Brolly Academy | Snowflake
Snowflake Openflow
What It Is, Architecture, Connectors, Setup and Cost (2026)
A plain-English guide to Snowflake Openflow for October 2026: what it is, why Snowflake built it on Apache NiFi, how its architecture and two deployment types work, the official connector list, how it compares with Snowpipe, COPY INTO and third-party ETL tools, the setup flow, how you are charged, and its limits.
Snowflake Openflow explained: what it is, its Apache NiFi basis, architecture, Snowflake vs BYOC deployment, connectors, Openflow vs Snowpipe and cost.
By Brolly Academy Team · Brolly Academy · Updated 24 October 2026

Table of Contents
1. Snowflake Openflow at a glance
2. What is Snowflake Openflow?
4. Openflow architecture: the building blocks
5. Deployment options: Snowflake Deployment vs BYOC
6. Openflow connectors: the official list
7. Openflow vs Snowpipe vs COPY INTO vs third-party ETL tools
8. How to set up Openflow: a high-level flow
9. Openflow use cases, with Indian examples
11. Openflow pricing: how you are charged
12. Limitations and things to check
Quick answer
Snowflake Openflow is Snowflake's managed data integration service, built on Apache NiFi, that moves data from almost any source (databases, SaaS apps, file stores, event streams) into Snowflake, and in some cases back out again. Snowflake's documentation marks Openflow as generally available (GA) in October 2026. You can run it inside Snowflake on Snowpark Container Services (a "Snowflake Deployment") in AWS, Azure and Google Cloud commercial regions, or in your own AWS account (a "BYOC" deployment). You use ready-made connectors, such as MySQL, PostgreSQL, Oracle and SQL Server change data capture, Kafka, SharePoint, Google Drive and Salesforce, or you build your own flows from NiFi processors.
- What it is: a managed integration (extract and load) service based on Apache NiFi
- Status: GA for Snowflake Deployments on AWS, Azure and GCP commercial regions, and for BYOC on AWS commercial regions (Snowflake documentation, October 2026)
- Connectors: about 26 official connectors listed, from database CDC to ads platforms, Slack and Workday
- Cost: billed through compute (compute pools or vCPUs), plus ingestion, warehouse and telemetry charges; it is not free
- Trial accounts: not switched on automatically; you contact your Snowflake account team
Key takeaways
- Openflow is the "get data into Snowflake" layer: it extracts and loads, then you transform inside Snowflake with SQL, dynamic tables or dbt.
- Openflow is Apache NiFi under the hood, so flows are built from processors and controller services on a visual canvas, or added as ready-made connectors.
- There are two deployment types: a Snowflake Deployment runs on Snowpark Container Services inside Snowflake, and BYOC runs the data plane in your own AWS cloud while Snowflake runs the control plane.
- Every new deployment is "gen 2", which lets you manage deployments, runtimes and connectors with SQL; gen 2 connector configuration is still in public preview.
- None of the 14 Snowflake job descriptions we read in October 2026 named Openflow, so learn it as an add-on to strong SQL, loading and pipeline basics, not as a replacement for them.
1. Snowflake Openflow at a glance
Short answer: Openflow is a NiFi-based integration service with two deployment types, a curated connector catalogue, SQL-based management in gen 2, and compute-based billing.
If Snowflake itself is new to you, read our guide on what Snowflake does first. This post assumes you know what a table, a warehouse and a load are.
2. What is Snowflake Openflow?
Short answer: Openflow is Snowflake's integration service that connects data sources to destinations, mainly Snowflake, using Apache NiFi flows that Snowflake hosts, manages and secures for you.
Snowflake's About Openflow page (October 2026) describes Openflow as an integration service that connects any data source and any destination with hundreds of processors. Those processors handle structured and unstructured data: text, images, audio, video and sensor data. The page also says Openflow is built on Apache NiFi and can run as a fully managed service in your own cloud.
What Apache NiFi is, in one minute
Apache NiFi is an open-source tool for moving data between systems. You design a data flow as a set of boxes on a canvas. Each box is a processor that does one job, such as "read from a database", "convert JSON", or "write to Snowflake". You join processors with arrows, and data flows through them. Controller services are shared settings that several processors reuse, for example a database connection pool.
Openflow takes this NiFi model and adds Snowflake sign-in, role-based access control (RBAC, where you grant privileges to roles and roles to users), managed infrastructure, curated connectors, and monitoring through Snowflake event tables and cost views.
Status in October 2026
The About Openflow page carries the label "Feature: Generally Available". It says Openflow Snowflake Deployments are available to all accounts in AWS, Azure and GCP commercial regions, and BYOC deployments are available to all accounts in AWS commercial regions (Snowflake documentation, October 2026). We do not give a GA date here because we could not confirm one on an official page.
