What kind of HDD connectors are this?

Solution 1:

UPDATE: the world works in mysterious ways.

TL;DR: This connector is SSA - Serial Storage Architecture, developed by IBM in 1990, and used in their storage arrays (paired with mainframes, which are predominantly consist of Processing, RAM and I/O) until at least the z9 series. Logically speaking SSA uses SCSI commands. At some point it was superseded by Fibre Channel

I recently stumbled across a tweet from T0py (@InfoSecFriends) showing progress on a z9 Mainframe they are playing with... they've have kindly shared a photo in a tweet from the storage array showing the other end of this connector.

So, I'd suggest that yes, this is very likely a hard disk from an IBM machine - possibly a z9 Mainframe (or another generation).

As @reben has mentioned, it could equally be another machine running AIX (though I can't immediately identify what hardware would have had SSA support).

From this photo we can confirm that the right-most connector (bottom here) is indeed power, and I'm still placing bets on the middle-right being for management.

Interestingly my expectations regarding the dual-path data channels are a little off - here you can see that the disks are daisy chained, forming a loop. This topology can offer redundancy for a single link failure, or a single disk failure (or removal)... Interestingly though, if multiple disks were to fail (or be removed, or multiple links), then you will loose access to a whole section of the loop. For this reason I suspect that multiple smaller loops will be employed - perhaps one per row.

Original Answer:

I can't make any bold claims as to what this is... but I can tell you it is not standard SATA or SAS - sorry.

I can also guide you through some of the interesting points from your photo.

I'd guess that "hard drive" is a reasonable assumption here... the black metal protrusion with what looks like 4× contacts is probably the spindle motor, and the mounting holes look reasonable for standard hard disk mounting...

The picture is pretty blurry, by we can see that there are parts from least three major manufacturers - TI, ST and Intel - which isn't particularly strange, but worth noting.

We cant clearly see the part numbers, but I'd hazard a guess along the following lines:

• TI - a DSP or other signal processing
• Intel - Flash memory - probably holding the firmware
• ST - probably general board control or motor control.

The big slab on the top-left is probably the main processor, handling I/O with the host(s)... below that is probably the RAM for caching.

I suspect (given the separated connectors) that this device supports two data paths - most likely for redundancy... the differential pairs are highlighted in yellow / purple below. These will probably be connected to two separate storage controllers, meaning that if a controller goes down, the disk is still accessible by the "other".

What is a bit strange, is the size of the inline resistors(?) - from a signal integrity point of view they are huge - especially when compared to the much smaller capacitors(?) just above.

Also, while the left two connectors appear to be 3-pin, I'd speculate that they are actually 6-pin (3× on either side). This allows for the differential signals (4× signals to each connector) to be separated by a pair of ground pins.

The connector to the far-right is most likely power. This is evident by the fact the some of the pins are elongated, providing a first-mate-last-break connection for the 0v rail. Additionally, the pins are ganged and feed directly into large capacitors and sizable copper pours - they are not signals.

This leaves the middle-right connector, which I'd guess is a low-speed interface for system health and monitoring purposes - something that will quite likely be handled by a totally separate subsystem - especially if this belongs in a system like what I have in mind...

Sadly a bit of rummaging provided no good information about what interface IBM use for hard disks in their "recent" System z9 and System z10 mainframes...

Bear in mind that these things are seriously over-engineered and very redundant...

• They have whole CPU dies that can be swapped in if another fails sanity checks.
• IIRC the System z10 is able to support a whopping 1.5 TB of fully redundant RAM.

These systems are designed for 99.999% uptime (five-nines), which equates to a downtime of approx 5m15s per year.

Redundancy here isn't just about failure, but also serviceability while the system remains fully functional.

I know that this question is quite old now, but if you have a chance to add more photos or information, I'd love to see it.