within Contiguous US
Enterprise-class Solid State Drives vs. Consumer-grade SSDs
Often in servers, popular retail drives such as the Samsung EVO are used because they are readily available and are priced at a lower $/GB. However, using consumer drives in servers is often limited to lab environments or non-critical use. Enterprise SSDs come standard with a set of technological features that are in line with other components such as memory, such as error-correcting code (ECC). In this article, we will briefly go over some of the main features found in many Enterprise SSDs which will allow you to determine if this class of drives is right for your use case.
End-to-end Data Protection
This feature extends error detection across the whole path between the storage and system. Utilizing error correction code (ECC) allows data integrity from many points ensuring the host doesn’t send out erroneous data. ECC will rectify errors thus enabling smooth transit of data from one point to another. This feature is important in servers because corrupted data can crash a system, similar to why ECC RAM is used in servers. Many servers are on the edge these days and communicate along many different paths, the more paths data has to travel the more important a feature like end-to-end data protection becomes. Having ECC on the storage itself allows better control as it offloads the host’s responsibility from verifying the data and thus skips extra steps.
eSSD manufacturers have built-in capacitors to power the SSD in the event of power loss. The capacitors have just enough power to flush the data from cache to be safely stored on the persistent NAND flash memory. Consumer hardware opts out of these capacitors because it costs more to implement and isn’t critical for the applications being used. For the enterprise, imagine a database server dynamically changing fields on the fly, a power loss can lead to data loss in this scenario. In comparison, a consumer application is able to restore from power loss with many applications available today, the database server might have no restore point as there’s just too much data being worked with and wasn’t actively backed up in that specific instance. Power-loss protection ensures that the data is still there and leaves no room for error.
Enterprise SSD manufacturers enable various options for encryption that’s part of the actual hardware and not reliant on only software. Crypto sanitize, also known as secure erase, is a command that overwrites the device with scrambled data effectively obfuscating the data previously written. Instant secure erase, or self-encrypting drive (SED), is similar but uses a hashed key that unlocks the written data. Deleting the key effectively bricks the data that’s been written on the device. Secure erase is a longer process as scrambled data has to be manually written, conversely with instant secure erase it is fast and useful for drive retirement in a pinch. Another level is FIPS 140-2, FIPS drives are compliant with requirements of the US government. Most encryption options for the enterprise is unique on hard disk drives, however it is standard on SSDs to usually have SED. Some consumer SSDs do require software utilities to actually use encryption however.
Enterprise-grade NAND Flash Memory
Overtime, writes to a solid state drive put wear on the NAND memory, eventually data cannot be written to an SSD as firmware locks the drive in read-only mode on most consumer SSDs. Enterprise SSDs are built to withstand more writes, exclaimed as TBW or total bytes written. In many cases TBW is used for warranty terms but in other cases it can mean as a rating for lifetime writes, often with 100% random write workloads which are more intensive. Most consumer and enterprise drives have vertical NAND technology (3D), which stacks flash cells to circumvent interference and increases storage density. Most SSDs use 3D (vertical) NAND flash technology now, with 3D TLC being the main type, with some popular consumer options using 3D QLC as well. 3D TLC NAND is 3 bits per cell while 3D QLC is 4-bits per cell, the more bits per cell there is, the less endurance the drive will have. There is still regular 2D or planar MLC SSDs on the market, although this is being phased out with 3D TLC as tiered storage and edge computing makes way.
