Ignore drive write messaging… SSDs are more reliable than HDDs, Period.

This blog mostly pertains to SSDs in server and storage arrays.

All NAND relies on electrical charges in silicon. Writing data involves a Program Erase (P/E) cycle, also known as a write cycle. And there are write cycle limits as to how many times a NAND cell can hold a charge:

  • SLC (1 bit/cell) write cycle is 100,000
  • MLC (2 bits/cell) write cycle 10,000
  • TLC (3 bits/cell) write cycle 3,000
  • QLC (4 bits/cell) write cycle 1,000

This is a race to zero. Why? Because more bits per cell equals a lower cost/bit. QLC NAND is cheap compared to SLC or MLC NAND.

Use cheap NAND for cat-videos or for USB thumb drives? Okay, fine.

But using cheap NAND for servers and storage arrays is a profoundly bad idea because high write workloads are a threat to SSD reliability. So using the cheapest possible flash for servers and storage arrays is unwise.

I am not a fan of the SSD marketing teams who repeatedly offered the lead marketing message of “the SSD I want you to buy … 3,000 P/E write endurance… they will surely fail”. Yuck – when that marketing message could have been “the SSDs I want you to buy are far more reliable than hard disk drives, which fail about one per cent per year.”

SSDs are more reliable than hard drives, period.

But we still have to deal with the reality of NAND write endurance. We should rightfully focus on the device – the SSD – and not the NAND itself.

Server/Storage SSDs added over provisioning, with spare cells held in reserve to take over when used cells reach their wear limit. Next, SSDs were rated with Drive Writes Per Day (DWPD), and then further improved with ratings of TWD (Terabytes Written per Day) so that the wear rate can be tracked.

It is obvious, but bears repeating: a 4TB SSD will have twice the TWD rating as a 2TB SSD. Buy bigger SSDs for write-heavy workloads (duh).

Our humble advice… Use SSDs. SSDs are more reliable than HDDs, period. And when the workload is write-heavy, simply buy bigger SSDs, as they are harder to fill up.

The endurance problem can also be solved another way; compression writes less data to SSDs, thereby preserving the available write endurance (P/E cycles) capacity. Today we know GZIP and its kin. These are antiquated and inefficient; you can expect to be less than delighted. Keep an eye out for new generation of “Cloud-native” compression storage services and software.

Note: Consultant and patent holder Hubbert Smith (Linkedin) has held senior product and marketing roles with Toshiba Memory America,  Samsung Semiconductor, NetApp, Western Digital and Intel. He is a published author and a past board member and workgroup chair of the Storage Networking Industry Association and has had a significant role in changing the storage industry with data centre SSDs and Enterprise SATA disk drives.