ETL from 250+ disparate customer host systems

My task is to produce an ETL process that pulls data daily from our customer systems (which can be most any database/OS). The data will be loaded into our central system, an analysis will be run and the aggregate data solution sent back to all customers.

Any recommendations for tools to achieve this scenario? I'd like to be able to connect to each customer's host (via whatever protocol) and execute the ETL from my central location. I will be able to install a client on the customer's system, in most cases. I've had some frameworks/tools recommended to me but I don't have enough experience yet to decide the best course of action.

Here are my current ideas, very open to suggestion:

1. Communication/messaging to heterogeneous systems (Apache Service Mix)
   
2. ETL process & data quality/cleansing/reporting (Apache Camel, Pentaho, Clover etc.) to pull data in
   
3. Automated analysis and processing (ANT, ETL tool, legacy apps) to generate solutions
   
4. Package/distribution of results (Apache Service Mix, FTP)
   

Any insight is greatly appreciated,
Bill