->read_latency_sq_sum, lat);
>
> Do we really need to calculate the std deviation on every update? We
> have to figure that in most cases, this stuff will be collected but only
> seldom viewed.
>
> ISTM that we ought to collect just the bare minimum of info on each
> update, and save the more expensive calculations for the tool presenting
> this info.

Yeh, that's probably the plan we want going forward when introducing
new metrics.

FWIW, we could start doing it with this itself. It's just that the
user land PRs are approved and those do the way it is done here.

I'm ok with moving math crunching to the tool and it should not be a
major change to this patchset.

>
> >       spin_unlock(&m->read_metric_lock);
> >  }
> >
> > @@ -368,23 +371,18 @@ void ceph_update_write_metrics(struct ceph_client_metric *m,
> >                              unsigned int size, int rc)
> >  {
> >       ktime_t lat = ktime_sub(r_end, r_start);
> > -     ktime_t total;
> >
> >       if (unlikely(rc && rc != -ETIMEDOUT))
> >               return;
> >
> >       spin_lock(&m->write_metric_lock);
> > -     total = ++m->total_writes;
> >       m->write_size_sum += size;
> > -     m->write_latency_sum += lat;
> >       METRIC_UPDATE_MIN_MAX(m->write_size_min,
> >                             m->write_size_max,
> >                             size);
> > -     METRIC_UPDATE_MIN_MAX(m->write_latency_min,
> > -                           m->write_latency_max,
> > -                           lat);
> > -     __update_stdev(total, m->write_latency_sum,
> > -                    &m->write_latency_sq_sum, lat);
> > +     __update_latency(&m->total_writes, &m->write_latency_sum,
> > +                      &m->avg_write_latency, &m->write_latency_min,
> > +                      &m->write_latency_max, &m->write_latency_sq_sum, lat);
> >       spin_unlock(&m->write_metric_lock);
> >  }
> >
> > @@ -393,18 +391,13 @@ void ceph_update_metadata_metrics(struct ceph_client_metric *m,
> >                                 int rc)
> >  {
> >       ktime_t lat = ktime_sub(r_end, r_start);
> > -     ktime_t total;
> >
> >       if (unlikely(rc && rc != -ENOENT))
> >               return;
> >
> >       spin_lock(&m->metadata_metric_lock);
> > -     total = ++m->total_metadatas;
> > -     m->metadata_latency_sum += lat;
> > -     METRIC_UPDATE_MIN_MAX(m->metadata_latency_min,
> > -                           m->metadata_latency_max,
> > -                           lat);
> > -     __update_stdev(total, m->metadata_latency_sum,
> > -                    &m->metadata_latency_sq_sum, lat);
> > +     __update_latency(&m->total_metadatas, &m->metadata_latency_sum,
> > +                      &m->avg_metadata_latency, &m->metadata_latency_min,
> > +                      &m->metadata_latency_max, &m->metadata_latency_sq_sum, lat);
> >       spin_unlock(&m->metadata_metric_lock);
> >  }
> > diff --git a/fs/ceph/metric.h b/fs/ceph/metric.h
> > index 103ed736f9d2..0af02e212033 100644
> > --- a/fs/ceph/metric.h
> > +++ b/fs/ceph/metric.h
> > @@ -138,6 +138,7 @@ struct ceph_client_metric {
> >       u64 read_size_min;
> >       u64 read_size_max;
> >       ktime_t read_latency_sum;
> > +     ktime_t avg_read_latency;
> >       ktime_t read_latency_sq_sum;
> >       ktime_t read_latency_min;
> >       ktime_t read_latency_max;
> > @@ -148,6 +149,7 @@ struct ceph_client_metric {
> >       u64 write_size_min;
> >       u64 write_size_max;
> >       ktime_t write_latency_sum;
> > +     ktime_t avg_write_latency;
> >       ktime_t write_latency_sq_sum;
> >       ktime_t write_latency_min;
> >       ktime_t write_latency_max;
> > @@ -155,6 +157,7 @@ struct ceph_client_metric {
> >       spinlock_t metadata_metric_lock;
> >       u64 total_metadatas;
> >       ktime_t metadata_latency_sum;
> > +     ktime_t avg_metadata_latency;
> >       ktime_t metadata_latency_sq_sum;
> >       ktime_t metadata_latency_min;
> >       ktime_t metadata_latency_max;
>
> --
> Jeff Layton <jlayton@redhat.com>
>


-- 
Cheers,
Venky