# Security & Stability

DataFlint for Spark is highly secure and stable, and this page is explaining why:

### Security

1. DataFlint is **running locally** on your spark driver or history server
2. DataFlint is using the **existing Spark UI endpoint**. so no new endpoints or ports are being exposed
3. DataFlint is **open source**, so you can see what the plugin does. No black-box wizardry!
4. The dataflint liberary jar is a stable version in maven central OSS repo, and maven does not enable editing or changing stable versions. **Meaning it's not possible that the code that runs in your cluster to change**.

### Stability

1. If DataFlint failing on startup it will throw a warning and **let the app continue**
2. Dataflint is running code in Spark **when you access the Web UI**
3. Errors in DataFlint in the driver are in a **separate thread and should not effect the app runtime**

### Performance

1. Most of the compute is being done **in the DataFlint Web UI side**
2. DataFlint **only runs compute on the driver**, not on the executors
3. DataFlint runs compute on the driver **only when the DataFlint Web UI is open and the tab is active**
4. DataFlint query the driver API ever 1 second, so the **performance impact is similar to looking at the existing Spark UI and refreshing constantly**

### Telemetry

We collect anonymous metrics via MixPanel on usage of DataFlint. \
We do not collect any data about your actual spark job beside:

1. Spark version
2. App id