Uncover needs
Information for preparing a proposal in CertiSearch
You can learn the process without a client or confirmed quantities. When preparing a proposal, record what is known and mark unknown information as pending; do not invent figures or choose a service before comparing alternatives.
Outcome and user defined, with a measure of success.
Initiative identified as new, migration or modernization.
Facts separated from assumptions; relevant gaps recorded.
What result, user and success measure warrant this case and what remains outside the first proposal?
What regions will have data and users, what network routes are allowed and what volume will cross clouds or data centers?
What roles, sensitive data, audit, encryption, secrets and access restrictions should be shown?
What availability, RTO, RPO, maintenance window and care responsibility require each component?
What budget, currency, monthly horizon, environments and growth can the client review, and what SKU, quantities and monthly hours support the complementary resources?
Who will operate every component and with what coverage, alarms, logs and skills?
Link needs to capabilities and candidate alternatives.
Represent components, flows, owners and decisions.
After identifying needs and comparing alternatives, the measurements depend on the workload. Only where relevant, collect data for the following questions according to the data, analytics, container, API or artificial intelligence workload.
What's the daily volume, file size, growth and maximum process window?
How many authors and consumers will have, how many will coincide at peak hour and what actions will they conduct?
How much CPU and memory do pods request and consume under normal and peak load, including platform components, and what OCPU and RAM per worker meet that demand?
How many requests monthly and by second at peak are expected, with what size and latency of response?
How many transactions, input and output tokens, and concurrent requests are expected, and what is their size distribution?
Each quantity has a unit, period and source.
Components without a verifiable rate are shown without a price.
Target and transition costs are separate, with uncertainty visible.
What type and quantity of CPU, memory per worker, number of workers and instances does each workload require, how many hours per month will they run, and what verified monthly total of AIDP units applies?
Building the compute stage by environment and type of work.
Record minimums, maximums and idle time per workload; verify the available configuration and measure with representative data. For heterogeneous topologies, request the monthly total of AIDP units verified using the official conversion, including catalog, Spark and AI Compute; do not multiply it by hours again or add it to a profile covering the same resources.
- Oracle AI Data Platform → Dimension and costs → Performance and capacity
Sources
How many GB will data, versions, checkpoints and volumes occupy, and what retention and monthly growth are expected?
Distinguish logical size, persistent storage and recovery data.
Separate Object Storage and other billing components Do not confuse input size with final storage.
- Oracle AI Data Platform → Dimension and costs → Economic viability
Sources
What is the first measurable outcome: a dashboard, an analytical model, a pipeline, a data product or an AI application?
To agree with the result that justifies the solution and its acceptance criterion.
Proof to contrast: The first deliverable determines the main engine and prevents overdimensioning the platform. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Identification of needs → Business Result
- Oracle Autonomous AI Lakehouse → Identification of needs → Business Result
Sources
Who will users be in the first six months and what can they do today?
Identify skills, tools and responsibilities to surrender and maintain the case.
Rounding direction: SQL and autoservice guide Lakehouse and Data Studio PySpark teams guide AI Data Platform. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Identification of needs → Delivery capacity
- Oracle Autonomous AI Lakehouse → Identification of needs → Delivery capacity
Sources
What sources, formats, sizes and frequency of update are necessary in the first phase?
To detect dependencies and compatibility conditions to be tested.
Proof to contrast: Check source and destination compatibility before choosing. Do not decide only by volume. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Access and data catalogue → Compatibility and integration
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Compatibility and integration
Sources
How much logic should be programmed and what part can be solved with assistants, SQL or visual flows?
Identify skills, tools and responsibilities to surrender and maintain the case.
Tackling: Data Studio and Data Transforms for guided work Spark when code and distributed processing add value. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Delivery capacity
- Oracle Autonomous AI Lakehouse → Analytics SQL → Delivery capacity
Sources
What query SLA, concurrency and data freshness are committed to the business?
To get a representative load and its measurable targets before dimensioning.
Proof to contrast: Separate production and consumption layers if both require different conduct. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Dimension and costs → Performance and capacity
- Oracle Autonomous AI Lakehouse → Dimension and costs → Performance and capacity
Sources
What budget, go-live date and operations team are available?
Compare scenarios with quantity and periods sustained in actual use.
Proof to contrast: Compare a first viable product with the cost of a joint solution, including compute at rest. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Dimension and costs → Economic viability
- Oracle Autonomous AI Lakehouse → Dimension and costs → Economic viability
Sources
What Warehouse uses today, what version and where is it deployed?