Enterprise SSDs are over-provisioned moreso than consumer SSDs. Take for instance a drive with NAND chips equaling 1024 GB, a client consumer SSD will be 1TB while an enterprise SSD would be 960GB or even 800GB. NAND flash is also binned just like processors, the best of the best gets used in the enterprise-class drives and yields a higher rated TBW for it’s lifetime. Another metric besides total bytes written is drive writes per day (DWPD), DWPD is the amount of times a drive can be written over and is derived from TBW. The DWPD is often used in warranty terms just like TBW, because even after 5 years somebody could put little wear by writing only a dozen or so TB of data. A warranty term rating would look like 600TB TBW over 5 years with 0.3 DWPD, when really the actual TBW would be a more exact number and the DWPD still in generic terms. Enterprise SSDs are grouped into 3 categories, read intensive, mixed use, and write intensive. Read-intensive drives are rated for typically 1 DWPD, with 3DWPD and 10DWPD for mixed-used and write-intensive respectively. Write intensive drives does not mean it’s only for writes, it just means it can handle a lot of writes! The trend for eSSD manufacturing is read-intensive and sparingly mixed-use devices as tiered storage and edge computing is becoming the norm instead of a consolidated data center due to bandwidth restrictions in today’s age. Regardless of the type of solid state drive, know that data retention when powered off is not unlimited, it may last 3 months or it may last a year. RAM has high endurance because it can leak it's charge, SSDs when powered off lose it's charge over time and should not be used for long-term cold storage.
Interface & Form Factor
Many enterprise solid state drives continue the lineage of hard drives by using the SAS interface along with the 2.5” SFF. Even the latest generation of PCI-Express drives in the enterprise use a 2.5” form factor because of the continuance of hot-plug front facing backplanes, along with the ease of combining the backplane with SAS/SATA/NVMe devices all together. Compared to M.2 drives, U.2 (Known as 2.5” for NVMe) also has better thermals and is serviced overall just like the standard 2.5” HDD. The consumer world now regularly uses NVMe or PCIe drives but the enterprise has options to use SAS still especially due to PCIe lane restrictions and in many cases cost. One of the biggest factors when it comes to enterprise drives is that many come with dual-port functionality. This allows no need of interposers and a built-in way to connect multiple paths for connecting to multiple systems at once, these are used in SAN environments (storage area network). There are many form factors when it comes to solid state drives, overall more compact form factors will be seen overtime with both consumer and enterprise SSDs. The interface used in the enterprise is moreso SATA and SAS with the newest generation of PCIe Gen4 SSDs being seen in consumer areas.
Looking from data sheets, one aspect to notice is that write speeds are mostly slower in comparison to consumer SSDs. At the firmware level, writes are tuned to enhance response times as greater writes cause more latency, just like RAM where latency increases the faster it goes. Consumer SSDs have read/write speeds shown that will depict burst speeds but once the cache buffer fills up or really any significant amount of data is transferred, the rated speeds drop off heavily. Enterprise SSDs will show sustained sequential and random operations, with enterprise applications it is expected for availability to be 24/7. Sustained transfers are important for many enterprise use cases and consumer SSDs can drop off and actually be slower than hard drives after the cache is filled, this is especially seen in 3D QLC drives as those often have limited cache buffer sizes and 4-bit NAND cells cannot withstand as much electrical interference.
Controller and Caching
The controller on an SSD controls how data is managed, enterprise SSDs will maintain consistency in terms of temperature, latency, and transfers. The controller has firmware that communicates with all the parts on a solid state drive to operate in a specific manner. The controller and cache both play a pivotal role when many drives are in parity and communicate between multiple users and systems. The cache on eSSD devices will almost always be DRAM memory, which is lower latency and more reliable than SLC NAND. SLC NAND (1-bit cell) is used in some consumer SSDs in place of DRAM in some cases, often times the cache will be of smaller storage size too. Some consumer SSDs on the market today won’t even have a built-in cache, instead it uses the host system’s RAM as a swap location, this saves on expenses and can result in higher latencies but overall is sufficient for many client system’s use cases.
All in all, it depends on your use case if enterprise SSDs are right for you. If you have multiple systems and many users all connecting to the storage all at once, then the benefits really arise. However, more often than not, the enterprise SSD options will cost a great deal more than comparable capacity options of consumer drives. The cost difference for better reliability and higher quality components would not be so apparent with a high price disparity, there's plenty of high quality consumer drives out on the market today that would get the job done. In my humble opinion, if you're consistently writing data each day and it's at least a third of the capacity of the device, then it's a good time to consider using enterprise drives, especially if the drives are in parity with one another. Otherwise, save yourself the money, you could always upgrade later.
If you want to view some of the different models out there, check out our catalog: