AWS wellarchitected: RTB Fabric multi-AZ/multi-Region resilience guidance rewrite
Summary
The ADVREL05 page was rewritten from generic 'Failure management' content into RTB Fabric-specific resilience guidance. The question changed to focus on surviving single-AZ and single-Region failures, the two old best-practice links (fault tolerance evaluation, DR runbooks) were replaced with five new best practices (multi-AZ bid gateways, AZ-level health monitoring, multi-Region gateways, per-Region gateway health monitoring, independent regional operational readiness), and new 'Capability intent' and 'Common issues to watch for' sections were added covering stateful dependency Multi-AZ failover, partner failover routing (Global Accelerator/Route 53), asymmetric degradation detection, and avoiding centralized control-plane dependencies.
Security assessment
This is Well-Architected reliability/availability guidance for RTB Fabric workloads (multi-AZ and multi-Region deployment, failover, monitoring). It describes resilience design patterns rather than a specific vulnerability, CVE, credential, encryption, or authorization control, so it is security-adjacent at most and does not document a security feature or fix.
Evidence
+RTB Fabric defaults to single-AZ, single-Region deployment. A failure at either scope, whether an Availability Zone impairment or a regional outage, causes a complete loss of bid connectivity unless every layer of the stack (gateways, compute, and stateful dependencies) is explicitly deployed for multi-AZ and multi-Region availability. Every layer must hold for the design to work. Multi-AZ gateways provide no protection if the bid-processing pipeline depends on a single-AZ database, and a second Region provides no protection if partner failover is undefined or capacity is under-provisioned.
Diff
diff --git a/wellarchitected/latest/video-streaming-advertising-lens/advrel05.md b/wellarchitected/latest/video-streaming-advertising-lens/advrel05.md index 99f988a..7283dc8 100644 --- a//wellarchitected/latest/video-streaming-advertising-lens/advrel05.md +++ b//wellarchitected/latest/video-streaming-advertising-lens/advrel05.md @@ -7 +7 @@ -# Failure management +Capability intentCommon issues to watch for @@ -9 +9 @@ -Failures are unavoidable, and every system eventually fails over time, especially in high volume advertising systems. Anticipate and manage failures before they happen through detection, testing, and quick recovery. +# RTB Fabric resilience @@ -11 +11,3 @@ Failures are unavoidable, and every system eventually fails over time, especiall -ADVREL05: How do you continuously evaluate the resilience of your advertising workload to meet availability and recovery requirements? +RTB Fabric defaults to single-AZ, single-Region deployment. A failure at either scope, whether an Availability Zone impairment or a regional outage, causes a complete loss of bid connectivity unless every layer of the stack (gateways, compute, and stateful dependencies) is explicitly deployed for multi-AZ and multi-Region availability. Every layer must hold for the design to work. Multi-AZ gateways provide no protection if the bid-processing pipeline depends on a single-AZ database, and a second Region provides no protection if partner failover is undefined or capacity is under-provisioned. + +ADVREL05: How do you keep your RTB Fabric workload available during single-AZ and single-Region failures? @@ -12,0 +15,26 @@ ADVREL05: How do you continuously evaluate the resilience of your advertising wo +[ADVREL05-BP01 Deploy partner-facing bid gateways across multiple Availability Zones](./advrel05-bp01.html) +[ADVREL05-BP02 Monitor Availability Zone-level health and detect asymmetric failures](./advrel05-bp02.html) +[ADVREL05-BP03 Deploy partner-facing bid gateways in multiple AWS Regions](./advrel05-bp03.html) +[ADVREL05-BP04 Monitor partner-facing bid gateway health independently per Region](./advrel05-bp04.html) +[ADVREL05-BP05 Maintain independent operational readiness in each Region](./advrel05-bp05.html) + +## Capability intent + + * RTB Fabric gateways are deployed across at least two Availability Zones, colocated with the Availability Zones where DSP/SSP compute runs, so that a single-AZ failure doesn't interrupt bid connectivity. + + * Every stateful dependency in the critical bid path (databases, caches, and message queues) is Multi-AZ with automatic failover tested against the bid response latency budget. + + * RTB Fabric gateways are deployed in at least two AWS Regions, with exchange partners able to route to either Region through a deliberate failover architecture (partner-managed, [AWS Global Accelerator](https://aws.amazon.com/global-accelerator/), or [Amazon Route 53](https://aws.amazon.com/route53/)). + + * Availability Zone-aware and Region-aware monitoring detects asymmetric and partial degradation before it erodes bid win rates unnoticed. + + * Each Region maintains independent operational readiness (deployment artifacts, capacity, and automated failover) so the surviving Region operates without depending on the impaired Region's control plane. + + + + +## Common issues to watch for + + * Multi-AZ gateways paired with a single-AZ database or cache, leaving the bid pipeline exposed despite gateway redundancy. + + * Gateways deployed in Availability Zones that have no compute fleet, adding cross-AZ latency without any resilience benefit. @@ -14 +42 @@ ADVREL05: How do you continuously evaluate the resilience of your advertising wo -Continuous resilience evaluation of advertising workloads can be achieved through two main approaches. + * Monitoring that treats Availability Zone health as binary (up or down) and misses partial degradation such as increased latency or packet loss. @@ -16 +44 @@ Continuous resilience evaluation of advertising workloads can be achieved throug -Perform regular fault tolerance testing and assessment using tools like AWS Gamedays, Well-Architected Framework Reviews, and Support Countdowns. Additionally, create and test disaster recovery procedures through documented runbooks and restoration processes, which helps you quickly recover during incidents. + * Centralizing deployment artifacts or operational tooling in the primary Region, so a regional impairment removes the ability to manage the surviving Region. @@ -18 +46 @@ Perform regular fault tolerance testing and assessment using tools like AWS Game -###### Best practices + * Relying on scaling up the secondary Region during an event when the control plane needed to provision gateways may itself be impaired. @@ -20 +48 @@ Perform regular fault tolerance testing and assessment using tools like AWS Game - * [ADVREL05-BP01 Perform routine evaluation of your workload's fault tolerance capabilities](./advrel05-bp01.html) + * Aggregating gateway metrics across Regions, which masks a single degraded Region. @@ -22 +50 @@ Perform regular fault tolerance testing and assessment using tools like AWS Game - * [ADVREL05-BP02 Create disaster recovery (DR) runbooks, and regularly test documented backup and restoration processes](./advrel05-bp02.html) + * Automatic failback to a still-recovering Region, causing a second outage. @@ -33 +61 @@ To use the Amazon Web Services Documentation, Javascript must be enabled. Please -ADVREL04-BP02 Deploy new code or resources in staggered phases, separated by sufficient time, to verify that the changes are successful +ADVREL04-BP05 Back up data in multiple locations with consideration for regulatory requirements @@ -35 +63 @@ ADVREL04-BP02 Deploy new code or resources in staggered phases, separated by suf -ADVREL05-BP01 Perform routine evaluation of your workload's fault tolerance capabilities +ADVREL05-BP01 Deploy partner-facing bid gateways across multiple Availability Zones