Backpressure

  Backpressure occurs because inflow tasks are outstripping our ability to process the data. Perhaps a denial-of-service attack is underway, rendering lots of data without pagination, or constrained shared resource such as a network printer.  Backpressure occurs more often in web/cloud environments when scaling challenges are encountered. 

  

Strategies

• Better Design -  a) Minimize Rendering - ex: design your customer grid to be on a tabbed notebook page or page that can be pinned, so does not need to populate the grid if not seen. ex2: add collapsible areas so user can control. ex3: have menus to take you to a different screen. ex4: pagination of grids; b) Minimize Data Traffic - ex: design your web page to only download to client first time and only send out updates when updates occur or user does action such as key strokes.  ex2: Filtering as specifically as possible, before retrieving data.
• Buffer/Queue - accumulate incoming data spikes temporarily, but this is switching to potential memory issue by deferring things
• Control the producer (block/slow down/speed up is decided by consumer). Locking mechanisms, synchronous programming, token ring network, block lists, security permission checking and blocking unless explicitly white listed.
• Drop (sample a percentage of the incoming data) - ex: a "first person shooter" skipping ray casting of objects behind walls. 
• Hand Off - ex: Load Balancer - manager switches to another processor when some processor is overwhelmed. ex2: Asynchronous programming - throw to another CPU.  
• Ignore Backpressure 
• Improve Capacity - ex: add more memory, upgrade the processor, etc.

• Progressive Rendering - ex. start with low quality or resolution image and as more data arrives sharpen the resolution
• Subscribe rather than polling