Finding the Sweet Spot — Coordination vs Apologies
more coordination: fewer inconsistency apologies, more outage apologies — total has a minimum
A inconsistency
B outages
TOTAL (A + B)
apologies
you must
make
(counts)
synchronous coordination
none
single leader / consensus / XA per op
A: apologies for INCONSISTENCY
constraint violations, lost updates
B: apologies for OUTAGES
unavailability, stalls, spreading faults
TOTAL apologies (A + B)
sweet spot: neither too many
inconsistencies nor outages
use sync coordination only where needed
small part of the app runs under it
oversold warehouse
apologize + discount
overbooked flight
refund / upgrade
ATM overdraft
fees, daily limits bound risk
cross-org settlement
corrections inevitable
ATM cash payout
cannot be recovered
Coordination cuts both ways
coordination reduces inconsistency
apologies BUT it reduces performance
and availability — raising outage
apologies (local faults escalate,
cross-region stalls).
Cheap vs expensive apologies
LEFT: tolerate violations — cheap fix:
warehouse restock, flight refund,
overdraft fee.
RIGHT: coordinate — cash once paid
out of an ATM is gone.
No zero-apology point
you cannot reach zero apologies —
find the best trade-off for your needs,
keep integrity, relax timeliness.
compensate later (compensating txn).