# Best Data Transfer Applications for Cloud Databases Compared > Compare AWS DMS, Datastream, Azure DMS, Erathos, Fivetran, Airbyte, Estuary, Matillion and dlt on pricing, CDC and hosting in 2026. Source: https://www.erathos.com/en/blog/best-data-transfer-applications-cloud-databases Em português: https://www.erathos.com/blog/best-data-transfer-applications-cloud-databases Published: 2026-10-26 Category: Tool Guides ![Best Data Transfer Applications for Cloud Databases Compared](https://cms-media.erathos.com/Best Data Transfer Applications for Cloud Databases Compared.png) Which data transfer application is best depends on what you need it for. If you're moving a database once inside AWS or Azure, the cloud's own migration service (AWS DMS or Azure DMS) is the cheapest option. If you need to keep a warehouse in sync with databases and SaaS apps, an ELT platform like Erathos, Fivetran or Airbyte is a better fit. In this article I compare the main tools for both jobs on how they charge and how they find changes, and note where each one runs. I took the 2026 prices from each vendor's pricing page. ## Do you need a one-time migration or continuous replication? A one-time migration moves a database to a new place once. Then you switch over and turn the tool off. Continuous replication keeps copying new changes from a live database into a warehouse every day, for as long as your business runs. On day one, both jobs look the same, because both start by copying the whole source. After that they're different. A migration only has to keep the target in sync long enough for you to switch over with little downtime. Replication has to stay correct for years, even as schemas change, rows get deleted and new tables get added. Here is how the two jobs map to tools: Job What it does Tool category Examples One-time migration Copies a database once, sometimes with a short sync window before cutover Cloud migration services AWS DMS, Azure DMS Continuous replication Keeps a warehouse up to date with a database or SaaS app, on a schedule or in near real time ELT (extract, load, transform) and CDC platforms Erathos, Fivetran, Airbyte, Estuary, Matillion, dlt The cloud providers make the same split in their pricing. Azure's Standard tier [supports only offline, "one-time" migrations](https://azure.microsoft.com/en-us/pricing/details/database-migration/), and the Premium tier adds online, "continuous" migration. For replication, you also need to know how the tool finds changes. A cursor-based sync asks the source for rows with a newer update timestamp. That's good for small tables with a reliable update column. But it misses deleted rows, and any state a row was in between two runs. Change data capture (CDC) reads the database's transaction log instead, so it sees every insert, update and delete. In Postgres that log is the WAL (write-ahead log), and the tool reads it through a logical replication slot. So CDC is about [capturing every change](https://www.erathos.com/en/blog/cursor-based-sync-vs-change-data-capture). Speed is a separate choice, and a CDC pipeline can still send data to the warehouse in hourly batches. Here is the path a change takes from Postgres to the warehouse: **\[IMAGE 1: upload gtm-1360-img1.png here\]** ## How do AWS DMS, Google Datastream and Azure DMS compare? AWS DMS, Google Datastream and Azure DMS are the cheapest way to move a database inside one cloud, and all three can stream changes. But they charge in very different ways. AWS charges by instance hour or capacity unit, Google by how much data it processes, and Azure by vCore hour with long free periods. Service How it charges Continuous sync Free allowance Best for AWS DMS On-demand instances, serverless capacity units (DCUs), or Savings Plans Yes, CDC Data transfer into DMS and schema conversion are free Moving databases into AWS Google Datastream Per GiB processed, from $2.00 down to $0.80 per GiB Yes, it is a CDC service First 500 GiB of backfill each month Streaming database changes into BigQuery Azure DMS Standard tier free; Premium 4 vCore at $0.37 per hour Premium tier only Premium is free for its first 6 months Moving databases into Azure **AWS DMS** has [no minimum fees and no upfront commitment](https://aws.amazon.com/dms/pricing/) on its on-demand and serverless options, so you pay only for what you use. Serverless measures capacity in DCUs (one DCU is 2 GB of RAM) and scales in steps from 1 to 384 DCUs. The service also supports continuous replication with CDC, so it can keep running after the migration ends. AWS lists hourly instance prices by region, so the cost depends on which region you pick. **Google Datastream** charges [per GiB of data processed](https://cloud.google.com/datastream/pricing). The price drops in tiers, from $2.00 per GiB down to $0.80 per GiB once you go above 10,000 GiB a month. But Google counts bytes in an unusual way. A "processed byte" is an internal format that is often 2 to 5 times larger than the source data. Google's own example, with a 600 GiB backfill and 3,000 GiB of CDC into BigQuery, comes to $5,790 a month before BigQuery charges. Datastream