To detect dependencies and compatibility conditions to be tested.
To contrast: Distinguish heterogeneous replacement Oracle to Autonomous migration compatibility should be measured. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Migration and modernization → Compatibility and integration
- Oracle Autonomous AI Lakehouse → Migration and modernization → Compatibility and integration
Sources
How many schemas, tables, views, packages, critical queries and dashboards exist?
To identify what exists and how far a verifiable transition can be achieved.
Countering: To inventory dependencies and prioritize the 20 processes and consultations with best impact. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Migration and modernization → Transition from today's state
- Oracle Autonomous AI Lakehouse → Migration and modernization → Transition from today's state
Sources
What are the current p95, concurrency, load windows and monthly peaks?
To get a representative load and its measurable targets before dimensioning.
Proof to contrast: The POC should match or improve the measured base line, not a suitable synthetic load. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Dimension and costs → Performance and capacity
- Oracle Autonomous AI Lakehouse → Dimension and costs → Performance and capacity
Sources
What percentage of ETL is implemented in SQL or PL/SQL, and what portion requires processing outside the database?
To identify what exists and how far a verifiable transition can be achieved.
Tackling: dominant SQL favours modernizing Lakehouse Additional Spark only if it reduces complexity or scale. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Migration and modernization → Transition from today's state
- Oracle Autonomous AI Lakehouse → Migration and modernization → Transition from today's state
Sources
Is querying files or Iceberg without moving data required, or must all storage be restructured?
Distinguish access, copies, publication and ownership of data.
Proof to contrast: Test links, catalogs, cache and acceleration before promise movement or zero copies. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Access and data catalogue → Access to data
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Access to data
Sources
What downtime, synchronization and rollback arrangements does the business accept?
To define continuity and recovery with a verifiable test.
Countering: select migration method and co-existence plan Validate data reconciliation. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Migration and modernization → Continuity
- Oracle Autonomous AI Lakehouse → Migration and modernization → Continuity
Sources
Do you wish to preserve the Warehouse as a consumption layer while adding a data lake?
To detect dependencies and compatibility conditions to be tested.
Proof to contrast: A joint architecture can add AI Data Platform for engineering without replacing the SQL service. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Access and data catalogue → Compatibility and integration
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Compatibility and integration
Sources
Where do Spark, Hadoop, Databricks or other pipelines run today, and who maintains them?
To identify what exists and how far a verifiable transition can be achieved.
Retiring: Identify relies on run time, supplier and current operation practice. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Migration and modernization → Transition from today's state
Sources
How many jobs are there, what languages, versions and libraries and what are critical?
To detect dependencies and compatibility conditions to be tested.
Counting: To select a sample of simple, medium and complex jobs to measure actual migration. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Migration and modernization → Compatibility and integration
Sources
What are current formats and catalogs: Delta, Iceberg, Parquet or other?
To detect dependencies and compatibility conditions to be tested.
Proof to contrast: To check compatibility of reading, writing, evolution of scheme and governance. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Access and data catalogue → Compatibility and integration
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Compatibility and integration
Sources
Are there Streaming, state, Checkpoint, low latency or particular Connectors?
To detect dependencies and compatibility conditions to be tested.
Proof to contrast: Compare the matrix supported by AI Data Platform with each critical flow. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Compatibility and integration
Sources
What costs and runtimes do the current clusters have, including idle periods?
Compare scenarios with quantity and periods sustained in actual use.
Proof to contrast: Compare total cost by job and by month, not only price by CPU. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Dimension and costs → Economic viability
Sources
How are code, dependencies, secrets and parameters deployed between environments?
Identify skills, tools and responsibilities to surrender and maintain the case.
Proof to contrast: To prove an actual promotion with tasks, notebooks, libraries and recovery. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Operation and continuity → Delivery capacity
Sources
Who consumes the final data and with what SQL or BI query pattern?
To get a representative load and its measurable targets before dimensioning.
Proof to contrast: If there's high analytic concurrence, evaluate Lakehouse as an additional service layer. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle Autonomous AI Lakehouse → Analytics SQL → Performance and capacity
Sources
What percentage of value is delivered through SQL queries and dashboards versus data transformation?
To agree with the result that justifies the solution and its acceptance criterion.
Rounding direction: SQL and boards favor Lakehouse distributed transformation favours AI Data Platform. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Business Result
- Oracle Autonomous AI Lakehouse → Analytics SQL → Business Result
Sources
Who builds the flow: analysts and SQL specialists or PySpark engineers and developers?
Identify skills, tools and responsibilities to surrender and maintain the case.
