To be fair, Seagate has ‘cheated’ a bit here on the UEBR side of things as they have included their Super-Parity configuration, and possibly even beefed it up. Either way, this is a technology they have included on Exos X models for a couple of generations now. If one is overly cynical, one could almost line up when the first gen Exos X got Super-Parity with when HAMR was originally getting ready to go into production… as it is meant to shore up concerns over UBERs in ultra-dense drives like this 30TB’er. After all, businesses spend a lot of time and resources ensuring UBERs does not result in corrupted data. Of course, when dealing with 30TB drives, a UBER of one read failure for every 10E15 bits (aka .125 Petabyte aka 125TB) is the same as saying that (statistically speaking) for every ~4.17 Full Drive Reads (FDR) you can count on a bit being wrong and not recoverable at the drive level. Compare and contrast that with the Exos X24’s 5.208 FDR, X22s 5.682 FDR, the X20s 6.25 FDR… the X16’s 7.8125 FDR, and this downward spiral must be stabilized before we see 50/60/100TB drives. Lest then not be taken seriously by the Enterprise consumer… as at some point even advanced Array level ECC (e.g. Raid6) will not be able to keep up with the potential for data corruption.
Either way, Super-Parity is excellent tech that is certainly helping to keep the chaos of UBER (somewhat) at bay. One where it has tangible real-world benefits beyond ‘trust me bro’ UBER promises. We say that as Super-Parity, in the most simplistic terms possible, draws heavily from old SandForce SSD controller algorithms, in that the tracks and platters are not considered as one cohesive whole storage unit. Instead, just as SandForce controllers did with their NAND ICs, Seagate has broken it down virtually into a (highly) modified (internal) RAID array type layout.
So, instead of Data+Parity, Data+Parity linear layout, Super-Parity means that the data is being laid down can take one of two known approaches (as laid out in Seagate and Hong et al. US patent 9,633,675 B2 and previous US20170322844 patent application, both circa 2017). In the first, all the data is written, and then separate “super blocks” consisting of nothing but said parity information (during a second silent write operation) are completed. Typically, to a separate adjacent track. An adjacent track that will only be filled with Super-Parity ECC data. To imagine this, think of RAID 5 with its data and parity stripe that spans all the drives in a given array. Just at the track level, not the drive array level.
(image courtesy of wikipedia.org)
The other (known) option is to instead of cutting all (for example, ten) platters up into data + super tracks on all (ten) platters… the controller can dedicate a full platter (or platters depending on ECC size) just for the parity data. The upside to this method is faster completion of writes (as the parity ‘super blocks’ are written at the same time as the data blocks. Thus no second and separate write operation occurs) but this would come at the expense of sequential r/w performance… as a full platter and its r/w arms are basically not being used for data read/write operations. Rather than platter and arm is just being used to write the ECC information… and only read from it when things go sideways. To imagine this operation in action, instead of imagining a RAID 5 array. Think of it as the odd-ball RAID level 3 or even 4, just with the dedicated parity disk being a dedicated parity platter (aka a surprisingly sane way to implement an insanely temperamental RAID design).
Seagate is mum on which Super-Parity option they use (and if they even use either option)… but option 2 in the older Exos X 24TB and yet option 1 now being used in the new Exos M would explain why the major differences in underlying write tech still result in very, very similar write performance. Since Seagate is not saying… they can easily silently swap from one to the other (or use a hybrid of both) as HAMR tech gets faster at writes. Either is acceptable as both work at ensuring the data is safe and secure.
Needless to say, that is a lot of decision trees (and parity calculations) that take a lot (by HDD standards) processor cycles to run through in real-time. Thus, Seagate has finally upped the horsepower in a significant way. While details are… “thin” to say the least, what is known is this 12nm RISC-V based SoC is an entirely in-house design. One that more than likely leverages all the IP Seagate obtained when they bought SandForce (a cutting-edge SSD controller manufacturer back in the day). Which means it probably is using a processor microarchitecture that features wide(r) pipelines and deep(er) parallelism to allow it to enable simultaneous execution of multiple instruction streams. That is just a S.W.A.G. so please take it with a grain of salt. Either way, this SoC is an entirely new beast compared to previous models. Thankfully, this new beast is being baked in a 12nm node oven… so it probably doesn’t use that much more power than previous SoCs Seagate used.
