No data is transmitted to or from the BigML system in normal operation. All computational requirements are installed locally. Once BigML is implemented On-Premises, BigML has no access to the environment or data at all, unless there is an explicit request by a customer for BigML to analyze a sample of the data exposing a problem. In that case BigML prefers a written grant (an email to email@example.com) of permission to examine a sample of data that would help the BigML Support Team debug a system failure. All copies of this data would be deleted upon completion of the troubleshooting.
With your On-Premises deployment you are free to firewall, block or otherwise restrict access to your environment. BigML will provide you with a maintenance guide and updates will be provided to your IT staff to be installed. Nevertheless, BigML recommends allowing for the possibility of giving access to the BigML Support Team for assistance with updates or troubleshooting. The method for granting access is at your discretion and can be arranged as needed. Some of BigML customers use VPN connection, some others simply use a Webex connection with remote control, which also has the benefit of allowing your staff to monitor what actions BigML’s Support Team is taking in your environment which is beneficial both for auditing and for training.
Note: BigML reserves the right to collect and review usage statistics. This is limited to metadata such as: the number of tasks that have run, what size they are, how long it took, etc., but it will never include any customer specific data. BigML uses this metadata to understand how the customer is using the platform and if it is performing adequately.