Proof to contrast: The available skill reduces delivery time and maintenance costs. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Identification of needs → Delivery capacity
- Oracle Autonomous AI Lakehouse → Identification of needs → Delivery capacity
Sources
Do the client agree to maintain code, libraries, tests and deployments?
Identify skills, tools and responsibilities to surrender and maintain the case.
Proof to contrast: A low code tolerance favours Data Studio and SQL custom code needs to favor AI Data Platform. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Operation and continuity → Delivery capacity
- Oracle Autonomous AI Lakehouse → Operation and continuity → Delivery capacity
Sources
Can business rules be expressed in SQL and visual maps or require algorithms, UDF and special logic?
Identify skills, tools and responsibilities to surrender and maintain the case.
Test Retirement: Test Data Transforms and SQL first If they don't cover the case, measure Spark. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Delivery capacity
- Oracle Autonomous AI Lakehouse → Analytics SQL → Delivery capacity
Sources
Must the same team build, operate and monitor multiple pipelines with dependencies and retries?
Identify skills, tools and responsibilities to surrender and maintain the case.
Proof to contrast: AI Data Platform's Workflows have more alignment. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Delivery capacity
Sources
What's the daily volume, file size, growth and maximum process window?
To get a representative load and its measurable targets before dimensioning.
Proof to contrast: Compare costs and duration of Lakehouse consultations with an actual job Spark. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Dimension and costs → Performance and capacity
- Oracle Autonomous AI Lakehouse → Dimension and costs → Performance and capacity
Sources
Are there mixed, semi-structured sources or a great number of archives that require distributed transformation?
To detect dependencies and compatibility conditions to be tested.
To contrast: AI Data Platform gain relevance The volume itself doesn't decide. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Compatibility and integration
Sources
Must the data remain in the lake without being copied, and what format is it in?
Distinguish access, copies, publication and ownership of data.
Countering: Both can consult outside sources under different conditions. Validate Iceberg, routes and catalogues. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Access and data catalogue → Access to data
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Access to data
Sources
Do you have to write or maintain tables Delta as its main format?
To detect dependencies and compatibility conditions to be tested.
Retiring: AI Data Platform offers a natural engineering path about Delta. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Compatibility and integration
Sources
Is the main requirement to query external Iceberg with SQL and accelerate repeated reads?
To get a representative load and its measurable targets before dimensioning.
Countering: Lakehouse with catalogs, Lake Cache or Data Lake Accelerator is a strong candidate. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Performance and capacity
Sources
Are there inherited Spark processes that are intended to migrate with few changes?
To detect dependencies and compatibility conditions to be tested.
To contrast: AI Data Platform reduces redesign if APIs, libraries and runtime are supported. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Migration and modernization → Compatibility and integration
Sources
Is continuous or micro-batch processing required, with checkpoint and recovery requirements?
To define continuity and recovery with a verifiable test.
Proof to contrast: To evaluate Spark Structured Streaming at AI Data Platform and its exact boundaries of origin and destination. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Data engineering → Continuity
Sources
Who consumes the data, and with what concurrency, latency and SLA?
To get a representative load and its measurable targets before dimensioning.
A SQL and Repetitive BI analytic layer favours Lakehouse. To be measured with actual data. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Dimension and costs → Performance and capacity
- Oracle Autonomous AI Lakehouse → Dimension and costs → Performance and capacity
Sources
Do users need assistants to load, link, scour and share data without writing code?
Identify skills, tools and responsibilities to surrender and maintain the case.
Tackling: Data Studio is a direct candidate To measure cover of each particular task. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle Autonomous AI Lakehouse → Analytics SQL → Delivery capacity
Sources
Are APIs, models, agents or knowledge bases required with data?
To detect dependencies and compatibility conditions to be tested.
Proof to contrast: To specify the function of AI. The AI.native capabilities of AI Data Platform require Autonomous AI Lakehouse 26ai or later. engineering without these functions should be independently evaluated. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → AI applications about data → Compatibility and integration
- Oracle Autonomous AI Lakehouse → AI applications about data → Compatibility and integration
Sources
Is data sharing between teams and engines without copying required?
Distinguish access, copies, publication and ownership of data.
Proof to contrast: To design catalogues and shared formats A combination can be more effective. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Access and data catalogue → Access to data
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Access to data
Sources
Where do data and users reside: OCI, other clouds or data centre?
To detect dependencies and compatibility conditions to be tested.
