HDD specs: Assess SATA vs SAS, sustained data rates and block size

Outer beats inner: HDD ‘sustained data rate’

In addition to reliability, the most important specs for HDDs are performance and energy consumption.

The highest performance is offered by HDDs that work at 10,500rpm or 15,000rpm, but they have been increasingly displaced by solid-state drives (SSDs). However, 7,200rpm enterprise HDDs still deliver sequential throughput of up to 280MBps and up to 400 input/output operations per second (IOPS). Storage systems with a few dozen of these drives can achieve more than 5GBps and 10,000 IOPS, which is sufficient for many modern applications.

Having said that, the performance of HDDs decreases with their fill level because the outer data tracks on rotating magnetic disks are longer and hold more data than those further in. So, the “sustained data rate” stated by the manufacturers in data sheets always refers to the outer tracks. Further inside, the value can drop to about two-thirds of that.

For companies that want to optimise energy costs, the most important consideration is the modernisation of their HDD infrastructure. With most of the energy used by an HDD needed for spindle rotation, storage capacity and the workload have only a small influence, so a few high-capacity HDDs are more economical than many small ones.

SATA vs SAS and energy costs

Enterprise HDDs are available with SATA or SAS interfaces, with SAS offering important features that SATA lacks, including higher signal strength, end-to-end data protection and dual porting.

But, usually, HDDs connect via a SATA interface – only 10,000/15,000rpm performance HDDs are available with a SAS interface. Today, a data rate of 6Gbps (called SATA 3.3) is standard, with backward compatibility to previous versions.

SAS is more expensive, however, and has slightly higher power requirements.

From 512 to 4: Different block sizes offer flexibility

In enterprise HDD data sheets, there is usually an indication of block size. This is the size of logical blocks that can be written or read from a hard disk. In the past, this was always 512 bytes, so drives had a native 512-byte sector. Later, larger sectors of 4kb were introduced to write and read larger blocks, which facilitates the management of high-capacity hard disks. In addition, error correction also works more efficiently with larger blocks.

Modern file and operating systems can handle native 4kb sectors on hard disks, but older versions often cannot. So, the 512e format was developed, which uses 4kb sectors but emulates eight 512-byte sectors in each of them. Older file and operating systems can write and read 512-byte blocks as usual.

When writing, however, there may be a loss of speed if the entire 4kb sector is not written. The hard disk must first read the entire 4kb sector to fill one or more of its emulated 512-byte areas and then write the sector back, so an additional read operation is incurred.

Different block sizes in enterprise HDDs give businesses the flexibility to choose drives that best fit their file and operating systems.

HDD security options: SEDs and SIE

Enterprise HDDs also offer flexible security options, such as self-encrypting drives (SEDs) and sanitise instant erase (SIE). The latter is Toshiba’s variant of widely available instant erase functionality.

SED is hardware-based encryption directly through the hard disk, which is very secure and offloads processing from the system in which the drive is installed. SIE is an option to securely erase all data immediately instead of going through lengthy overwriting processes.

SMR and MAMR for higher storage density

HDD models also differ in relation to recording technology used with conventional magnetic recording (CMR), SMR and MAMR available.

CMR has been in use for years and was formerly called perpendicular magnetic recording (PMR) to distinguish it from a predecessor technology, longitudinal magnetic recording (LMR). PMR has been in use for 15 years, hence why it is now called “conventional” and has reached its limit at 16TB per drive.

SMR increases storage density by working with overlapping data tracks, and so provides higher recording density. Reading the tracks works as before, but when overwriting an existing track, the data of the overlapping track must first be read and then written back with the new data. This can cause fluctuations in write speed, but caches and caching algorithms are built to handle them.

SMR is primarily used with PC and surveillance HDDs because they do not have to handle sustained high write loads with random accesses. For occasional writes or sequential data streams, such as those delivered by surveillance cameras, SMR is ideal.

Higher-capacity enterprise HDDs, meanwhile, rely on MAMR. A microwave-generating element on the write head helps to focus the magnetic flux so that less magnetic energy is needed for writing. The write head can thus be smaller and write bits more densely.

Currently, MAMR is used in 18TB and 20TB hard drives, and with advancements in this technology, hard drives up to 30TB can be expected in the future.

Since no data needs to be written via overlapping, MAMR is not subject to the restrictions and performance limitations associated with SMR technology. A combination of MAMR and SMR is also technically possible, but not very prevalent. With a combination of these two methods, it will not be long before the industry can benefit from capacities of up to 40TB, but with SMR-typical performance limitations when it comes to random write access.

Rainer W. Kaese is senior manager for business development in storage products at Toshiba Electronics Europe.