One more status detail matters. Snowflake's gen 2 quickstart says gen 2 deployment and runtime SQL objects are GA, while gen 2 connector configuration is still in public preview (October 2026). So the platform is GA, but parts of the newest connector setup are preview.
Do not confuse Openflow "gen 2" with "Gen2 warehouses". Gen2 standard warehouses are a newer type of virtual warehouse. Openflow gen 2 is the second management model for Openflow itself. They are unrelated.
3. Why Openflow exists
Short answer: Teams were stitching together many separate tools to get data into Snowflake; Openflow puts extraction and loading for many source types into one managed service governed by Snowflake.
Think about a typical data team at a bank's global capability centre (GCC) in Hyderabad or a retail analytics team in Bengaluru. In one week they might need to:
- Copy changes from an Oracle or SQL Server core system every few minutes.
- Pull marketing numbers from Google Ads, LinkedIn Ads and Meta Ads.
- Stream click events from Kafka.
- Index policy PDFs from SharePoint so business users can ask questions over them.
Before Openflow, that often meant a CDC tool for databases, a SaaS ELT tool for marketing apps, a separate Kafka connector, custom Python scripts for files, and Snowpipe or COPY INTO to land files from a stage. Each tool had its own login, its own security model and its own bill.
Openflow tries to cover these patterns in one place. The About Openflow page names its benefits as: open and extensible (NiFi-based), a unified platform for bi-directional extraction and loading, enterprise-ready security and observability, ingestion of structured and unstructured data in batch and streaming modes, and near real-time unstructured ingestion from sources such as SharePoint and Google Drive so you can chat with that data using Snowflake Cortex.
Two terms come up often here:
- ELT (extract, load, transform): you extract data from the source, load it raw into Snowflake, and transform it later inside Snowflake. Openflow does the "E" and "L".
- CDC (change data capture): instead of copying a whole table every night, you capture only the rows that changed (inserts, updates, deletes) and apply them to the copy in Snowflake.
4. Openflow architecture: the building blocks
Short answer: Openflow has a control plane that manages and observes everything, deployments that hold your runtimes, runtimes that actually run the flows, and connectors or custom flows that run inside the runtimes.
Here are the components, as Snowflake's docs describe them (October 2026):
How data moves through Openflow
Use this numbered flow as a mental diagram (we will add an image later):
- An admin sets up roles and privileges in Snowflake.
- An engineer creates a deployment (Snowflake Deployment or BYOC).
- Inside the deployment, the engineer creates one or more runtimes, often one per team, project or environment (dev, test, prod).
- A connector (for example, PostgreSQL CDC) is added to a runtime, or a custom flow is built from processors.
- The connector reads from the source system, such as a database, a SaaS API, Kafka or SharePoint.
- The runtime writes data into Snowflake tables using standard Snowflake ingestion, such as Snowpipe or Snowpipe Streaming. Some database connectors can also write to Apache Iceberg tables.
- Data engineers transform the raw "bronze" data into cleaner "silver" and "gold" layers inside Snowflake.
- Analysts, dashboards and AI tools use the gold layer.
The control plane watches steps 2 to 6 the whole time. Logs and metrics can go to a Snowflake event table, so you can monitor flows with SQL.
A short word on Apache Iceberg: it is an open table format that stores data as Parquet files with metadata, so several engines (Snowflake, Spark, Trino and others) can read the same tables. The Openflow docs list "Iceberg table destinations" pages for the MySQL, Oracle, PostgreSQL, SQL Server and Salesforce Bulk API connectors.
5. Deployment options: Snowflake Deployment vs BYOC
Short answer: Choose a Snowflake Deployment if you want Snowflake to run everything inside your account; choose BYOC if data must be processed inside your own AWS cloud first.
Openflow - Snowflake Deployment
This option runs on Snowpark Container Services (SPCS), Snowflake's managed container platform. Containers run on compute pools, which are groups of virtual machine nodes that Snowflake manages. Snowflake says this option is easy to deploy and manage and integrates natively with Snowflake's security model for authentication, authorization and network security (Snowflake documentation, October 2026).
Snowflake lists use cases that suit Snowflake Deployments well:
- Landing full-fidelity raw data in the bronze layer.
- Enriching tables that already exist in Snowflake.
- Keeping ingest, processing and serving all inside Snowflake.
- Ingesting unstructured data for AI, for example to search it with Snowflake CoWork (formerly Snowflake Intelligence) alongside structured data.
- Reverse ETL: sending insights back to operational systems through APIs or messaging.
