Jump to content
RESET Forums (homeservershow.com)
Alex.Covecube

Successfully Enabled Port Multiplier on eSATA

Recommended Posts

Joe_Miner

Nice to see that the MicroServer has port multiplier support. My thanks to Alex for exposing this little nugget.

 

I've been messing with it for a few days, and have a few things to report:

 

1) The MicroServer's port multiplier (PM) support is command-based switching (CBS) only. This isn't "bad"; it's just not (quite) as good as it could be. The other mode of PM support is frame-based [aka FIS-based] switching (FBS), which is more time- and bandwidth-efficient. (Results below)

 

2) [With the appropriate BIOS upgrade, and settings] Both the eSATA port and the on-motherboard SATA port have PM support, and can be used concurrently. I don't have the weird cable/connector required to test the caddy SATA ports.

 

I'm running a Linux kernel (v3.2.29). I connected two SansDigital TR4M 4-drive enclosures (they use the Silicon Image SiI3726 port multiplier chip), each containing 4 drives ... plus 4 more drives in the caddies. 12 drives all active and transfering together--a joy to behold. Performance is decent, and should be satisfactory for typical media-serving duties. I'll attempt to quantify:

 

First a general (non-PM-specific) tidbit that Joe_Miner may want to include in the appropriate summary post: (Both my MicroServers are N40L's, but this very likely holds for the 36L & 54L too) The combined sustained bandwidth for the MicroServer's SATA sub-system is ~650 MiB/sec. For example, if I connect 6 x 125 MiB/sec drives (e.g. Hitachi 7K2000 2TB), and crank them all up transferring from their outer (ie, fastest) zone, they all measure in the 105-115 MiB/s range. Crank up only 5 at once, and they all achieve their top speeds.

 

Now, PM-specific performance: with both enclosures holding 4 drives, and all 8 drives powered and recognized/enumerated.

 

When transferring from (or to) only a single drive in an enclosure, that drive will exhibit pretty much the full extent of its performance. The only limitation I could expose was that a fairly fast SSD (Samsung 830) only achieved ~215 MiB/s(vs 250+ when it had a direct [no PM involved] connect). Access time performance was not hampered in the least (mechanical or SSD). All of the above held true when both enclosures were performing the same (single drive each) test concurrently.

 

It is when we operate two or more drives (in the same enclosure) simultaneously where CBS operating mode exposes its inferiority (vs FBS). When simultaneously transferring from two drives (in one enclosure), the transfer rate for each drive is ~60 MiB/s; for 3 drives, ~34 MiB/s each; for 4 drives, ~20 MiB/s each. The same results are obtained even with both enclosures operating concurrently--i.e. no further degradation (with 2 drives transferring in each enclosure, those 4 drives will each achieve ~60 MiB/sec).

 

With all 12 drives trying to do their fastest, the 8 drives in enclosures reach ~16-18 MiB/s each, with the 4 "direct" drives sharing the remaining 500-525 MiB/s.

 

All of the above transfer testing involved files with no fragmentation (essentially no seeking required). For multi-drive (simultaneous) operation, CBS mode PMs can really suffer. This is because, in CBS, only one drive can be acting upon a command at a time, and that command must complete before the next command can be issued to that PM/enclosure. A glaring example of this would be running an Access Time test on all 4 drives at once. Using 4 of my 7K2000s as a test case:

 

These drives have an average access time (seek + latency) of ~13 milliseconds. So a test comprised of 1000 randomly-chosen sectors (LBAs) will take ~13 seconds. And when we perform that test on just one such drive (in the enclosure) it does take 13 seconds. But if we run the test on two drives (in same enclosure) simultaneously, they both complete after 26 seconds; 3 drives, 39 sec; 4 drives, 52 sec.

 

However, those same tests on a FBS mode PM enclosure would all complete after 13 seconds, regardless of 1, 2, 3 or 4 drives concurrently. True multiplexing vs round-robin operation. For data transfers, FBS is only limited by the SATA II connection which is generally good for 230-260 MiB/sec. 3 drives concurrently will each run in the 75-90 MiB/s range; 4 drives 55-65 MiB/s. Yes, FBS sure does sound appealing, but let me warn you: with FBS, the controller AND the driver must be *known* to work properly *with* the PM-equipped enclosure you have. Else it is headaches and heartaches, I promise you. If the CBS provided by the MicroServer can meet your needs (even if not your desires), don't go looking for trouble [you will find it].

 

--UhClem

 

UhClem:

Very Very Nice bit of research and write up! +1

Joe

Share this post


Link to post
Share on other sites
ikon

Yes, very nice indeed UhClem. All that to say though, in the end, the box does perform significantly faster with the Mod BIOS than without ;)

Share this post


Link to post
Share on other sites
ziki

How can you take advantage of a port multiplier?

Share this post


Link to post
Share on other sites
ikon

A Port Multiplier aware SATA port allows for the connection of a multi-drive enclosure with a single SATA cable while retaining the ability to see each of the drives in the enclosure individually. This can be advantageous in some situations. For example you could group the drives as you want them using DrivePool or DriveBender. IOW, you don't have to depend on the enclosure's inbuilt RAID or JBOD functionality.

