There’s been a lot of talk in the storage industry recently about solid-state disks (SSDs) and their ability to dramatically speed up a computing environment’s performance. One thing that is rarely discussed is the difference between different types of solid-state technology.
There are two basic types of memory used by SSDs: DRAM and flash memory. A general perception in the computing industry is that only DRAM is robust enough for enterprise use. That sentiment doesn’t give enough credit to flash memory. As with any two different technologies, each has its advantages. This paper explains the differences in effort to help determine which technology is best suited for one’s IT environment.
It’s common knowledge that writing to flash memory is much slower than writing to DRAM. Isn’t speed the main purpose of installing an SSD in the first place? Yes. However, to say that flash technology is “slower than DRAM” is to sell it short. First, reading data from flash memory is very similar to the speed of reading from DRAM. Second, the better manufacturers of flash SSDs incorporate a DRAM cache in the drives to speed up writes. The best of those manufacturers have algorithms inside the devices which are able to flush that data from cache to flash in the background without impacting performance. If we graph the relative performance of the two types of SSD and a traditional rotating disk, it looks like:
This graph is close to scale- one pixel width represents 10 microseconds (Âµs). Typical access times are: DRAM SSD: 10-50Âµs, Flash SSD: 35-100Âµs, Rotating Disk: 5000-10000Âµs (5-10ms). We can see from the above graph that the DRAM-based SSD is indeed faster than the flash-based SSD; it may even be three times as fast. However, we must ask the question: “Is that performance difference significant?” Considering how much faster each SSD technology is than rotating disk, the answer may well be “No.” Chances are good that other differentiators below will be more important to many IT environments.
Unlike DRAM, flash memory chips have a limited lifespan. Further, different flash chips have a different number of write cycles before errors start to occur. Flash chips with 300,000 write cycles are common, and currently the best flash chips are rated at 1,000,000 write cycles per block (with 8,000 blocks per chip). Now, just because a flash chip has a given write cycle rating, it doesn’t mean that the chip will self-destruct as soon as that threshold is reached. It means that a flash chip with a 1 million Erase/Write endurance threshold limit will have only 0.02 percent of the sample population turn into a bad block when the write threshold is reached for that block. The better flash SSD manufacturers have two ways to increase the longevity of the drives: First, a “balancing” algorithm is used. This monitors how many times each disk block has been written. This will greatly extend the life of the drive. The better manufacturers have “wear-leveling” algorithms that balance the data intelligently, avoiding both exacerbating the wearing of the blocks and “thrashing” of the disk: When a given block has been written above a certain percentage threshold, the SSD will (in the background, avoiding performance decreases) swap the data in that block with the data in a block that has exhibited a “read-only-like” characteristic. Second, should bad blocks occur, they are mapped out as they would be on a rotating disk. With usage patterns of writing gigabytes per day, each flash-based SSD should last hundreds of years, depending on capacity. If it has a DRAM cache, it’ll last even longer.
Most flash SSD makers employ error-checking algorithms and are able to correct a few bytes in a 512-byte block. Some of the less-robust error-checking will miscorrect three byte errors about 20% of the time. The best flash SSD providers can correct six random byte errors (and detect nine) in a 512-byte block. They will also never miscorrect a three-byte error. This level of error-checking gives security that the data integrity of the drive will last much longer than we as IT professionals will have to worry about it.
Unlike DRAM, flash is inherently non-volatile. There’s an old axiom which states that “a computer’s attention span is only as long as its power cord.” This definitely holds true for DRAM as well. While flash memory will retain its data beyond 10 years without power, little more than 10 milliseconds without power will give DRAM a most annoying case of amnesia. To prevent this, DRAM-based SSD makers must add batteries and disks to keep the data from being lost during a power failure. Though rechargeable, these batteries must be maintained (replaced) on a regular basis (maintenance cycles vary; consult the SSD’s manufacturer) to ensure their ability to completely backup the data in the SSD. The batteries maintain power to the memory and disk(s) long enough to transfer the data from DRAM to the non-volatile storage. Two things to consider are: Some power failures happen in rapid succession- This may cause the backup operation of the SSD to start over, which essentially drains the batteries prematurely. This may mean that the batteries will not retain enough power to complete a backup cycle. Second, backup and restoration of the data takes time. It can take 30 to 60 minutes or more to backup and restore the data. The backup time usually isn’t painful, but the restoration can cause extended downtime. Consider the scenario of a power failure and successful data backup to disk. When power returns, the server(s) can be up and ready long before the SSD’s data is restored from its backup disk. This can mean that the server will be unavailable for an extra hour or so. Depending on the application, this could range from a mere annoyance to a business-threatening outage.
Most DRAM-based SSDs are large, rack-mount devices. They require large internal power supplies, fans, batteries and disk drives to provide non-volatility. In comparison Flash-based SSDs are much smaller, usually the same form factor as a conventional disk.
Because the form factor of flash-based SSDs is so much smaller, they are inherently more flexible in their use. They can often be used in place of traditional disks in storage arrays or in a server’s internal disk bays. Embedded applications or mobile systems often require the much smaller footprint of a flash-based SSD.
Both types of SSD are quite reliable since there are few, if any, moving parts. Even the backup disks of the DRAM-based SSDs are typically spun down during normal operation. This means that both types of SSDs are much more reliable than a traditional disk. However, for more demanding environments the smaller, more rigid flash-based SSDs are often more desirable. They typically withstand greater vibration and temperature ranges than DRAM-based SSDs. Some flash-based SSDs are even considered “ruggedized” by NASA and the U.S. Military. These drives will withstand intense extremes that would reduce a rack-mount box to rubble.
Power Consumption/Heat Dissipation
One benefit of flash memory is that is uses much less power than DRAM chips. Because of this, flash-based SSDs generate much less heat than their DRAM counterparts. This also means that they don’t need cooling fans, whereas the DRAM-based SSDs do. Again, fans take space and require power themselves, which in turn generates heat and noise.
No IT department would purchase a solution without looking at its price tag. While DRAM chips and flash memory chips are similar in price, the overall cost per megabyte is generally lower for flash-based SSDs. This is due to the simpler design, and the lack of need for backup batteries and disks, and the enclosures in which to hold them. Some of the cost of the DRAM-based SSDs is the extra sheet metal for holding the batteries and disks, as well as the labor involved in assembling it all.
Now that we’ve taken an in-depth look at the different functionality and features, it should be clear to see that flash memory is useful for more than just consumer devices- it’s also well-suited for the enterprise. Of course, one type of SSD will likely be better suited to the needs of your applications and IT environment than the other. Clearly though, it can be seen that flash memory has quite a list of capabilities that make it a “superstar” technology for many IT organizations.
Kelly Cash is the Technical Evangelist for BiTMICRO Networks, a leading supplier of flash-based solid-state disk storage solutions. He has been in the systems performance arena over 15 years, specializing in server and storage optimization. During his career he’s been an engineer focusing on performance for such companies as Sun Microsystems, Data General, and AIM Technology. He’s also held positions in the IT space, such as the Chief UNIX IT Technologist for Cadence Design Systems.