Openflow - Bring Your Own Cloud (BYOC)
With BYOC, the Openflow data processing engine (the data plane) runs in your own cloud environment, while Snowflake manages the Openflow service and control plane. Snowflake positions BYOC for connecting public and private systems securely and for preprocessing sensitive data locally, inside your organisation's cloud boundary. BYOC is GA in AWS commercial regions (Snowflake documentation, October 2026).
Side-by-side comparison
Gen 1 vs gen 2
According to Snowflake's "Openflow gen 1 and gen 2" documentation page and the gen 2 quickstart (October 2026):
- Gen 1: new gen 1 deployments can no longer be created. Existing gen 1 deployments keep working. Connectors are NiFi process groups installed from a catalogue and configured on the runtime canvas.
- Gen 2: all new deployments are gen 2. Deployments, runtimes and connectors are Snowflake objects you manage with SQL commands such as CREATE OPENFLOW DEPLOYMENT, CREATE OPENFLOW RUNTIME and CREATE OPENFLOW CONNECTOR. Gen 2 connectors have versioned configuration that you can commit and roll back, a setup wizard, and standard RBAC grants such as GRANT USAGE and GRANT OPERATE. Gen 2 objects are also supported by the Snowflake Terraform provider.
- Gen 2 runtimes can run both gen 1 and gen 2 connectors.
Each account can have up to three Snowflake Openflow deployments, and gen 1 and gen 2 share that limit (Snowflake documentation, October 2026).
6. Openflow connectors: the official list
Short answer: Snowflake's connector page lists about 26 connectors in October 2026, covering database CDC, streaming, SaaS apps, file and content stores, and ads platforms.
Snowflake's Openflow connectors page describes connectors as curated, versioned Apache NiFi flow definitions, built from open-source and proprietary NiFi components and following strict design patterns for performance and fault tolerance. Here is the list, grouped by type (October 2026):
A few points worth knowing:
- Box: besides ingestion, the page says you can use Box AI to extract metadata and also push enriched metadata from Snowflake back to Box content.
- SQL Server: there are two connectors. Change Tracking (CT) replicates table changes. CDC captures every row-level change, which suits audit and history use cases.
- Snowflake to Kafka: this one runs the other direction. It replicates Snowflake table changes into Apache Kafka for event-driven systems.
- Oracle: the Oracle connector has a page titled "Enable and manage commercial terms", so check those terms before you plan on it.
- Need something not listed? You can build a custom flow from NiFi processors and controller services. Snowflake also publishes guidelines for Python extensions in Openflow.
The connector list changes over time, so always check the live page before you design a pipeline.
7. Openflow vs Snowpipe vs COPY INTO vs third-party ETL tools
Short answer: Snowpipe and COPY INTO load files that are already in a stage; Snowpipe Streaming loads rows sent by your own code; Openflow goes out to the source system, extracts the data and loads it for you.
A stage is a location that holds files for loading, either inside Snowflake (internal stage) or in your own S3, Azure or GCS storage (external stage). Our stages in Snowflake guide covers stages in depth.
- COPY INTO is a SQL command that bulk-loads staged files into a table. See our COPY INTO tutorial.
- Snowpipe is Snowflake's serverless service that loads files automatically as they land in cloud storage. See our Snowpipe guide.
- Snowpipe Streaming loads rows directly into tables without staging files. Snowflake's Snowpipe Streaming page lists Java, Python and Node.js SDKs, a REST API and the Kafka connector (October 2026).
Openflow sits one step earlier. It talks to the source (a database log, a SaaS API, SharePoint), pulls data out, and then uses Snowflake ingestion such as Snowpipe or Snowpipe Streaming to land it. The BYOC cost page says exactly this: ingestion through services such as Snowpipe or Snowpipe Streaming appears on your bill under those services.
What about Fivetran, Informatica, Matillion, Airflow and ADF?
Tools such as Fivetran, Informatica, Matillion and Airbyte are commonly used with Snowflake for ingestion and ETL. Several Indian job descriptions we read in October 2026 name them: one Capgemini listing asked for ELT ingestion with Fivetran, Airbyte or Snowpipe, and a TCS listing named Informatica Cloud (IICS) alongside COPY and Snowpipe. We are not comparing their features or prices here, because we have not verified those on official pages. A fair way to decide:
- If your company already pays for and runs one of these tools well, Openflow is an option, not a must.
- If you need CDC plus unstructured content plus streaming, and want it all governed by Snowflake roles and billed on your Snowflake account, Openflow is worth a pilot.
Apache Airflow and Azure Data Factory (ADF) are mostly orchestration tools: they schedule and coordinate jobs. Openflow is mainly about moving data. Many teams use an orchestrator (or Snowflake tasks) to run transformations after Openflow lands raw data. Inside Snowflake, streams (which track row changes in a table) and tasks (which run SQL on a schedule) often handle that next step; see our streams and tasks lab.
