INFRASTRUCTURE LEARNINGINFRASTRUCTURE LEARNING
インフラ大全インフラ大全IT INFRASTRUCTURE / LEARNING
Infrastructure learning guide
Find a lesson or lab/

Esc to close · Search terms stay on your device

LEARNING日本語

Topic lessonTopic: Backup and recovery foundations

RTO vs RPO explained: recovery timelines and backup basics

RTO limits how long a service can be down; RPO limits how much recent change can be lost, expressed as time. Read the two intervals on one timeline using a fictional report service, then distinguish recovery evidence from the mere existence of a backup.

What you will learn

Distinguish interruption, recovered data point and service restoration; assess RTO and RPO separately and identify missing evidence rather than assuming success.

Prerequisite

Know that a backup supplies material used for recovery. No AWS or SQL Server configuration experience is required. This lesson uses fictional times without live restore operations or input of real data or credentials.

RTO limits downtime; RPO limits tolerable data loss

RTO (Recovery Time Objective) sets the acceptable duration from service interruption to restored service. RPO (Recovery Point Objective) sets the acceptable data loss expressed as a time gap before the interruption. These are requirements set in advance, not the measured duration of a restore job or its backup schedule.

Keep each target separate from the result compared with it
ObjectiveWhat does it limit?Result compared in this lesson
RTO: 30 minutesMaximum acceptable service interruptionElapsed time from interruption until the agreed service restoration condition passes
RPO: 5 minutesMaximum tolerable data loss before interruption, measured as timeGap between interruption and the point represented by the recovered data

Read two intervals on the same timeline

A fictional report service fails at 10:00. We stipulate that validated recovered data contains committed changes through 09:50 and users can read the required report at 10:25. All times use the same day and Japan Standard Time (JST). These are supplied teaching facts, not AWS measurements; no new writes occur after the interruption.

One fictional outage, two different recovery intervals

Same day, JST. Cards show time order; their spacing is not a time scale. These are not AWS measurements.

  1. Recovered data point

    Committed changes through this point are stipulated to be included.

  2. Service interruption

    The internal report becomes unavailable to its users.

  3. Report check passes; service restored

    Required data and user report access are stipulated to pass.

Compare with RPO

Data-point gap: 10 minutes

10:00 − 09:50 = 10 min

0 min30 min

Exceeds the 5-minute target. The required data point is 09:55 or later.

Compare with RTO

Outage duration: 25 minutes

10:25 − 10:00 = 25 min

0 min30 min

Within the 30-minute target. Service restoration is due by 10:30.

Both bars use a 0–30 minute scale. Fill shows the supplied interval; the dashed line marks its target.

The ten-minute gap describes the time range of changes potentially absent from the restored data. It does not mean a known number of records was lost, or that every record from those minutes is irretrievable. Identify missing changes and any possible reconciliation separately.

A restore job is only part of the outage
TimeSupplied event
10:00Service interruption: start counting the outage
10:05Restore job starts
10:20Job ends; the report check is not yet complete
10:25Required data, destination and report checks pass; service is restored

Backup interval, retention and PITR answer different questions

Separate three backup conditions
ConditionQuestion it answersWhat it does not establish
Backup intervalHow often are captures scheduled or performed?Successful capture, the data point available, or successful recovery
Retention periodHow far back is restore material retained?The latest usable point or all material required for a chosen time
Point-in-time recovery (PITR)Which point within the supported window should be restored?Recovery outside the window, a guaranteed duration or business consistency

If all you know is a five-minute schedule, achievement of a five-minute RPO is unconfirmed. Check capture success, the data point covered, required material such as logs, and usable restored results. A job completion timestamp is not automatically the point represented by its data.

For Amazon RDS DB-instance PITR, inspect the actual earliest and latest restorable times. Configured retention does not establish recovery through the present instant. PITR creates a new DB instance rather than overwriting the source, so plan application cutover separately. This scope does not cover every Aurora or backup product.

Choose a usable recovery point, not merely the newest one

With the same 10:00 outage, an erroneous update at 09:54 may remain in a 09:59 recovery point. Selecting a validated clean point at 09:53 gives a seven-minute gap. Choosing correct data and meeting a five-minute RPO are separate judgments. Report the missed target and review missing changes and next actions.

  1. Record the workload, incident, RTO/RPO targets and service restoration condition; use a consistent time basis.
  2. Match required data, configuration and dependencies to the restore material, point and destination.
  3. Restore in an appropriate isolated environment and validate required data and consistency.
  4. Check the intended application destination, permissions and required reads or writes.
  5. Record cutover and completion of the restoration condition, then compare outage duration and data-point gap separately with their targets.

Existing lessons below cover SQL Server recovery models, log continuity and RDS for SQL Server restore methods. RDS SQL Server also has a cross-database transaction consistency boundary. Choosing a timestamp is not evidence that the complete business state is consistent.

Check your decisions using times and evidence

Choose an option to see the result and explanation.

Outage 10:00, validated data point 09:50, report service restored 10:25. Targets: RTO 30 minutes and RPO 5 minutes. Which result follows?
Read the answer and reason

Answer: Within RTO; outside RPO. · The 25-minute outage is within 30 minutes; the 10-minute data-point gap exceeds 5 minutes. Evaluate the objectives separately.

The outage starts at 10:00; the restore job runs from 10:05 to 10:20; the user-facing check passes at 10:25. How long until service restoration?
Read the answer and reason

Answer: 25 minutes: outage to successful user-facing check. · The stipulated restoration condition is the completed report check: 10:25 minus 10:00 is 25 minutes, distinct from the 15-minute job.

Backups are scheduled every five minutes. Does that alone show that a five-minute RPO target has been met?
Read the answer and reason

Answer: No: inspect usable recovery points and restored results. · Failed capture or missing required logs can leave an older usable point. Check the actual restore material and validation results.

A fictional RDS DB instance has a restorable window from yesterday 00:00 through today 09:56. The requested point is today 09:58. What follows?
Read the answer and reason

Answer: The supplied bounds do not permit 09:58. · A requested point must fit the actual earliest/latest restorable bounds, not merely the retention setting. Even an eligible time still needs restore and application validation.

Outage 10:00. To avoid an erroneous 09:54 update, the plan selects a validated clean point at 09:53. With a five-minute RPO target, what should be reported?
Read the answer and reason

Answer: The seven-minute gap exceeds the target; review missing changes and next actions. · 10:00 minus 09:53 is seven minutes. A necessary clean point does not make the target met. Review the missing changes and any reconciliation with stakeholders.

The restored instance is available, but required data, application connectivity and report access are untested. Is workload recovery complete?
Read the answer and reason

Answer: Not yet: complete the agreed data, connectivity and operation checks. · Separate restore status from the workload restoration condition. Record when that condition passes, then compare the outage duration with the target.

Check, then explain in your own words

  1. Separate targets from observed results.
  2. Calculate outage duration and data-point gap on one timeline.
  3. Identify backup, PITR and post-restore validation evidence.
Check your understanding: Does a five-minute backup schedule show that the 30-minute RTO and 5-minute RPO targets have been met?

No. Establish the usable data point and time until restored service separately.

A schedule is not evidence that targets were met. Required material, recovered data and the service restoration condition need validation.

Sources and conditions checked

Primary references are collected here with their scope, conditions and corresponding sections. Independent expert review and live-environment validation have not been performed.