ITS#9599 Tiered balancing implementation
This patch set changes how backends are configured, adding tiers
- groups of backends that define the strategy used to select a backend. Next tier is chosen only if all of the current backends are unreachable - we stop with LDAP_BUSY
if they are overloaded.
Three strategies are introduced:
- roundrobin - this has been the default until now
- weighted - a backend is chosen more often the higher its weight
- bestof2 - operation latency is measured and two backends are picked at random, the one with a lower latency is chosen more often