FireWire 800 daisy-chain Vs USB 2: which is more efficient?

[noire anqa]

I'm running two of these enclosures:
http://eshop.macsales.com/item/Other World Computing/MEPT934AL2C/

each with 2x2TB drives in RAID0 (i.e. 4TB per enclosure). I use one 4TB array for my user files, and the other for TimeMachine.

My user array is connected via FW800, my question is - is it more efficient to connect the other (TM) array via USB2 or via daisy chained FW800?

The enclosures use the Oxford 934DSb chipset, which is capable of saturating the FW bus, so thats 100MBps. As i understand it very few interfaces are capable of saturating a USB2 bus such that practically speaking even FW400 gives greater real-world throughput than USB2, despite the 60Mbps difference in spec.

Since one is a TM volume, when backing up from the user array to the TM array will the data have to pass through the computer's controller? It definitely will if it's FW800 -> USB2, but will it have to if it's daisy chained FW800? I'm told this might not be the case as some FW controllers support P2P communication but i can't find too much info about it. Even if there is P2P support the OS needs to know the progress of file transfers and so forth so can it do this even if the drives are transferring via P2P or can it only monitor that sort of thing if the data goes through it's controller? Essentially is it up to the OS or the controller to select the optimal route for data? If it's up to the OS does Mac OS support FW P2P?

That said, even if the data does have to go to the computer from one array and back through the daisy chain to the other array, is it still practically faster than the USB2 route?

Any help is much appreciated! Thanks!

 

[hcuar]

TL;DR

But FW800 is faster than USB 2.0. Even when daisy chained. You can daisy chain 63 devices (including the computer). Provided there is enough power, the devices support that many devices on the bus, and no FW400 devices plugged in.

 

[FourCandles]

I would daisy chain via the FW800. I've currently got 3 FW800 drives daisy chained (scratch disk, media disk, backup disk) and all works fine.

My rudimentary copy speed tests are showing pretty much the same speed from one external drive to another as from the computer to an external drive.

 

[noire anqa]

TL;DR

But FW800 is faster than USB 2.0. Even when daisy chained. You can daisy chain 63 devices (including the computer). Provided there is enough power, the devices support that many devices on the bus, and no FW400 devices plugged in.

My concern was that though FW800 is faster, by having to send data from one array to the computer, and then back through to the other array via the same port - for a sustained read/write task such as this i'd effectively be halving the bandwidth. i.e. of the total 100MBps available 50MBps would be consumed reading from one drive, and 50MBps for writing to the other drive.

In this case USB2 has a theoretical peak of 57.5MBps, exceeding the 50MBps as in the above example.

And yet .. since USB2 more than likely isn't capable of hitting its peak throughput it's probably going at around 50MBps anyway: essentially it probably makes very little difference!

I think i've answered my own question:-

1) Worst case, the FW800 bandwidth is halved: at which point its practically equivalent to the USB2 route. Though FW800 still wins in this case because the daisy chain free's up a USB2 port.

2) Best case, the drives can P2P and the transfer runs at the full speed, far in excess of USB2.

Either way you've got to hand it to the FW800 bus ..

Thanks for the help!



EDIT: Yep, that confirms it - USB2 sucks. Long live FW.

Additional notes from Alex Esquenet - our engineer friend based in Belgium: "A fast usb host can achieve 40 MBytes/sec. The theorical 60 MB/sec cannot be achieved, because of the margin taken between the sof's (125 us), so if a packet cannot take place before the sof, the packet will be rescheduled after the next sof. On top of that, all the USB transactions are handled by software on the PC. For instance, a USB host on a PCI bus will send or receive the data via the PCI bus; the stack will prepare the next data in memory and receive interrupt from the host."

 

[noire anqa]

I would daisy chain via the FW800. I've currently got 3 FW800 drives daisy chained (scratch disk, media disk, backup disk) and all works fine.

My rudimentary copy speed tests are showing pretty much the same speed from one external drive to another as from the computer to an external drive.

Just out of interest, what are your read/write stats?

From one drive to another, is it going at around 50MBps (half bandwidth) or closer to 100MBps (full bandwidth)?

And from the computer to the external drive? same again?

 

[FourCandles]

Just out of interest, what are your read/write stats?

From one drive to another, is it going at around 50MBps (half bandwidth) or closer to 100MBps (full bandwidth)?

And from the computer to the external drive? same again?

I can't remember the actual figures, and to be honest like I said they were rudimentary timings of copying a 2GB file rather than actual read/write stats. But I do remember clearly that copying from one external to another took the same time as from the computer to an external.

 

[noire anqa]

I can't remember the actual figures, and to be honest like I said they were rudimentary timings of copying a 2GB file rather than actual read/write stats. But I do remember clearly that copying from one external to another took the same time as from the computer to an external.

I was just wondering if maybe the drives you were using maxed out at say 50MBps, in which case it would be the same speed if the data went via the computers controller and back, or if it went straight from drive to drive.

Either way, it's bound to be different for different controllers, i.e. my system from yours - so i'll just have to test it to know how it works on mine!

Thanks for your help