Proof to contrast: Check connectivity, permissions and available functions by region and deployment. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Connectivity and residence → Compatibility and integration
- Oracle Autonomous AI Lakehouse → Connectivity and residence → Compatibility and integration
Sources
What policies of identity, roles, audit, encryption and residence are required?
Identify controls, responsible and required government evidence.
Proof to contrast: require end-to-end demonstration Do not assume equivalence between platform RBEC and basic privileges. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Security and Government → Control and traceability
- Oracle Autonomous AI Lakehouse → Security and Government → Control and traceability
Sources
What catalogue is the true source and how are metadata and lineage managed?
Identify controls, responsible and required government evidence.
Proof to contrast: Define owner from metadata and update frequency. review line boundaries in streaming. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Access and data catalogue → Control and traceability
- Oracle Autonomous AI Lakehouse → Access and data catalogue → Control and traceability
Sources
What does the life cycle require: Git, change review, CI / CD, rollback and interenvironment promotion?
Identify skills, tools and responsibilities to surrender and maintain the case.
Proof to contrast: run an actual deployment test with service and version capabilities. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Operation and continuity → Delivery capacity
Sources
Who will respond to incidents, and what evidence do they need: query plan, rejected load, Spark UI or executor logs?
To define continuity and recovery with a verifiable test.
Test direction: The operating model should be followed by the dominant engine. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Operation and continuity → Continuity
- Oracle Autonomous AI Lakehouse → Operation and continuity → Continuity
Sources
What is the acceptable total cost of compute, storage, networking, logs and complementary services?
Compare scenarios with quantity and periods sustained in actual use.
To be tested: To model both alternatives from measured use and not from nominal prices. To confirm with an evidence of the case and the conditions of the documentation; That's not an automatic selection.
- Oracle AI Data Platform → Dimension and costs → Economic viability
- Oracle Autonomous AI Lakehouse → Dimension and costs → Economic viability
Sources
What ECPU, storage at GB, monthly hours and license regime support the Lakehouse alternative?
Record billing units and the SQL service capacity assumption.
Compare regular and beak load Validate configuration, autoscaling and license before commenting. A ECPU does not amount to a OCPU.
- Oracle Autonomous AI Lakehouse → Dimension and costs → Economic viability
Sources
What result, user and success measure warrant this case and what remains outside the first proposal?
Define scope and success criterion before choosing products.
To register current status, target, date and owner to validate; Do not derive a product only by the project label.
- Oracle AI Data Platform → Identification of needs → Business Result
- OCI API Gateway → Identification of needs → Business Result
- OCI Generative AI → Identification of needs → Business Result
- Oracle Autonomous AI Lakehouse → Identification of needs → Business Result
- Oracle Analytics Cloud → Identification of needs → Business Result
- OCI Kubernetes Engine (OKE) → Identification of needs → Business Result
Sources
What regions will have data and users, what network routes are allowed and what volume will cross clouds or data centers?
Discover residence, latency and traffic that affect the design.
Draw from and destination of traffic and confirm regional service availability separate public and private access and data output.
- Oracle AI Data Platform → Connectivity and residence → Compatibility and integration
- OCI API Gateway → Connectivity and residence → Compatibility and integration
- OCI Generative AI → Connectivity and residence → Compatibility and integration
- Oracle Autonomous AI Lakehouse → Connectivity and residence → Compatibility and integration
- Oracle Analytics Cloud → Connectivity and residence → Compatibility and integration
- OCI Kubernetes Engine (OKE) → Connectivity and residence → Compatibility and integration
Sources
What roles, sensitive data, audit, encryption, secrets and access restrictions should be shown?
Transform policies into verifiable controls.
To assign owner and evidence to each monitoring permissions OCI, permissions data and permissions application are independently validated.
- Oracle AI Data Platform → Security and Government → Control and traceability
- OCI API Gateway → Security and Government → Control and traceability
- OCI Generative AI → Security and Government → Control and traceability
- Oracle Autonomous AI Lakehouse → Security and Government → Control and traceability
- Oracle Analytics Cloud → Security and Government → Control and traceability
- OCI Kubernetes Engine (OKE) → Security and Government → Control and traceability
Sources
What availability, RTO, RPO, maintenance window and care responsibility require each component?
To define continuity according to the impact of the failure.
To design and test recovery of the set; The SLA of a service does not replace the continuity of the solution.