also only delivers the change stream, so you still have to build the processing that turns it into clean tables. **Azure DMS** is the cheapest of the three for a project. The Standard tier is free and handles offline migrations. The Premium tier adds online migration at [$0.37 per hour for 4 vCores](https://azure.microsoft.com/en-us/pricing/details/database-migration/), free for the first 183 days. Each service instance can run for up to one year from creation, which shows Microsoft built it for projects that have an end date. Here are both Azure tiers: **\[IMAGE 2: upload gtm-1360-img2.png here\]** All three are built to move databases into their own cloud. If your sources are SaaS apps like ad platforms, CRMs or ERPs, or your warehouse is in another cloud, an ELT platform is a better fit. ## How do Erathos, Fivetran, Airbyte, Estuary, Matillion and dlt compare? The six platforms differ mostly in what they bill for and who runs them. Erathos bills for rows written to the warehouse, Fivetran for rows changed per connection, Airbyte and Estuary for credits or gigabytes, Matillion for task hours, and dlt is free code you run yourself. Platform How it charges Entry price CDC Hosting Connectors Erathos Flat tiers by rows written to the warehouse each month Free up to 1M rows; $29 for 2M; $250 for 5M Postgres, on the Pro plan Managed only 100+ Fivetran Monthly active rows (MAR) per connection, $5 minimum per connection Free up to 500,000 MAR Mature, including enterprise databases Managed; hybrid on Enterprise 700+ Airbyte Credits on Standard and Plus; capacity (Data Workers) on Pro 40 credits for $195 (Standard) or $189 (Plus) Postgres, MySQL, SQL Server, MongoDB; Oracle, SAP HANA, Db2 paid Cloud, self-hosted open source, or hybrid 700+ Estuary $0.50 per GB plus $100 per connector instance Free up to 10 GB a month Log-based, millisecond latency Managed; private or bring-your-own-cloud on Enterprise 200+ Matillion Credits per task hour No public dollar price Streaming CDC pipelines Varies by plan Not published as a single count dlt Free open-source Python library; dltHub license is paid Free Only in the paid dltHub license You run it Limited verified connectors **Erathos** has the most predictable bill on this list. It only counts rows written to your warehouse, and each tier on the [pricing page](https://www.erathos.com/en/pricing) has a fixed price. The $250 Pro plan adds CDC, custom backfill, custom connectors and syncs every 5 minutes. You can also set up each table as its own job, with its own schedule and retry rules. Their catalog is smaller than Fivetran's or Airbyte's, and Erathos is managed only. Erathos has three self-serve plans: **\[IMAGE 3: upload gtm-1360-img3.png here\]** **Fivetran** has the widest catalog and the most mature CDC for enterprise databases. In 2026 their pricing got harder to predict. Since January 1, 2026, [deletes count toward paid MAR](https://fivetran.com/docs/core-concepts/usage-based-pricing/pricing-updates/2026-pricing-updates), and every standard connection with 1 to 1M MAR costs at least $5. In the example on their [pricing page](https://www.fivetran.com/pricing), three marketing connectors on the Standard plan cost $511.15 a month. The Standard plan syncs every 15 minutes, and you need Enterprise for 1-minute syncs. **Airbyte** is a good choice if you want open source you can host yourself. Airbyte Core runs on Kubernetes. On Airbyte Cloud, [databases cost 4 credits per GB](https://www.erathos.com/en/blog/erathos-vs-airbyte-data-integration-platform-2026), measured after serialization, so you can get billed for more than the data on disk. The Standard plan syncs at most once an hour. For API rows, Airbyte is cheaper than Erathos on sticker price. 6.64M API rows cost $195 on Airbyte Standard, compared to $250 for 5M rows on Erathos Pro. **Estuary** is built for real-time CDC. It charges [$0.50 per GB plus $100 per connector instance](https://www.estuary.dev/pricing/), and their own calculator shows 800 GB with two connector instances costing $1,000 a month. It's a good fit if you need sub-second latency and can predict how many gigabytes you'll move. **Matillion** prices in credits. Orchestration pipelines use 1 credit per task hour, transformation pipelines use 3, and each developer user costs 70 credits a month. It's for teams that want ingestion and transformation in one product. There's no public dollar price, so you have to ask sales for a quote. **dlt** is a free, Apache 2.0 Python library. It's good for engineers who want to write pipelines as code and take care of running, testing and operating them. CDC and Apache Iceberg support are only in the paid [dltHub license](https://dlthub.com/pricing), which starts at about $12,000 a month with a 12-month minimum. ## When is Erathos the best pick? Erathos is a good choice for small or mid-size data teams that want a fixed monthly bill, Postgres CDC and control over each table. It's also the best option on this list for Brazilian SaaS sources and for teams that want to pay in BRL. Erathos fits best when these match your setup: - **You need to forecast the bill.