Adaptec 39160 Very SLOW

A

alphaa10

n00b
Joined
Jun 9, 2011
Messages
7
I have an Adaptec 39160, and its data rate suddenly went from 80 MBps to 40MBps when I added a second, but identical drive on channel A.

Of course, I have checked all the usual suspects, including making sure jumper settings on the two drives match (except for SCSI ID), and even restoring the 39160 to its factory defaults, but came up with no discovery (except that one of my SCA interface adapters was faulty, and I replaced that, but without improvement beyond 40 MBps). Termination is not an option on this drive, so I use a passive terminator at cable's far end, and auto-termination at the adapter end.

What puzzles me is the adapter easily took the original installation of a single IBM DNES 318350 (17gb) drive on channel A as an SE device and ran it happily at the drive's max speed of 80 MBps, so this seems to be an adapter logic issue, not a true hardware limit.

This is a new Dell / Adaptec unit, sold in white box without the colorful Adaptec retail box, which means I provide my own troubleshooting whenever I can . But I am nearly out of bullets, unless I concede that putting two identical drives on the same channel and watching their data rate drop to 40MBps is "normal" for this configuration.

Yes, I can use the second channel for the other drive, but nothing in the Adaptec literature suggests I must use one drive per channel to obtain maximum data rate on each drive.

Put another way, the big question is whether available channel MBps bandwidth is divided among all devices on that channel. In the terms of bandwidth, alone, this certainly makes sense. But if adding more devices progressively reduces the data rate, this makes little sense for the typical enterprise installation of many drives, if every machine with three or more drives on the channel must slow to a crawl.

BTW, when I discovered the 40 MBps speed limit on adding the second drive, I immediately removed the second drive, hoping to revert to the previously reported 80MBps, but the first drive-- now alone on the channel-- has remained at 40 MBps despite all efforts, including returning the adapter to factory defaults.
 
The reason a SCSI bus syncs at a lower rate is usually cable related. SCA adapters especially seem to be very picky things. The cable is apparently OK for 80, so I'd look at the SCA adapters. I've had ones that were rated for 160 not want to go over 80, replacing with an apparently identical one fixed it. The cheap ones seem to be inconsistent.
 
Thanks for the excellent news-- if anything needs to be replaced, this is the least expense possible (other than a mis-jumpered drive). The news is also comforting-- I have run SCSI for several years, and am familiar with normal setup, as well as routine mistakes (I've made many of them, at some point). So, your guess is likely, and as I indicated earlier, one of my SCA adapters actually died during the process of attaching a second drive (the deceased SCA adapter was the older of the two). SCA adapters are often delicate and balky creatures.

Now, in regard to the same speed issue, today I telephoned some businesses that sell the 39160 and similar adapters, and was told conflicting stories. Dell, for example, had a Tier 1 tech who was the soul of good intention, but his/her advice was that two drives "split" the available bandwidth, and I should expect two 80MBps max drives to run at 40MBps each. I then asked, how do enterprise systems use as many as 15 drives per channel, or a total of 30 per adapter, and still manage to get serviceable performance?

Telephoning another Adaptec vendor, I was assured there is no limit to bandwidth imposed by the actual number of devices. While there is also every reason to believe this vendor is correct, it is surprising the SCSI speed issue is not well understood through the industry. My growing suspicion is SCSI is so old, at this point, younger techies do not bother to study it as deeply / intensively as they once did, and instead invest their efforts in SAS and SATA devices.

An Adaptec salesman said the market is moving toward a 6GBps data rate for applications, especially RAID, and various new adapter-based RAID options offer better security, but with fewer drives.
 
I'm so glad they phased SCSI out and developed sas. It is far superior. There are too many moving parts so.to.say in a SCSI bus.
 
You have helped the diagnostic a great deal-- as a result of your comment about the cable and SCA adapter, I have determined I have vintage 2000 SCA-80 adapters that are not at least U160 certified. Although the two SCA adapters and the two LVD/SE multimode IBM DNES 318350 drives were sold together, I have no assurance the adapter, itself, is suited to LVD operation. The original vendor may have presumed all buyers would run each drive as SE.

Additional clue-- the drive ran well at 80MBps on the SCA adapter when it was the only drive on the LVD cable. But with two identical drives on the cable (with unique SCSI IDs), the 39160 controller reverted to 40MBps, despite my assigning 80MBps in SCSI Select. I may be in the market for a good SCA-80 LVD/SE adapter, certified for at least U160.
 