For a wider map of the ecosystem, read our Snowflake tools guide.
8. How to set up Openflow: a high-level flow
Short answer: Set up an admin role and privileges, create a deployment, create an execute-as role and network access, create a runtime, allow the source domains, then add a connector.
Snowflake's setup page for Snowflake Deployments lists these tasks in order (October 2026). The persona column shows who usually does each step.
The execute-as role is the Snowflake role your connectors use to write data. Give it access only to the destination databases it needs. An external access integration is a Snowflake object that allows outbound network calls to specific domains, so a runtime can reach, say, your PostgreSQL host or the Salesforce API.
Step 3 is usually done once. Steps 4 and 5 repeat for each runtime you add.
Simplified example: privileges and a Snowflake Deployment
This simplified example follows the gen 2 quickstart (Snowflake documentation, October 2026). Run it only in an account where Openflow is enabled.
-- Simplified example (gen 2 Openflow)
USE ROLE ACCOUNTADMIN;
CREATE ROLE IF NOT EXISTS openflow_admin;
GRANT ROLE openflow_admin TO ROLE ACCOUNTADMIN;
GRANT CREATE OPENFLOW DEPLOYMENT ON ACCOUNT TO ROLE openflow_admin;
GRANT CREATE DATABASE ON ACCOUNT TO ROLE openflow_admin;
GRANT CREATE INTEGRATION ON ACCOUNT TO ROLE openflow_admin;
GRANT CREATE COMPUTE POOL ON ACCOUNT TO ROLE openflow_admin;
USE ROLE openflow_admin;
CREATE DATABASE IF NOT EXISTS openflow_db;
CREATE SCHEMA IF NOT EXISTS openflow_db.openflow_schema;
CREATE OPENFLOW DEPLOYMENT my_snowflake_deployment
DEPLOYMENT_TYPE = SNOWFLAKE
DISPLAY_NAME = 'My gen 2 Snowflake Deployment';
A few notes on this example:
- openflow_db.openflow_schema is the control schema that holds runtimes and connectors. It is not where your data lands.
- The CREATE COMPUTE POOL grant is needed only for Snowflake Deployments.
- PrivateLink settings must be chosen when you create the deployment. The quickstart says you cannot change them later with ALTER OPENFLOW DEPLOYMENT.
- For BYOC, you set DEPLOYMENT_TYPE = BYOC, accept the Openflow BYOC terms, and complete the AWS infrastructure steps. If you need AWS PrivateLink, set it up before you apply the CloudFormation template.
A first project idea: PostgreSQL CDC into Snowflake
If your company enables Openflow for you, a good first flow is:
- Pick a small PostgreSQL test database with two or three tables.
- Create a runtime in your deployment.
- Allow the PostgreSQL host in the runtime's network settings.
- Add the PostgreSQL connector and point it at a destination database.
- Let it take the initial snapshot, then insert, update and delete a few rows.
- Check that the changes appear in Snowflake.
- Write a dynamic table or view on top for a clean "silver" version.
Note that database CDC connectors need a standard Snowflake warehouse for the initial snapshot and for incremental CDC, per the BYOC cost page (October 2026). Use an X-Small warehouse with auto-suspend for practice.
9. Openflow use cases, with Indian examples
Short answer: Snowflake lists five main use cases: unstructured content for AI, database CDC, real-time streaming, SaaS ingestion and custom NiFi flows.
The About Openflow page lists these use cases (October 2026). We have added typical Indian settings in brackets. These are illustrations, not named customers.
- Unstructured content for AI: ingest files from Google Drive, Box or SharePoint and make them ready for chat with Snowflake Cortex. (An insurance GCC in Hyderabad indexes policy documents from SharePoint so support teams can ask questions.) Cortex is Snowflake's set of AI services; our Cortex AI tutorial explains the SQL functions.
- Database CDC: replicate changes from database tables into Snowflake for central reporting. (A bank's Pune centre replicates an Oracle or SQL Server system into Snowflake every few minutes instead of a nightly batch.)
- Real-time events: ingest events from Apache Kafka for near real-time analytics. (A food-delivery or e-commerce team in Bengaluru streams order events.)
- SaaS ingestion: pull data from platforms such as LinkedIn Ads for reporting. (A digital marketing agency in Chennai combines Google Ads, Meta Ads and LinkedIn Ads data.)
- Custom flows: build your own dataflow from Snowflake and NiFi processors and controller services, for a source that has no ready connector.
Snowflake also names enrichment of existing Snowflake tables and reverse ETL (sending results back to operational systems) as good fits for Snowflake Deployments.