** Each [plan](https://www.erathos.com/en/pricing) has a row limit and a fixed price, and Erathos counts only rows written to the warehouse. Adding a small source only adds the rows it writes. - **Your main source is Postgres.** The [Postgres connector](https://docs.erathos.com/connectors/databases/postgresql) reads the WAL for CDC, in both overwrite and append modes. - **You want to tune tables one by one.** Each table is its own job, so you can set its schedule, backfill window and retry rule separately. Alerts can go to email, Slack or Discord. - **You run Brazilian tools.** The catalog includes [Bling, Omie, RD Station and Conta Azul](https://www.erathos.com/en/vs/fivetran), and they offer support in Portuguese and billing in BRL. - **You orchestrate with code.** Airflow or Dagster can trigger syncs through the open API, and dbt can run right after ingestion. Among Erathos customers, UoouShop connected [6 data sources in a single day](https://www.erathos.com), and Kamino [cut its data costs by more than 90%](https://www.erathos.com/en/blog/erathos-vs-airbyte-data-integration-platform-2026). Other tools beat Erathos in a few cases: - **You need CDC from Oracle, SQL Server, SAP or Db2, or PCI DSS Level 1 (card data) and HIPAA (US health data) compliance.** Fivetran has deeper coverage here, plus 1-minute syncs on Enterprise. - **You need to self-host.** Erathos runs only as a managed service. Airbyte Core is open source and runs on your own Kubernetes cluster. - **You need millisecond latency.** Estuary is built for real-time streaming. - **You're only moving one database into AWS or Azure once.** AWS DMS or Azure DMS is cheaper for that. - **Snowflake is your warehouse.** Erathos still listed Snowflake as coming soon in September 2026, so it may not be on the [connector list](https://www.erathos.com/en/connectors) yet. ## How do you choose a data transfer application? I'd look at the job first, then at your sources, how fresh the data needs to be, and how the vendor counts volume. I'd look at price last, because each vendor measures something different. The first few decisions narrow the tool category: **\[IMAGE 4: upload gtm-1360-img4.png here\]** I'd go through these questions in order: 1. **What's the job?** A database moving once to a new cloud needs a migration service. A warehouse that has to stay in sync needs an ELT platform. 2. **Which sources do you have?** A tool with 700+ connectors can still miss your ERP, so it's worth checking each catalog for every source. Enterprise databases like Oracle and Db2 often need a paid tier. 3. **Do you need CDC?** You need it when tables have no reliable update column, or when deletes have to reach the warehouse. On Postgres, CDC needs some setup on the database: `wal_level` set to logical, free replication slots, and a limit on how much WAL a slot can hold. 4. **How fresh does the data need to be?** Faster syncs usually mean a higher plan. Erathos syncs hourly on Movement and every 5 minutes on Pro, Fivetran every 15 minutes on Standard and every minute on Enterprise, and Airbyte at most hourly on Standard. If you need millisecond latency, a streaming tool like Estuary is the better fit. 5. **How does the vendor count volume?** Erathos counts rows written, Fivetran counts monthly active rows per connection, Airbyte counts database gigabytes after serialization, and Datastream counts processed bytes that can be 2 to 5 times the source size. 6. **Is your warehouse available?** Some vendors list destinations as beta or coming soon. 7. **Where is the data hosted, and which rules apply?** The vendor should confirm in writing where the data is processed, which rules apply (such as LGPD, Brazil's data protection law, or PCI DSS for card data), and which language support speaks. The pricing models differ the most in step 5. Running a trial on your real tables answers it faster than any calculator. [Erathos](https://www.erathos.com/en/pricing), Fivetran, Airbyte and Estuary all have a free tier or trial. ## FAQ **What is a data transfer application?** A data transfer application copies data from one system to another, such as from a Postgres database into BigQuery. Some move a database once, like AWS DMS. Others keep a warehouse in sync on a schedule, like Erathos, Fivetran and Airbyte. **Can AWS DMS replace an ELT tool?** AWS DMS can run continuous replication with CDC, so it can keep a database copy in sync. But it only moves databases, so for SaaS sources and warehouses outside AWS you still need an ELT platform. **Which data transfer tool is free?** Azure DMS Standard is free for offline migrations, and dlt is a free open-source library. Erathos is free up to 1M rows a month, Fivetran up to 500,000 MAR, and Estuary up to 10 GB a month. **What changed in Fivetran pricing in 2026?** Since January 1, 2026, deletes count toward paid MAR, and each standard connection with 1 to 1M MAR has a $5 monthly minimum. **Can I self-host Erathos?** No. Erathos is a managed service. It connects to databases over an open connection, a fixed IP or an SSH tunnel, and dbt can run transformations after each load.