I just responded in the other thread http://www.satelliteguys.us/private.php?do=newpm&u=161374&postid=2999958 why I don't think they are storing a single stream for the whole transponder.
They can't store 8 days of full transponder anymore. They could store current day as the whole transponder and break it down. Don't know that that's the case, nor what type of CPU power it takes to do this. I'm sure this is one of several reasons Hopper has a "higher-end" CPU than Joey.
Even with a 6 tuner box you'd only be serving about 100mbps combined read/write requests for 10-12 "clients." I was putting single disk servers into production doing more work than that 10 years ago with less capable hardware.
It's still not the bandwidth, the inherent latency of rotating rusted iron when you have to move the heads constantly to drop your data for the correct streams. Seeks + Rotational Latency and the limits of either a 5400 RPM or 7200 RPM spindle. Double from 3 to 6 and increase the read only makes it worse. I don't recall that Hopper has tons of memory on board for FS buffer, which could help.
No, that functionality is already built-in to the Hopper. The current switch technology sends all the even or all the odd transponders for a sat to a tuner, even though the tuner really only needs a single transponder at a time. There is not enough bandwidth in the cable to stack six tuners that way. They would need to have switching smart enough to skinny it down to a single transponder per tuner, again much like D*'s SWM tech.
I knew I didn't know that, good information.