Web UI Dashboards performance at 100,000 datapoints is wonky

[jscottsf]

We need to start optimizing across the stack to improve performance for high-volume queries.

Initially we can reduce the timeframe for queries and increase the timeouts.

Some next steps to consider:

• Breaking out the Influx select service into it's own container for more parallelism
• Add caching on the back-end to speed up repeat queries for datapoints
• Consider fetching less data initially (e.g. random sample when zoomed out)
• Fetch less, then allow scrolling
• Add caching in the client (e..g don't refetch entire datasets)
• Only fetch recent data on the client (e.g. be more reactive on new data)
• Ensure we're using compression
• Move to protobufs versus JSON for datapoints (make it a content type option in the API)

[jscottsf]

Look for bottlenecks too. It's possible the issue is related to SQL queries for legacy datasets.