- Oracle AI Data Platform → Operation and continuity → Continuity
- OCI API Gateway → Operation and continuity → Continuity
- OCI Generative AI → Operation and continuity → Continuity
- Oracle Autonomous AI Lakehouse → Operation and continuity → Continuity
- Oracle Analytics Cloud → Operation and continuity → Continuity
- OCI Kubernetes Engine (OKE) → Operation and continuity → Continuity
Sources
What budget, currency, monthly horizon, environments and growth can the client review, and what SKU, quantities and monthly hours support the complementary resources?
To prepare comparable investment and operation scenarios.
Separate verified quantities from assumptions and record unpriced components without treating them as free. For each necessary complement, confirm the exact SKU, quantity in its Oracle unit and active hours per month if the metric is hourly. Distinguish storage, networking, observability and additional compute; do not duplicate resources included in another profile.
- Oracle AI Data Platform → Dimension and costs → Economic viability
- OCI API Gateway → Dimension and costs → Economic viability
- OCI Generative AI → Dimension and costs → Economic viability
- Oracle Autonomous AI Lakehouse → Dimension and costs → Economic viability
- Oracle Analytics Cloud → Dimension and costs → Economic viability
- OCI Kubernetes Engine (OKE) → Dimension and costs → Economic viability
Sources
Who will operate every component and with what coverage, alarms, logs and skills?
Check that the design has actual operational capacity.
Assign liability, fault sign and care path to agree to training if a divide exists.
- Oracle AI Data Platform → Operation and continuity → Delivery capacity
- OCI API Gateway → Operation and continuity → Delivery capacity
- OCI Generative AI → Operation and continuity → Delivery capacity
- Oracle Autonomous AI Lakehouse → Operation and continuity → Delivery capacity
- Oracle Analytics Cloud → Operation and continuity → Delivery capacity
- OCI Kubernetes Engine (OKE) → Operation and continuity → Delivery capacity
Sources
What applications, data, versions, integrations and owners form the current state and what dependencies require them to move them together?
To create a transition inventory with verifiable units.
Distinguish what to retain, migrate, replace and retire; assign an owner and a test to each dependency.
- Oracle AI Data Platform → Migration and modernization → Transition from today's state
- OCI API Gateway → Migration and modernization → Transition from today's state
- OCI Generative AI → Migration and modernization → Transition from today's state
- Oracle Autonomous AI Lakehouse → Migration and modernization → Transition from today's state
- Oracle Analytics Cloud → Migration and modernization → Transition from today's state
- OCI Kubernetes Engine (OKE) → Migration and modernization → Transition from today's state
Sources
What consumption, latency, availability, errors and observed costs form the basic line and what period do they come from?
Avoid comparing the destination against an assumption of its origin.
Record the unit, percentile, interval and source of each measurement; acceptance must use a comparable workload.
- Oracle AI Data Platform → Migration and modernization → Performance and capacity
- OCI API Gateway → Migration and modernization → Performance and capacity
- OCI Generative AI → Migration and modernization → Performance and capacity
- Oracle Autonomous AI Lakehouse → Migration and modernization → Performance and capacity
- Oracle Analytics Cloud → Migration and modernization → Performance and capacity
- OCI Kubernetes Engine (OKE) → Migration and modernization → Performance and capacity
Sources
Which interfaces, formats, extensions, versions, licenses and behaviors must be preserved or changed?
Identify incompatibilities before promising equivalence.
To prepare a representative sample and record coverage, necessary changes and boundaries; A similar API does not guarantee functional equivalence.
- Oracle AI Data Platform → Migration and modernization → Compatibility and integration
- OCI API Gateway → Migration and modernization → Compatibility and integration
- OCI Generative AI → Migration and modernization → Compatibility and integration
- Oracle Autonomous AI Lakehouse → Migration and modernization → Compatibility and integration
- Oracle Analytics Cloud → Migration and modernization → Compatibility and integration
- OCI Kubernetes Engine (OKE) → Migration and modernization → Compatibility and integration
Sources
How many GB should be moved, how much change per day and how much bandwidth and actual window are available for copies and synchronization?
Dimension initial and incremental synchronization.
Relate data volume, effective transfer rate and available time window; verify the mechanism specific to each service and reconcile the final data.
- Oracle AI Data Platform → Migration and modernization → Transition from today's state
- OCI API Gateway → Migration and modernization → Transition from today's state
- OCI Generative AI → Migration and modernization → Transition from today's state
- Oracle Autonomous AI Lakehouse → Migration and modernization → Transition from today's state
- Oracle Analytics Cloud → Migration and modernization → Transition from today's state
- OCI Kubernetes Engine (OKE) → Migration and modernization → Transition from today's state
Sources
What functions have to be performed at the same time, for how many months and who will decide how soon to complete co-existence?