Yeah, I was using one of those and yes, it is definitely time for an upgrade. SAS is far superior in all areas, definitely go for the upgrade. ;)
 
SCSI bandwidth is shared among the devices on the bus, you don't get a lower bus speed from attaching more devices as long as everything meets the spec. Also check that the cable is plugged into the correct connector on the adapter, some adaptecs have seperate ports for LVD/SE. I know the 29160 is like that and the 39160 is just one of those with 2 channels I believe. The 39160 is capable of U160, that's what the 160 in the model number signifies.
 
dustNbone said, "The cable is apparently OK for 80, so I'd look at the SCA adapters. I've had ones that were rated for 160 not want to go over 80, replacing with an apparently identical one fixed it. The cheap ones seem to be inconsistent."

----

That was your key comment for my diagnostic, and I appreciate your reading my problem description so carefully.

1. I was reluctant to believe the vendor did not understand LVD well enough to know he must sell LVD/SE rated adapters when he sells multiple LVD/SE drives to the same customer, in the same order. (Of course, even if the adapters are not specifically labeled LVD/SE, it is also possible the adapter(s) are simply too old, and wandering out of spec.)

2. You appear to have substantial SCSI experience. Are you now using something other than Ultra 160 or 320? Are you planning to migrate to SAS?

3. What is your favorite vendor, based on (1) price (2) experience and (3) willingness to help (to some degree) with tech support?

BTW-- Only two internal ports are located on the 39160 for a 68p cable, with an additional port (what appears to be a 50p) on the card end.
 
QUESTIONS FOR ANYBODY

More detaills about my diagnostic--

1. The Adaptec 39160 controller auto-terminates the 68p LVD cable, but the other end of the cable is served by a terminator which I believe gets power from the cable (ie. from the adapter, whose SCSI Select assignment is terminator power ON). Of course, to be sure, I can check the order record whether I have a passive or active terminator installed. According to many sources, however, only an active terminator permits LVD operation beyond 20MBps, and I have 40MBps for each drive.

QUESTIONS:
Considering that I appear to have an adapter-based power termination, but any device can wander off-spec over time, is it possible I could have a problem with the current (old) terminator, itself? Do I need to purchase a power terminator (active negation) specifically labeled LVD? The current terminator at cable-end is not labeled, at all.


2. At a SCSI Tutorial provided by scsita.org (http://www.scsita.org/aboutscsi/SCSI_Termination_Tutorial.html#5), under Section "5. Connecting WIDE SCA-2 (80 pin) drives to a WIDE bus (68 pin)" the tutorial states, "There is no term power line on the 80 pin connector, termination must be provided in the adapter or backplane... Power must be provided by the adapter or backplane. The adapter has the ID switches, there are no ID settings on the SCA-2 drives..." Which last statement contradicts the fact each of my IBM DNES 318350 drives has a jumper block (J4) which specifically provides for a jumpered SCSI ID from 0-15.

QUESTIONS:
Why do I have a J4 jumper block for SCSI ID assignment, and the Adaptec SCSI Select setup utility, itself, does not provide SCSI ID assignment, if "... The adapter has the ID switches..." ? Is this simply one of the perils of making generalizations?
 
GOOD NEWS, AT LAST

Thanks to all, especially poster dustNbone, who all but found the problem with a shrewd guess based on his experience with bad cables and drive adapters.

As it turns out, the problem was two faulty SCA-68 drive adapters, which left me with two drives running at speeds like 40MBps or worse. The adapters masked the problem at first, since I had not changed anything before the drop in speed except add (1) one drive and (2) one adapter.

A problem with one drive / adapter pair should not affect the other, but apparently it did.

Of course, my first step in starting the diagnostic was to revert to the original configuration, but the first drive-- which had run at max speed moments before the second drive was attached-- now ran at half speed, even when it was the only drive on the cable. Worse, still, only minutes later, the first drive no longer appeared on the boottime drive list.

To isolate the problem, I had to find a known-good cable, adapter and terminator. Fortunately, I was able to create a test link (cable, adapter and terminator) for comparison with other devices. In a matter of minutes, by substitution of parts, it was possible to isolate any device which reduced performance.

Again, thanks to all who offered help.
 
You must log in or register to reply here.
Back
Top