Now one of the accidental discoveries of our architecture very early on was that because we were always translating this virtual layer to a real layer, we know every transaction on a file in file namespace. Storage has typically been in block space.
Your IT guys can tell you that you’re getting a very high block rate, but they can’t help you identify a specific or local bottleneck in a workflow. Why is this application suddenly running slow? “Well I’m not really sure.”
But in PeerCache, we know the name of the file, we know the writer, we know the reader, we know every stat, we know every open, we know every close, we know every rename, we know absolutely everything.
And this spews out a massive log of events that get consumed by something like Elasticsearch, and then you can build your own dashboard.