 

Hope that's the kind of info you were looking for....

Share this post


Link to post
Share on other sites
ziki

Hmm I see. And what if it there's no mutiplier support? What would be different then? Cause i'm thinking of grabbing a Raidsonic IB-3640SU3.

 

http://www.raidsonic...e_objectID=8248

 

IB-3640U3_big.jpg

 

It is a 4 drive enclosure that connects to an eSata or USB3. If the machine doesnt allow port multiplier, will it not support this at all with the eSata port, or will it not see each drive seperately?

 

Is it also nesecarry for a setup like this?

 

P73055437.jpg

Edited by ziki

Share this post


Link to post
Share on other sites
ikon

I had to do a little research on this box because they didn't make clear which type of JBOD the box is providing. Unfortunately, there are 2 definitions of JBOD: the real JBOD, which is where each drive is presented to the computer individually, and a second erroneous definition where all the drives are presented together as a single spanned volume.

 

The definition makes a huge difference. Although the spanned volume definition of JBOD is not really correct, it is nevertheless quite commonly used, even by equipment manufacturers.

 

That said, it looks like RAIDSonic is using the correct definition, which means each drive is presented individually, which means each drive gets its own drive letter in Windows. This is problem if the modded BIOS is not used on an N40L. For JBOD to work with an external enclosure I believe you require port-multiplier on the host system. On the N40L this means using the modded BIOS.

 

So, if you don't have port-multiplier on the eSATA port, I think you will see either no drives, or only 1 of the 4.

 

BTW, if the RAIDSonic had used the spanned volume definition of JBOD, it would have presented all drives in the enclosure as a single large drive. It would be similar to DE in WHSv1 wherein the drives would effectively be stacked on top of each other as if they were 1 drive, even if the drives are of different sizes. Data is written to the 1st drive. When it's full, data is written to the 2nd drive, and so on. There is no load balancing by writing to all 4 drives at once.

Share this post


Link to post
Share on other sites
Puulima

I posted in my other thread on an alternative to RoboCopy that RoboCopy kept stalling - after some testing looks like its something to do with the AMD_SATA drivers (or something related?). I looked in the event log and see the same WARNING every few minutes:

 

WARNING - Source: amd_sata, event 129. In the details it notes: Reset to device, \Device\RaidPort0 was issued

 

so I googled that and found this posting:

 

http://social.techne...8b-e4d245bc33eb

 

If someone with more technical knowledge can look at the link above and advise - I'd be truly appreciative as I'm not sure what to do at this point to troubleshoot further. In the posting there are a few possible solutions being proposed - the person that stared the thread ends by disabling something in Storage Controllers - but I don't see anything in Storage Controllers except for the Windows 8 Storage Spaces Controller? One other person also notes a failing drive - but all 4 of my drives in the external enclosure are brand new and were tested for 24 hours with zero errors so seems odd that one could be failing already/

 

And If a hard drive is failing - how do I find out which drive? I've only got new drives installed in the N40L and External enclosure right now...

 

Thanks for any assistance on what to look for or try next.

 

EDIT: wondering if the copy operation I have running will be ok also? It keeps stalling due to this error, but resumes again.

Edited by Puulima

Share this post


Link to post
Share on other sites
Puulima

Referring to my post immediately above - seems the solution to this error on many other machines was to switch form the AMD driver to the Generic Windows Driver and this goes away. Obviously if I do that I will lose the Port Multiplier functionality though - and given others are using the AMD driver succesfully - can't be the driver right?

 

Couple of things though:

- I'm using the LATEST AMD drivers which I had note in an earlier post in this thread were NEWER than those noted in Post#1 to this thread - so wondering if that could be the issue? I'm not sure how to find/locate the PRIOR version though to test - and if I do locate the prior version - can I simply change the driver? I'm going to ditch my Win 8 Storage Space on the External enclosure for now so assuming I can try this

- I'm using the Windows 8 driver (not the Win 7) but that would seem to make sense given I'm running Windows 8 on the N40L - so not so sure that I should switch to the Window 7 version noted in the first post to this thread

 

Additional question: since I will be going to WHS 2011 - I'm wondering which version of the AMD drivers I should install then? I'd appreciate some feedback on that also.

 

Any ideas on the above issue?

Share this post


Link to post
Share on other sites
ikon

What I got from that forum thread was that the guy had an AMD RAID controller on his system that was causing errors. But, since he wasn't actually using RAID, he was able to disable that controller and get rid of the errors without losing any functionality. otherwise.

Share this post


Link to post
Share on other sites
Puulima

What I got from that forum thread was that the guy had an AMD RAID controller on his system that was causing errors. But, since he wasn't actually using RAID, he was able to disable that controller and get rid of the errors without losing any functionality. otherwise.

 

Others in the thread noted a bad hard drive being detected after the generic Windows driver was installed - but I would lose the PM functionality. See the notes in my latest posting )probably crossed paths with yours). Who on the forums would be in the best position to decipher this error message and how to troubleshoot? If I roll back to the Windows driver - can't do PM :-(

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...