10. Security and governance
Short answer: Openflow uses Snowflake roles, short-lived managed tokens, TLS encryption, Snowflake secrets and optional private connectivity.
From the About Openflow page (October 2026):
- Authentication: runtimes use Snowflake Managed Token by default. Tokens are short-lived and refresh automatically, so you do not store or rotate long-lived keys. In Snowflake Deployments the runtime uses SPCS session tokens. In BYOC it uses workload identity federation, exchanging its cloud identity (for example an AWS IAM role) for a Snowflake token. BYOC can also use key-pair authentication.
- Authorization: fine-grained roles for RBAC. ACCOUNTADMIN grants the privileges needed to create deployments and runtimes.
- Encryption in transit: connectors use TLS, and all traffic between deployments and the control plane is encrypted with TLS.
- Secrets: integration with Snowflake secrets in both deployment types; BYOC can also use AWS Secrets Manager or HashiCorp Vault.
- Private connectivity: inbound AWS PrivateLink (BYOC) or outbound private connectivity (Snowflake Deployments).
- Tri-Secret Secure: connectors are compatible with Tri-Secret Secure, the customer-managed key feature, when writing to Snowflake.
A practical tip: give each runtime its own execute-as role with access only to its target schema, so a misconfigured flow cannot write elsewhere.
11. Openflow pricing: how you are charged
Short answer: Openflow is not free; you pay for compute (compute pool credits for Snowflake Deployments, vCPU hours for BYOC) plus ingestion, any warehouse used, and telemetry, at rates listed in Snowflake's Service Consumption Table.
Snowflake does not publish one flat Openflow price. Instead, its cost pages explain the categories. We quote no per-credit or per-vCPU rates here, because rates vary by region and edition and sit in Snowflake's Service Consumption Table.
Snowflake Deployment costs
Snowflake's Snowflake Deployment cost page (October 2026) lists compute pool costs, SPCS infrastructure, data ingestion, telemetry data ingestion and other costs. The key points:
- Compute shows on your bill as Openflow Compute Snowflake.
- A management compute pool (its name ends in _CONTROL_POOL) starts when you create a deployment and keeps running, and costing credits, while the deployment is active, even with no runtimes running.
- Runtime compute pools scale with the runtimes you run. When all runtimes on a pool stop, the pool scales to 0 nodes and costs nothing.
- Credits are billed per second with a 5-minute minimum.
- Per-runtime cost attribution is not available, because runtimes of the same size share a compute pool.
A credit is Snowflake's unit of compute usage; your bill is credits used multiplied by your price per credit.
BYOC costs
The BYOC cost page (October 2026) lists:
- Openflow Compute BYOC: based on vCPUs used by active connector runtimes, billed per second with a 60-second minimum. Management processes are excluded from this charge.
- Infrastructure: you pay AWS directly for EC2 compute, networking and storage in your account.
- Ingestion: Snowpipe or Snowpipe Streaming charges by data volume; CDC connectors also need a warehouse for snapshots and incremental CDC.
- Telemetry: charges for sending logs and metrics into Snowflake.
The page maps runtime sizes to vCPUs: Small 1 vCPU, Medium 4 vCPUs, Large 8 vCPUs. For example, one Small runtime running one node for one hour uses 1 vCPU-hour.
Track Openflow spend with SQL
This simplified example is adapted from the Snowflake Deployment cost page. It shows daily Openflow compute credits for the last 30 days.
-- Simplified example: daily Openflow compute credits
SELECT TO_DATE(usage_date) AS date,
service_type,
SUM(credits_used) AS credits_used
FROM snowflake.account_usage.metering_daily_history
WHERE service_type = 'OPENFLOW_COMPUTE_SNOWFLAKE'
AND usage_date >= DATEADD(month, -1, CURRENT_TIMESTAMP())
GROUP BY 1, 2
ORDER BY 1 DESC;
For BYOC, filter on OPENFLOW_COMPUTE_BYOC instead.
Cost-control checklist
- Delete practice deployments you no longer use, since the management pool runs while a deployment is active.
- Stop runtimes outside working hours; gen 2 lets you script this with SQL.
- Start with Small runtimes and scale only when you see backlogs.
- Schedule MERGE operations for CDC connectors to control warehouse cost, as the BYOC cost page suggests.
- Use an X-Small warehouse with auto-suspend for snapshot and CDC work during testing.
12. Limitations and things to check
Short answer: Check region and trial access, the three-deployment limit, preview status of gen 2 connector setup, login rules, account suspension risk and edition needs for PrivateLink.
From Snowflake's documentation (October 2026):
- Trial accounts: Openflow is not automatically available in trial accounts; you must contact your Snowflake account team.
- China: Openflow is not available in the People's Republic of China.
