When WAN works like LAN: RStor's network acrobatics helps it wolf $45m from Cisco 'n' pals

Re: When WAN works like LAN

True... but you see, you have a whole load of potential collaborators colocating in these enormous data halls... and then you have geo-distributed object store with erasure coding... and then you have a compute system sitting on top of one node of an object store...

So depending on your geo-location and your erasure coding policy you can have 1 complete copy of your dataset sat physically underneath your compute, with all the bandwidth you can afford to buy to cover that 2.4m or so, whilst at the same time having 4/9ths of the entire object/dataset sat a few metres from some collaborator that you'd like to share or move or copy the data with/to... and your object store is slowly distributing itself 24/7/265 to this location so in terms of the speed that someone sees when they hit GO! on the copy tool, they are already ~ half-way there to getting the full dataset. And of course, if you have TWO massive data halls linked by very fat pipes in addition to your local datacentre where you are generating the data, and/or are processing that data, then that might be a very fast transfer indeed. Because it's the PERCEIVED speed for the user / owner of the data that's what is going to make them want to hand over the money for that service. As well as efficiency of course.

So I think there's some mileage in RStor's approach, even if in trying to run it like they say, they're going to miss out on a lot of market.