Enterprise users are facing some tough choices right now as solid state drives (SSDs) begin to proliferate in the datacenter. Not only must they sort through the performance benefits and documented drawbacks of typical SSDs, they must also determine which SSDs are suitable for use in mission critical applications. But as they do, new evidence is emerging that an SSD’s classification as “enterprise ready” is not determined by an SSD’s use of “MLC” or “SLC” but rather if it possesses the ability to detect and correct soft errors as they occur.
Historical “Consumer” versus “Enterprise” SSD Classification
According to some estimates, there are well over 100 OEMs of SSD storage systems on the market today with SSDs finding its way into more storage systems that typically receive the “enterprise” label. As they do, an aging criteria that differentiated between “consumer” and “enterprise” grade SSDs is based on whether or not the SSD is a multi level cell (MLC) or single level cell (SLC) design.
MLC SSDs are most often associated with “consumer” grade. The knock on MLC SSDs is that they lack the performance and reliability of SLC SSDs. SLC SSDs experience wear-out after approximately 100,000 program/erase cycles while MLC SSDs experience wear-out after 10,000 or less cycles. The upside of MLC SSDs is that they can reach higher capacities (over 200 GB) and are available at a lower price point because so many manufacturers are starting to produce them for desktops and laptops.
Conversely, SLC SSDs are the ones finding their way into enterprise storage systems. While the storage capacity of SLC SSDs is lower and they have a higher price than MLC SSDs, they support faster write speeds and have higher cell endurance so they have lower error rates. Since SLC SSDs do a better job of mitigating the possibility of data corruption, they are finding their way into the enterprise space since, at this level, data integrity arguably is more important than performance.
Flash is Unreliable
However at their core both MLC and SLC SSDs are flash drives, not hard disk drives (HDDs). SSD manufacturers have taken steps to make SSDs appear as HDDs to operating systems and applications but this action of making a flash drive look like an HDD is where the potential – and likelihood – for soft errors (an error in a computer’s memory system that changes an instruction in a program or data value) to occur.
To make flash look like block storage requires a tremendous amount of processing power. Embedded microprocessors and DRAM are placed on the individual SSD devices that handle the task of making the SSD device appear as an HDD to the operating system and applications by handling the exchange of data between the application and the underlying SSD media.
While manufacturers are taking the appropriate steps to protect data once it is stored on the SSD to ensure that the data is preserved and does not become corrupted over time, this newly created process of handing off the data and retrieving it from the flash media is when soft errors can and do occur.
The “Unexamined” SSD Data Integrity Risk
This new compute process within the SSD devices that makes flash appear as an HDD is creating different data integrity risks that, to date, have gone largely unexamined as to the impact that it can have on data.
In brief, the specific problem that these new processes create is that it is still uncommon for most SSD devices to put enough sophistication into this process to even detect errors should they occur, much less do error correction. Because they are not looking for soft errors much less taking any steps to correct them should they occur, there is a small but distinct possibility that during the metadata handling that occurs with the processor instruction stream, something could go awry and the wrong piece of data could be stored or retrieved from the underlying flash media.
This soft error could later show up in financial applications displaying inaccurate financial information or healthcare software retrieving inaccurate information about a patient. The danger here is that because there is no error checking going on during the conversion process, invalid data could be presented as valid data due to the lack of data integrity checking that is occurring at this point in the process.
The True “Consumer” versus “Enterprise” SSD Differentiator
These anomalies in system behavior – and how they are addressed – are exactly what differentiates consumer from enterprise SSD solutions. In enterprise environments SSDs are more likely to be used for applications with high transaction rates. This increases the likelihood of this soft error anomaly occurring from “improbable” to “highly likely” so these applications demand SSD solutions that deliver performance while also taking these data integrity issues into account.
The increased frequency of these errors was confirmed in a recent study completed by the University of Toronto. The researchers found that DRAM error rates are “orders of magnitude higher than previously reported, with 25,000 to 70,000 errors per billion device hours per Mbit and more than 8% of DIMMs affected by errors per year.” While the researchers found that memory errors are dominated by hard rather than soft errors, any system using memory without support for error correction and detection can lead to a memory error resulting in a machine crash or applications using corrupted data.
Because of the possibility that soft errors can occur, enterprises need SSD solutions that account for this behavior by detecting and correcting soft errors. Indeed, this ability to detect and prevent such anomalies from occurring is becoming the true differentiator between consumer and enterprise grade SSD solutions.
Fusion-io Takes the Appropriate Steps to Detect and Correct Soft Errors in SSDs
Enterprise SSD solutions will take the appropriate the steps to protect both the memory and CPU instruction stream from being corrupted prior to the data being stored to the SSD and then when it is retrieved. So any enterprise organization that is looking at SSD and has concerns about the integrity of its data stored on SSD needs a solution that mitigates the possibility of soft errors ever occurring.
Enterprise organizations are right to have a fair amount of skepticism about the appropriate role that SSDs should play in their data centers in light of their high cost and documented problems. This latest issue of how soft errors can occur within the SSD device itself is only one more reason that enterprise users should remain skeptical.
However the emergence of SSD solutions such as the Fusion-io ioDrive is one such product that takes into account the possibility of the occurrence of these soft errors and takes steps to prevent them from occurring. In an upcoming blog, I will take a look at two different techniques that currently exist for addressing this issue and why Fusion-io has selected the approach it has for addressing soft errors.