- Clouds: BYOC is AWS only. Snowflake Deployments cover AWS, Azure and GCP commercial regions. Snowflake's India regions are AWS Mumbai (ap-south-1) and Azure Central India (Pune); confirm Openflow support for your account's region on the docs before you plan a project.
- Deployment limit: up to three Snowflake Openflow deployments per account.
- Preview parts: gen 2 connector configuration is public preview.
- Login rule: users whose default role is ACCOUNTADMIN cannot log in to Snowflake Deployment runtimes.
- Account suspension: if the account is suspended (for billing, trial expiry or other reasons), existing Snowflake Deployments show as "Not Reporting" and cannot be recovered, even after reactivation.
- PrivateLink: outbound PrivateLink for Snowflake Deployments needs Business Critical edition.
- BYOC specifics: shared security responsibility, images only from the Snowflake system image registry, and authorization only through roles granted directly to the user (not inherited through the role hierarchy).
When Openflow may not be the right choice
- You only load files that already land in S3 or Azure: Snowpipe or COPY INTO is simpler.
- Your producer is your own application: Snowpipe Streaming may be more direct.
- You only need a couple of small, nightly CSV loads: a task plus COPY INTO is cheaper to run.
13. Skills to learn and the career angle
Short answer: Openflow is a useful add-on skill, but Indian Snowflake JDs in October 2026 still ask mostly for SQL, loading, dbt, Python and cloud basics.
We read 14 Snowflake job descriptions in India on 25 October 2026. SQL appeared in all of them, dbt in 9 and Python in 9. Named ingestion and integration tools included Snowpipe, Fivetran, Airbyte, Informatica IICS, Talend and Matillion. None of the 14 named Openflow. So Openflow is not yet a common hiring filter in this sample, but it sits right next to skills that are.
To be ready for Openflow work, learn in this order:
- SQL and data modelling: joins, window functions, star schema, slowly changing dimensions.
- Snowflake loading basics: stages, file formats, COPY INTO and Snowpipe.
- CDC concepts: snapshots, inserts, updates, deletes, MERGE, and Snowflake streams and tasks.
- Roles and security: RBAC, least privilege, secrets and network rules.
- NiFi basics: processors, controller services, connections and back-pressure (how a flow slows down when the next step is full).
- One cloud: AWS is the safest pick because BYOC runs on AWS.
- Transformation: dynamic tables or dbt for the silver and gold layers.
Informatica or Talend developers bring a useful flow-design mindset, DBAs bring replication knowledge that helps with CDC connectors, and B.Tech, BCA, B.Sc or MCA freshers should master steps 1 to 4 first. Our Snowflake data engineer guide shows where ingestion fits in the full role.
14. How Brolly Academy helps
Short answer: Brolly Academy's 45-day Snowflake course builds the loading and pipeline foundations that Openflow sits on; it does not include an Openflow module.
Our Snowflake training in Hyderabad runs for 45 days and comes in three formats: Video Recording at ₹8,999, live Online at ₹17,999 and Classroom at ₹22,999 (course page, October 2026). The trainer, Mr. Bala Krishna, has 13+ years in data warehousing, SQL and Snowflake.
The 15 modules cover the skills you need before and around Openflow:
- Data loading and unloading: file formats, internal and external stages, COPY INTO, and loading from AWS, Azure and Google Cloud.
- Snowpipe: PUT and GET, bulk loading and continuous loading.
- Scheduling with tasks and streams, including calling procedures from tasks.
- Partner Connect, the Python connector and BI connectors.
- Security basics, secure data sharing, Time Travel, cloning and performance.
To be clear, Openflow, Snowpark, Dynamic Tables and Cortex are not syllabus modules. You also build 6 projects, including an ETL pipeline project, and receive a Brolly Academy course completion certificate. SnowPro exam fees are separate and paid to Snowflake.
Build the Snowflake loading skills Openflow depends on
Openflow lands data, but you still need to know stages, COPY INTO, Snowpipe, streams, tasks, roles and cost control to run real pipelines. Book a free demo class of Brolly Academy's 45-day Snowflake course in Hyderabad (classroom, live online or video) and practise these foundations on hands-on projects with a working trainer.
Frequently asked questions
What is Snowflake Openflow?
Snowflake Openflow is Snowflake's managed data integration service, built on Apache NiFi. It extracts data from sources such as databases, SaaS apps, Kafka and file stores like SharePoint, and loads it into Snowflake. You can use ready-made connectors or build custom flows from NiFi processors. It runs either inside Snowflake on Snowpark Container Services or in your own AWS cloud through a BYOC deployment.
Is Snowflake Openflow based on Apache NiFi?
