data paths

[ilkhan]

This may seem like a dumb question, but it really makes a difference in a seperate discussion Im having.

When you give a copy command, say, from one SATA (onboard port) to another SATA (RAID card) HDD, in high levels terms what traces does the data travel upon.

Is the path HDD->cache->cable->ICH10R->DMI->X58->PCI-E->RAID->cable->cache->HDD?
or does the data travel through the CPU? RAM? Is there another route? What if its SATA HDD -> network, any (path) differences there?

 

[SonicIce]

I think you really need an engineer to answer that type of question :Q

 

[Idontcare]

I'm no expert on data topology (the word you are looking for to describe the nature of your query is topology, not traces...topology vs. traces is kinda like ISA vs. architecture) for PC's but what you laid-out in the OP is pretty much how I envision it as well.

The data passes thru a controller chip on the raid card (like the ICH10R logic chip in your proposed topology) in addition to local cache within the hard-drive itself (assuming the "cache" you show in your raid controller card side of the topology was meant to represent the cache on the controller card itself).

Regarding CPU involvement in handling the data: PIO mode versus DMA

In PIO mode the CPU is heavily involved in "handling" the data as well as "managing" it thru it's journey, whereas DMA involves the CPU only in a rather subtle manner during the initiation phase and ending phases of the transfer but the data itself doesn't "touch" the cpu at any point¹.

Over networks, there is a protocol called Remote DMA.

¹ A caveat here might be anti-virus software, which for some "realtime" settings would seem to require the data itself to be cycled thru the cpu as the AV software analyses it to confirm there are no signs of embedded virus in the file(s) while the files are being handled by the system.

 

[ilkhan]

how about for something like an MP3 being played?

I was thinking traces like wires, but topology is probably better.