BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//sf26-agenda//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:SESS-28@sf26-agenda
DTSTAMP:20261020T151709Z
DTSTART:20261105T193000Z
DTEND:20261105T201500Z
SUMMARY:When an SLA Isn't Enough: Cluster Linking as the Last Line of Defen
 se for Brazil's Capital Markets
LOCATION:Room 2008
DESCRIPTION:Kaio Lebid\n\nA clearing house sits at the heart of a capital m
 arket. It guarantees settlement\, manages counterparty risk\, and cannot s
 top — an outage doesn't degrade a feature\, it freezes a national market
 . As part of its multi-year partnership with Nasdaq to build a next-genera
 tion clearing platform\, B3 — the Brazilian stock exchange — is modern
 izing this Tier-0 workload to the cloud\, with Confluent Cloud as the cent
 ral nervous system: the backbone for messaging\, system-to-system integrat
 ion\, and real-time data streaming across the entire clearing flow. This t
 alk tells that story from the trenches. We'll explain why we chose Apache 
 Kafka and Confluent Cloud as the integration hub for an ultra-critical pla
 tform\, how we model clearing events as streams\, and the governance and o
 perational discipline required to run streaming at this level of criticali
 ty — including a SaaS product layered on top of one of the most critical
  platforms in the Brazilian market. The core of the session is resilience.
  For a workload where "the platform is down" is not an acceptable sentence
 \, a managed SLA alone isn't enough — we needed a defensible answer to "
 what happens if we lose Confluent itself?" Our answer is a dedicated conti
 nuity layer built on Cluster Linking. We'll go deep on the mechanics: byte
 -for-byte\, offset-preserving asynchronous replication that keeps a standb
 y cluster continuously hydrated with topic data\, consumer offsets\, and A
 CLs — with no Connect cluster to operate. We'll cover our failover desig
 n (promote / fail-forward)\, how we reason about RTO and RPO under a regio
 nal or provider outage [insert your real targets]\, how Schema Linking kee
 ps consumers reading immediately after failover\, and the trade-offs we ma
 de. Attendees leave with a concrete blueprint for running mission-critical
 \, regulated workloads on Confluent: positioning Cluster Linking as a cont
 inuity layer — not just a migration tool — treating mirroring lag as a
  first-class SLO\, rehearsing failover\, and the architectural decisions t
 hat let a financial market infrastructure sleep at night. If you operate a
 nything that genuinely cannot go down\, this is the playbook.
URL:https://sf26.rmoff.net/s/when-an-sla-isnt-enough-cluster-linking-as-the
 -last-line-of-28
END:VEVENT
END:VCALENDAR