Yes. Snowflake's documentation says Openflow is built on Apache NiFi, the open-source data flow tool. Openflow connectors are curated, versioned NiFi flow definitions made from open-source and proprietary NiFi components. You can also build your own flows from NiFi processors and controller services on the runtime canvas, while Snowflake adds managed infrastructure, Snowflake sign-in, roles and monitoring.
Is Snowflake Openflow an ETL tool?
Openflow is best described as an extract and load tool in an ELT pattern. It pulls data from sources and loads it into Snowflake, often into a raw bronze layer. Heavy transformation usually happens afterwards inside Snowflake with SQL, dynamic tables or dbt. NiFi processors can do light conversions during the flow, but Openflow is not mainly a transformation engine.
Is Snowflake Openflow generally available?
Yes. In October 2026, Snowflake's documentation marks Openflow as generally available. Snowflake Deployments are available in AWS, Azure and Google Cloud commercial regions, and BYOC deployments in AWS commercial regions. Gen 2 deployment and runtime SQL objects are GA, but gen 2 connector configuration is still in public preview, so check the status of the exact feature you plan to use.
Is Snowflake Openflow free?
No. Openflow has no free tier described in Snowflake's documentation. Snowflake Deployments are billed as compute pool credits, and the management pool keeps running while a deployment is active. BYOC is billed by vCPU use of active runtimes, plus your own AWS infrastructure bill. You also pay for ingestion, any warehouse used by CDC connectors, and telemetry.
How much does Snowflake Openflow cost?
There is no single flat price. Snowflake Deployments are billed per second in credits, with a 5-minute minimum, based on compute pool usage. BYOC is billed per second on vCPUs of active runtimes, with a 60-second minimum, plus AWS costs. Exact rates sit in Snowflake's Service Consumption Table and vary by region and edition, so check your own contract and region.
What connectors does Openflow support?
Snowflake listed about 26 connectors in October 2026. They include MySQL, PostgreSQL, Oracle, SQL Server and MongoDB CDC, Google BigQuery, Kafka, Kinesis, Snowflake to Kafka, SharePoint, Google Drive, Box, Google Sheets, Salesforce Bulk API, HubSpot, Microsoft Dataverse, Workday, Veeva Vault, Jira Cloud, Slack, Shopify, and Google, LinkedIn, Meta and Amazon Ads. The list changes, so check the live page.
Can Openflow handle CDC from databases?
Yes. Openflow has change data capture connectors for MySQL, PostgreSQL, Oracle, MongoDB and SQL Server, where SQL Server has both a Change Tracking and a CDC connector. They take an initial snapshot and then apply ongoing inserts, updates and deletes. Database CDC connectors need a standard Snowflake warehouse for the snapshot and incremental CDC, which adds warehouse cost.
What is the difference between Openflow and Snowpipe?
Snowpipe automatically loads files that are already sitting in cloud storage or a stage. Openflow goes out to the source system, such as a database, SaaS API or SharePoint, extracts the data and then loads it into Snowflake, using services such as Snowpipe or Snowpipe Streaming under the hood. If your files already land in S3 or Azure, Snowpipe alone is often simpler.
Openflow vs Fivetran: which is better?
Neither is better for every team. Fivetran is a third-party ELT tool commonly used with Snowflake, while Openflow is Snowflake's own NiFi-based service, governed by Snowflake roles and billed on your Snowflake account. If your company already runs Fivetran well, keep it. If you need CDC, streaming and unstructured content under one Snowflake-managed service, pilot Openflow and compare.
How is Openflow different from Airflow or Azure Data Factory?
Airflow and Azure Data Factory are mostly orchestration tools that schedule and coordinate jobs across systems. Openflow is mainly a data movement service that extracts from sources and loads into Snowflake. Many teams use both: Openflow lands raw data, and an orchestrator or Snowflake tasks run the transformations that follow. They solve different parts of a pipeline.
What is BYOC deployment in Openflow?
BYOC means Bring Your Own Cloud. The Openflow data plane, the engine that processes data, runs in your own AWS account, while Snowflake manages the Openflow service and control plane. It suits teams that must preprocess sensitive data inside their own cloud boundary. You pay Snowflake for vCPU use of active runtimes and pay AWS for the infrastructure.
What is an Openflow Snowflake Deployment?
An Openflow Snowflake Deployment runs entirely inside Snowflake on Snowpark Container Services, using compute pools that Snowflake manages. It integrates natively with Snowflake's security model and is simpler to set up than BYOC. It is available in AWS, Azure and Google Cloud commercial regions. Each account can have up to three Snowflake Openflow deployments.
Does Openflow work on Azure?
