Your enterprise storage subsystem is a crucial part of your IT infrastructure, whether on one server and desktop, or 1,000 servers and 10,000 desktops. With so many choices out there for hard drives, solid state drives (SSDs), consumer drives, enterprise storage components and capacities, you’re bound to find the right ones to fit your needs. But what types of buses and protocols are those storage devices using, and when should you aim to equip your machines with which types of devices?
Storage interfaces connect the actual drives where your data resides to your system’s main processing bus; think of the interfaces as the highway data travels to get from your drive to your processor and memory.
There are two main types of bus interfaces in most PC production systems today: SATA and PCI Express, or PCIe. Each has its own history, design and proper placement.
Bus Interfaces: A Brief History
SATA, or serial advanced technology attachment, is a computer bus interface that was designed during a time when hard disks were dependent on spinning platters, and therefore had a relatively low physical limit to the amount of data they could transfer at any given time. When you look at computers and especially servers from the early 2000s and before, no matter how powerful your processor was, the most likely performance bottleneck was the hard drive. The SATA protocol was designed to make storage transfers using this relatively slower medium as fast as possible, but it was still designed for spinning media. Its most current revision, SATA 3.0, dates back to 2008 — well before SSDs were entering the mainstream.
But then SSDs came along. SSDs eliminated the need to seek individual sectors on a spinning platter and offered very fast data retrieval and transmission. They could handle sustained data transfer rates about an order of magnitude faster than spinning media, and thus SATA data transfer limits became reality. A new SSD interface was needed.
In came the NVMe protocol. NVMe, or non-volatile memory express, is first and foremost designed for speed. It eschewed the common command sets — SCSI and ATA — and instead uses its own commands and protocols to transfer data between a fast storage device like an SSD or a super-fast cache memory, and a host system over the PCIe peripheral bus. Speeds and capabilities depend on the specific drive and controller combination you select, but for a ballpark estimate of the performance differences, remember this rule of thumb: NVMe drives are, in general, about five times faster than a SATA-based SSD, and about 50 times faster than a SATA-based traditional hard disk drive (HDD).
NVMe was built in the era of fast solid state storage and doesn’t try to be backwards compatible; rather, it is a storage protocol designed for the future. According to IDC, SSD shipments will grow at a 15.1 percent compound annual growth rate through 2021. As SSDs continue to grow in ubiquity, NVMe will become increasingly important.
Pros and Cons: SATA vs. NVMe
Choosing the right storage interface bus can be the difference between getting the right solution for your workloads and being disappointed in performance or cost. Here’s a handy reference comparing the advantages and disadvantages of both SATA and NVMe:
Advantages of SATA
- There’s a long history of SATA support in hardware devices and form factors, so there’s a remarkable level of compatibility with a variety of hardware from the early 2000s to today.
- SATA devices are relatively inexpensive due to their broad availability and lower performance levels than spinning media.
- SATA is preferable for workloads where transfer speed is a secondary or tertiary concern, and low cost for larger capacities is the primary priority — for archiving data, storing little-used files and digital components, and keeping backup and recovery solutions.
Disadvantages of SATA
- It is the slowest storage transfer protocol of all and was designed for yesterday’s storage device era, not tomorrow’s.
- There are some M.2 devices that still use SATA connectors and command sets, and they perform much more slowly than their NVMe counterparts; the format can breed confusion when selecting devices.
- SATA and its related devices are poor choices for data-intensive, busy environments that depend on fast transfers and low latency.
Advantages of NVMe
- The time required to get data from storage medium to host processor is slashed, making NVMe a much better choice for data-intensive applications and workloads.
- The NVMe protocol has much lower latency, allowing for better performing sustained transfers and more consistent data delivery.
- The M.2 format used by a lot of NVMe devices allows for a great deal of capacity in a very small form factor, such as a thumb drive, making it perfect for systems requiring fast performance and lots of storage where physical space or perhaps total device weight is a limiting factor.
Disadvantages of NVMe
- NVMe depends on relatively high-performance SSDs that are more expensive than spinning drives on a capacity basis.
- For client PCs, NVMe generally depends on devices in the M.2 format, which limits the selection of drives compared to other solutions. (U.2 drives, which are 2.5″ form factor drives, are very relevant in the datacenter space.)
- There is not a lot of legacy NVMe support on older systems, making a storage system upgrade to an otherwise serviceable older machine unlikely.
- NVMe devices are not a cost-effective choice for storing large amounts of data.
Learn more about NVMe and how it can enhance the value of your SSD investments for enterprise data storage.