Make the period with dual resources and time dependencies visible.
Distinguish the monthly cost of destination from the cost and duration of transition avoid adding them as permanent expenditure.
- Oracle AI Data Platform → Migration and modernization → Economic viability
- OCI API Gateway → Migration and modernization → Economic viability
- OCI Generative AI → Migration and modernization → Economic viability
- Oracle Autonomous AI Lakehouse → Migration and modernization → Economic viability
- Oracle Analytics Cloud → Migration and modernization → Economic viability
- OCI Kubernetes Engine (OKE) → Migration and modernization → Economic viability
Sources
Which steps, owners, cutover window, point of no return and rollback triggers does the business accept?
Prepare a controlled and reportable transition.
Document criteria for proceeding or rolling back, and rehearse recovery; rolling back traffic and rolling back data may require different actions.
- Oracle AI Data Platform → Migration and modernization → Continuity
- OCI API Gateway → Migration and modernization → Continuity
- OCI Generative AI → Migration and modernization → Continuity
- Oracle Autonomous AI Lakehouse → Migration and modernization → Continuity
- Oracle Analytics Cloud → Migration and modernization → Continuity
- OCI Kubernetes Engine (OKE) → Migration and modernization → Continuity
Sources
What functional, data, security and performance tests will approve the transition and who will sign acceptance?
To define sufficient evidence to complete migration.
Compare results with the baseline and record unresolved defects; retire the source system only when the agreed criteria are met.
- Oracle AI Data Platform → Migration and modernization → Transition from today's state
- OCI API Gateway → Migration and modernization → Transition from today's state
- OCI Generative AI → Migration and modernization → Transition from today's state
- Oracle Autonomous AI Lakehouse → Migration and modernization → Transition from today's state
- Oracle Analytics Cloud → Migration and modernization → Transition from today's state
- OCI Kubernetes Engine (OKE) → Migration and modernization → Transition from today's state
Sources
What changes to operation or architecture bring measurable value and what can be postponed to reduce risk?
Separate technical relocation to modernise the process.
Prioritize changes by outcome and dependency; do not require a complete redesign or software sprints to produce a proposal.
- Oracle AI Data Platform → Migration and modernization → Business Result
- OCI API Gateway → Migration and modernization → Business Result
- OCI Generative AI → Migration and modernization → Business Result
- Oracle Autonomous AI Lakehouse → Migration and modernization → Business Result
- Oracle Analytics Cloud → Migration and modernization → Business Result
- OCI Kubernetes Engine (OKE) → Migration and modernization → Business Result
Sources
How many authors and consumers will have, how many will coincide at peak hour and what actions will they conduct?
Characterize concurrence and pattern of use.
To measure the combination of navigation, creation and execution The total number of registered users does not by itself describe the load.
- Oracle Analytics Cloud → Analytics visual → Performance and capacity
Sources
What dashboards, workbooks, semantic models and fixed-format reports are needed, and which are critical?
Link the analytics with the editing functions.
To inventory content and permissions check that the chosen edition and service cover each function.
- Oracle Analytics Cloud → Analytics visual → Compatibility and integration
Sources
Which sources will be queried live or through extracted data, how many GB will be loaded, and how often must they be refreshed?
To define connectivity and production of the analytical data.
Separate the time for charging and update consultation. Test credentials, connectors and network access.
- Oracle Analytics Cloud → Analytics visual → Access to data
Sources
How many rows, columns and MB production reports and how many scheduled exports or deliveries coincide?
Dimension of report work as well as interaction.
To check the size and modality boundaries chosen; step up tasks when their concurrence damages interactive use.
- Oracle Analytics Cloud → Dimension and costs → Performance and capacity
Sources
What edition, license, user-based or capacity-based billing, OCPU and monthly hours apply to each environment?
To get the billing profile of the alternative.
Validate availability and scaling options; separate production and nonproduction. Confirm the exact SKU according to edition, license and billing mode; do not add users and OCPU as though they were the same billing mode.
- Oracle Analytics Cloud → Dimension and costs → Economic viability
Sources
What users can see each dataset and what filters, identity and audit have to be maintained?
To design consistent access between source and analytics.
Check access to the source and permissions content with representative users. Do not infer row security simply by belonging to a group OCI.
- Oracle Analytics Cloud → Security and Government → Control and traceability
Sources
What contents, connections, calendars and permissions of the current analytics system should be migrated and how will their equivalency be established?
Determining the transition of analytics and their acceptance.
Test a sample of contents, re-connect sources and compare results, times and safety before withdrawing the source.