Yes, through Snowflake Deployments. In October 2026, Openflow Snowflake Deployments are available to all accounts in AWS, Azure and Google Cloud commercial regions. BYOC deployments are available only in AWS commercial regions. So if your Snowflake account runs on Azure, you would use the Snowflake Deployment type rather than BYOC.
Can I use Openflow in Snowflake India regions?
Snowflake's India regions are AWS Mumbai and Azure Central India in Pune. The Openflow documentation says Snowflake Deployments are available in AWS, Azure and Google Cloud commercial regions, and BYOC in AWS commercial regions. Before planning a project, confirm Openflow support for your own account's region on the Snowflake documentation or with your account team.
How do I set up Openflow in Snowflake?
Set up an Openflow admin role and privileges, then create a deployment with CREATE OPENFLOW DEPLOYMENT or the Openflow UI. Next, create an execute-as role and external access integrations, create a runtime, allow the external domains your connectors need, and finally add a connector with SQL or the setup wizard. Follow Snowflake's gen 2 quickstart for exact steps.
Can I use Openflow in a Snowflake free trial account?
Not by default. Snowflake's documentation says Openflow is not automatically available in trial accounts, and you need to contact your Snowflake account team to enable it. For practice, learn the surrounding skills in a trial, such as stages, COPY INTO, Snowpipe, streams and tasks, and study NiFi concepts separately until you get access.
What is the difference between Openflow gen 1 and gen 2?
All new Openflow deployments are gen 2, and new gen 1 deployments can no longer be created. Gen 2 makes deployments, runtimes and connectors Snowflake objects you manage with SQL, adds versioned connector configuration, a setup wizard and standard grants. Existing gen 1 deployments keep working. Gen 2 connector configuration is still in public preview.
What is an Openflow runtime?
A runtime is the engine that actually runs your data flows. It lives inside a deployment, and you deploy connectors or build custom flows on it. Teams often create several runtimes in one deployment to separate projects, teams or dev, test and production work. In BYOC, runtime sizes map to vCPUs: Small 1, Medium 4 and Large 8.
Can Openflow load unstructured data like documents from SharePoint?
Yes. Openflow handles structured and unstructured data, including text, images, audio, video and sensor data. Snowflake's SharePoint, Google Drive and Box connectors can ingest content and make it ready for chat with Snowflake Cortex, or for your own custom processing. This is one of the main use cases Snowflake lists for Openflow.
Can Openflow send data out of Snowflake?
Yes, in some cases. Snowflake describes Openflow as bi-directional, and lists reverse ETL, sending insights back to operational systems, as a good fit for Snowflake Deployments. The Snowflake to Kafka connector replicates Snowflake table changes into Kafka, and the Box connector can push enriched metadata from Snowflake back to Box content.
Can Openflow write to Apache Iceberg tables?
Yes, for several connectors. Snowflake's Openflow documentation has Iceberg table destination pages for the MySQL, Oracle, PostgreSQL, SQL Server and Salesforce Bulk API connectors. Apache Iceberg is an open table format that lets several engines read the same tables. Check each connector's page for the exact setup and any limits before you choose Iceberg as a target.
What skills do I need to work with Openflow?
Start with strong SQL and data modelling, then Snowflake loading basics such as stages, COPY INTO and Snowpipe. Add CDC concepts, MERGE, streams and tasks, plus RBAC and secrets. Then learn NiFi basics such as processors and controller services, one cloud (AWS fits BYOC), and a transformation tool such as dynamic tables or dbt.
Do Snowflake job descriptions in India ask for Openflow?
Not often yet. In the 14 Indian Snowflake job descriptions we read on 25 October 2026, none named Openflow. SQL appeared in all 14, and dbt and Python in 9 each. Ingestion tools named included Snowpipe, Fivetran, Airbyte, Informatica IICS, Talend and Matillion. Treat Openflow as a useful extra on top of these core skills.
Does Brolly Academy's Snowflake course teach Openflow?
No. Brolly Academy's 45-day Snowflake course does not have an Openflow module. It covers the foundations Openflow depends on: stages, file formats, COPY INTO, Snowpipe, tasks and streams, Partner Connect, security, data sharing and performance, with 6 projects. Once you know these, learning Openflow from Snowflake's documentation becomes much easier.
Learn with Brolly Academy
Review the syllabus, practical projects and batch timings, and book a free demo before you enrol.
Related Articles
Continue Your Learning
Snowflake Training in Hyderabad
45-day course covering Snowflake architecture, warehouses, data loading, security, performance and real-time projects, in classroom, live online and recorded plans.
View Coursedbt Course
Learn dbt, the transformation tool most Snowflake job descriptions ask for, to build tested, documented data models.
View CourseDatabricks Course
Add the lakehouse side of modern data engineering with Databricks, Spark and Delta Lake, often used alongside Snowflake.
View Course