[tarantool-patches] Re: [PATCH 6/9] vinyl: set range size automatically

[Vladimir Davydov]

On Wed, Feb 06, 2019 at 12:23:31PM +0300, Vladimir Davydov wrote:
> On Tue, Feb 05, 2019 at 08:09:09PM +0300, Konstantin Osipov wrote:
> > * Vladimir Davydov <vdavydov.dev at gmail.com> [19/01/21 06:58]:
> > > +int64_t
> > > +vy_lsm_range_size(struct vy_lsm *lsm)
> > > +{
> > > +	/* Use the configured range size if available. */
> > > +	if (lsm->opts.range_size > 0)
> > > +		return lsm->opts.range_size;
> > > +	/*
> > > +	 * It doesn't make much sense to create too small ranges.
> > > +	 * Limit the max number of ranges per index to 1000 and
> > > +	 * never create ranges smaller than 16 MB.
> > > +	 */
> > > +	enum { MIN_RANGE_SIZE = 16 * 1024 * 1024 };
> > > +	enum { MAX_RANGE_COUNT = 1000 };
> > > +	/*
> > > +	 * Ideally, we want to compact roughly the same amount of
> > > +	 * data after each dump so as to avoid IO bursts caused by
> > > +	 * simultaneous major compaction of a bunch of ranges,
> > > +	 * because such IO bursts can lead to a deviation of the
> > > +	 * LSM tree from the configured shape and, as a result,
> > > +	 * increased read amplification.  To achieve that, we need
> > > +	 * to have at least as many ranges as the number of dumps
> > > +	 * it takes to trigger major compaction in a range.
> > > +	 */
> > > +	int range_count = vy_lsm_dumps_per_compaction(lsm);
> > > +	range_count = MIN(range_count, MAX_RANGE_COUNT);
> > > +	int64_t range_size = lsm->stat.disk.last_level_count.bytes /
> > > +						(range_count + 1);
> > > +	range_size = MAX(range_size, MIN_RANGE_SIZE);
> > > +	return range_size;
> > > +}
> > 
> > OK, you could say the value is rarely used, so can be calculated
> > each time it is used, but why instead not recalculate it on each
> > major compaction? This would spare us from technical debt and
> > having to think about potential performance bottleneck in the
> > future.
> 
> Well, okay, I think we can do that, too.

Come to think of it, I'd prefer not to, because this would one more
member to struct vy_lsm, which we would have to update on dump and
compaction - it's more error prone IMO. Since calculation of the target
range size is trivial and called rarely (on split/coalesce), I think we
should leave it as is.