- Oracle Analytics Cloud → Migration and modernization → Transition from today's state
Sources
What containers and services will be run, how many minimum and maximum replicas they require and what do they maintain state?
To describe deployment units and data dependence.
Separate applications, jobs and platform services identify persistence and distribution requirements.
- OCI Kubernetes Engine (OKE) → containers run → Compatibility and integration
Sources
How much CPU and memory do pods request and consume under normal and peak load, including platform components, and what OCPU and RAM per worker meet that demand?
Convert load measurements to nodes capacity.
Consider system requests, boundaries, system resources and operational range; Do not compute nodes using only nominal memory.
- OCI Kubernetes Engine (OKE) → Dimension and costs → Performance and capacity
Sources
What signal will trigger scaling, what peak load must be absorbed, and what limits, startup times and constraints apply?
Check how the capacity will respond to the shift of load.
Distinguish escalates from replicas and nodes; Test startup, bounds and behaviour if there's no capacity available.
- OCI Kubernetes Engine (OKE) → containers run → Performance and capacity
Sources
How many clusters and environments will there be, what kind of cluster and nodes require and how many hours monthly will they operate?
To prepare units and liability of the platform.
Reason for isolation and nodes modality account for flat monitoring, workers and supplements according to the chosen profile.
- OCI Kubernetes Engine (OKE) → Dimension and costs → Economic viability
Sources
What persistent volumes, boot disks, VPU per GB, load balancers, data egress and logs does each environment need?
Include dependencies that do not suit the price of nodes.
Link each complement to the component using it and to a quantity. In the managed profile, confirm boot GB per worker and VPU per GB of that volume; persistent volumes, traffic and observability are sized separately.
- OCI Kubernetes Engine (OKE) → Dimension and costs → Economic viability
Sources
What model or GPU workload will be hosted, how much memory and how many GPU per replica does it require, and who maintains the runtime, image and drivers?
To evaluate own AI load accommodation.
A GPU load at OKE requires shape and image compatibility and operation of the application. does not equal Dedicated AI Clusters from OCI Generative AI.
- OCI Kubernetes Engine (OKE) → containers run → Monitoring model and its operation
Sources
What policies for upgrade, distribution, backup and recovery will be tested for applications and data?
Agreed to secure operation of the cluster and its loads.
Prove maintenance and failure of nodes with representative services the availability of flat monitoring does not prove data recovery.
- OCI Kubernetes Engine (OKE) → Operation and continuity → Continuity
Sources
Which Kubernetes versions, APIs, images, charts, volumes and dependencies of the current cluster must be adapted?
Identify actual migration work of containers.
Validate compatibility and persistent state by application Test deployment, traffic cut and reversion before moving the set.
- OCI Kubernetes Engine (OKE) → Migration and modernization → Transition from today's state
Sources
What apis, routes, backends and consumers will be published and what should be public or private?
To define area and network boundaries.
Draw client, gateway and backend To check connectivity and responsibilities for each leap.
- OCI API Gateway → Publication and protection of APIs → Compatibility and integration
Sources
How many requests monthly and by second at peak are expected, with what size and latency of response?
Separate instantaneous billing volume.
To include re-attempts and stationary in volume; Prove that boundaries and backend tolerate beak.
- OCI API Gateway → Dimension and costs → Performance and capacity
Sources
What authentication, authorization, validation and transformations does each route require and who manages the credentials?
Convert policies to verifiable entry behaviour.
Prove permitted and rejected access and assign ownership to policies do not place a backend as an authorization.
- OCI API Gateway → Publication and protection of APIs → Control and traceability
Sources
What subscribers require quotas or requests limits and how will excesses be reported?
To design backend protection and consumer monitoring.
The use plans work with requests API A share of requests does not constitute a tokens budget of a model.
- OCI API Gateway → Publication and protection of APIs → Control and traceability
Sources
If an AI gateway is requested, are model or provider selection, token budgets, response evaluation and guardrails also required?
Separate API access from AI orchestration functions.
Enterprise AI Gateway describes a solution pattern. Those functions require specific components and validation. they are not assumed to be included in OCI API Gateway.
- OCI API Gateway → Government of consumption of AI → Monitoring model and its operation
Sources
What metrics, logs, alerts, retention and response to errors do each API and consumer require?
To make the complete course of the request observable.
To co-relate dateway and backend without registration of unnecessary sensitive data; defining decision makers and consumers.
- OCI API Gateway → Operation and continuity → Continuity
Sources
What API contracts, policies, certificates, DNS names and customers should maintain compatibility during the change?
Plan co-existence and acceptance of API access.
Prove contracts and mistakes, migrate consumers by group and define traffic revisions. includes expiry of certificates and credentials.
- OCI API Gateway → Migration and modernization → Transition from today's state
Sources
What specific task will the model solve, which model is selected, and does it require text, images, audio, embeddings or reranking?
Identify input and output modality and quality criterion.
Validate each capability in the selected model and region; the service does not offer every modality in every model. Identify the exact public SKU for the model and consumption type; input, output, cache, audio and context may require separate lines.
- OCI Generative AI → Inference of models → Compatibility and integration
Sources
What set of examples, languages, expected responses and criteria will measure model quality, safety and usefulness?
Define acceptance before choosing capacity.
Compare candidates with the same sample and document human errors and review A velocity metric doesn't prove useful.
- OCI Generative AI → Inference of models → Business Result
Sources
How many transactions, input and output tokens, and concurrent requests are expected, and what is their size distribution?
Dimension consumption and load mix.
Separate input and output and measure their percentiles; record peaks and retries. For tokens, confirm the monthly quantity per SKU type; for transactions, apply the Oracle definition for the model without assuming equivalence between token, call and transaction.
- OCI Generative AI → Dimension and costs → Performance and capacity
Sources
What time at first token, total latency and tokens by second does the case require at their target concurrence?
To define a comparable performance test.
To measure with actual length and concurrence Do not extrapolate a sign from another requests mixture as a guarantee.
- OCI Generative AI → Inference of models → Performance and capacity
Sources
Which mode should be compared: on-demand, dedicated cluster or import of a compatible model?
Clarify if shared consumption, dedicated capacity or monitoring of the imported model are requested.
Dedicated Model Import uses dedicated capacity and an endpoint isn't synonymous with all Dedicated AI Cluster. Check restrictions and availability by model.
- OCI Generative AI → Model accommodation → Monitoring model and its operation
Sources
What bursts, limitation errors, reattempts and acceptable degradation do the case admit under demand?
Check the application response to variable capacity.
Throtting at demand may vary; Test backoff and measure user experience without assuming a guaranteed RPS.
- OCI Generative AI → Inference of models → Continuity
Sources
What model, cluster type and count, verified AI Units, replicas and hosting or fine-tuning hours does the dedicated capacity require?
Building an administered infrastructure scenario.
Confirm regional capacity, quotas and model commitment; separate hosting from fine-tuning and verify SKU, AI Units, clusters and hours. Rerank uses its cluster-and-hours profile. If a GPU reservation is owned, confirm BYOR and its separate SKU; do not add that fee to standard Model Import.
- OCI Generative AI → Dimension and costs → Economic viability
Sources
What are the exact model, version, license, source, format and size of the model weights to be imported?
Check that the artifact can be housed in the chosen modality.
Validate documented compatibility and use permissions To import doesn't mean that any architecture or quantification will be admitted.
- OCI Generative AI → Model accommodation → Compatibility and integration
Sources
The imported model's already adjusted, its full weights are available and who did and valid that tuning?
Distinguish external titration of service capabilities.
The service does not directly suit imported models. For an externally adjusted model, check the complete artifact compatible; an insulated adapter isn't good enough.
- OCI Generative AI → Model accommodation → Monitoring model and its operation
Sources
What shape, GPU, replicas, monthly hosting hours, context and regional availability does the imported model need under the measured workload?
Link the model and its test with a concrete configuration.
Validate model and service capacity requirements Do not exchange the clilling unit with an GPU Compute without checking the profile.
- OCI Generative AI → Dimension and costs → Performance and capacity
Sources
What data can be entered into the model, what retention and traceability are allowed and who controls models, budgets and evaluations?
Define AI's government and use limits.
To separate policies of access, data management and model appraisal assign components and liability to additional controls.
- OCI Generative AI → Government of consumption of AI → Control and traceability
Sources
Do we require continuity between OCI regions or selection between different models and providers and with what criteria?
To clear up what decision the routing should make.
A OCI file routing for an admitted model does not amount to a multimodel or multisupplier router. design and test that component independently as necessary.
- OCI Generative AI → Government of consumption of AI → Continuity
Sources
Which prompts, API contracts, tools, context, embeddings and evaluations of the current model must preserve their behavior?
To identify the change of supplier, model or modality.
To vertir the evaluation sample, compare quality and latency and anticipate rollback. changing embeddings may require re-dexing the knowledge base.
- OCI Generative AI → Migration and modernization → Transition from today's state