Understanding the technology behind the technology
(It’s not about the flash—it's about the flash controller.)
Flash is great. Flash is good. Flash gets all the press.
When talking about solid-state drives (SSDs), “flash” serves as both storage media and an important part of the drive—but what’s often overlooked is the controller inside the drive.
In the realm of SSDs, the controller is extremely complex to design and build, but it’s practically given away, with the NAND becoming everyone’s focus when being sold on each drive.
The controller, however, is the brains of the SSD solution. So when looking for a difference in reliability, performance, endurance, capability or security, bypass the flash—and look at the controller on the drive.
The controller communicates with the host computer, moves data into and out of the flash, and handles all the flash management tasks such as wear leveling, error correction, and data reliability activities. The latter includes—if designed for the enterprise—full data-path protection and features like STEC’s proprietary CellCare™ Technology to extend the endurance of the MLC media from the inside of the flash chip.
SSD selection for applications and use is very dependent on the design of the controller, and consumer SSDs come with a certain set of requirements that mostly focus on extending the longest battery life. Enterprise SSDs have controllers that are much more involved, and intended to ensure the highest capability of drives running 24/7/365 under extreme conditions with demanding workloads and duty cycles.
Author Information
Scott Stetzer
Vice President
Technical Marketing
Two types of media today: SLC and MLC
Today, flash media technology comes in two categories: single-layer cell (SLC) and multi-layer cell (MLC) media.
SLC media, in the simplest of technical terms, means that for each “cell” used to store data, a single bit is stored in that cell, with a single layer in the cell media. As the technology evolved, engineers and developers improved design and development methods to enable more than two bits to be stored in each cell, creating multiple layers in each cell (MLC) media. Work in progress to store three bits per cell media is being developed and tested.
Without a doubt, MLC media signals a tremendous improvement in capacity for flash and solid-state drive (SSD) devices. This technology also gives a cost advantage in delivering twice the capacity for the same price of the chip.
In terms of disadvantages, MLC media writes slower and offers less endurance than SLC media because it takes time to layer more than one bit of data into each cell. Higher current is needed to erase, and more current is needed for the longer times required to write multiple layers of data into each cell. Both result in increased wear on the NAND media.
These endurance challenges aren’t considered critical within typical consumer mobile applications; for example, the write duty cycles are typically very low in laptop computers. Such challenges become more pronounced in enterprise server or storage systems where an unmanaged MLC drive can wear out in less than one year without the deployment of specialized technology designed to extend the life of the media.
STEC proudly pioneered the use of MLC media within enterprise applications with the August 2010 introduction of revolutionary CellCare™ and (Secure Array of Flash Elements™ (SAFE) technologies, which are employed to extend the life of MLC media through active flash cell management. These technologies have been proven to extend the life of storage media—and, in turn, enable it to meet the demanding enterprise requirements for duty cycle and life.
When considering MLC media in enterprise environments, CellCare is the only technology proven to deliver performance, endurance and reliability.
Author Information
Scott Stetzer
Vice President
Technical Marketing
Stored Data is Only Getting Bigger
The analysts at IDC estimate that 1.8 zettabytes of data are now being created or replicated each year. So, when it comes to data handling, shrinking latency is mission-critical. Remember it’s not about how many records you can access, but rather, when a host asks for the record, how fast it comes back.
This massive task is only getting bigger, and this chart from Gizmodo puts the sheer size of this data into perspective and illustrates why accelerated access to data is increasingly important.
Author Information
Michael Bingham
Director
Corporate Marketing & Communications
Our Customers Are Always Right
A few weeks ago we got the opportunity to support a customer serving the United States Army. They had purchased SSDs from us almost six years earlier but the slow pace of government acquisitions had just now brought their design to production levels. The drive warranties expired two years ago and tight programmatic constraints prevented the Army from upgrading to current technology. With little to no upside potential, it made justifiable financial sense to point to the expired warranty and end-of-life notices rather than provide the requested support.
Our actual response however, was to fly out our Senior Field Application Engineer along with our Director of Defense Programs to assist in troubleshooting and integration. We analyzed the old drives and gave their engineers tools that would help them maintain the drives until the Army’s next technology insertion opportunity. We also spoke directly with Army officers to both educate them and ease their apprehension about any programmatic risk.
Rather than picking up a purchase order, this visit helped us develop tremendous insight into the demands facing our defense customers and ultimately our soldiers. It also helped us develop a level of trust and confidence that cannot be won over the phone or through email. I have no doubt that these engineers and program managers will remember who helped them when they needed it most.
No matter how small, each customer deserves our best and STEC is fortunate to have a wealth of engineering expertise that we can move to support seemingly insurmountable problems.
Author Information
Mark Flournoy, LtCol USMC (ret)
Vice President
Government and Defense
It Takes a World Class Manufacturing Facility to Build a World Class SSD!
Any MIS Director will tell you that the data that is stored on an SSD is more valuable than the SSD itself. Reliable access to error-free data is critical for any modern organization. At STEC, we take extra care to ensure that data errors do not occur on our SSDs.
One area to which STEC pays very close attention is how the SSDs are built. STEC employs a holistic approach that seamlessly integrates “in-house manufacturing” of our SSDs with the management of the entire product life-cycle from new product launches though product ramps and all the way to “end-of-life”.
To achieve this, STEC had started laying plans to build a “green-field” manufacturing facility that was optimized for SSD production back in 2007. By January 2008 we opened our brand new, state-of-the-art facility in Penang, Malaysia. This 210,000 square foot (19,500 square meters) facility incorporates good manufacturing practices and workforce empowerment, as the fundaments to achieving consistent and predictable output from the applied process and test technologies. The aim is to minimize “waste” and maximize both quality and timely output in the production of SSDs within a “build to order” environment.
Many of the manufacturing techniques deployed by STEC, will be familiar to those who have a keen interest in manufacturing excellence. These techniques have been customized by us to specifically address the challenges of manufacturing SSDs. Although there is not enough room in this blog to describe each of these techniques in detail, here is a brief summary of the key processes we use to manufacture a world class SSD:
• Lean Six- Sigma manufacturing and Kaizen (continuous improvement) mentality which focus on driving for higher efficiencies and effectiveness whilst eliminating various forms of “wastes” and minimize any potential “variability”
• Demand driven supply chain practices such as VMI (vendor managed inventory) and Kanban (just-in-time) replenishment of raw components
• Six Sigma quality methods to quantify our manufacturing processes and enable our drive for continuous improvement
• Qualification and refreshers with various quality and regulatory standards, e.g. ISO-9001, ISO-14001, OSHAS 18001: 2007, to ensure ongoing compliance to industry standards
• Streamlined process controls and monitor, using IT enabled workflows and shop-floor control systems
• Full traceability systems to track the purchasing and production history in great detail of each SSD built and every component used as it is manufactured.
• Automated optical inspection systems that use machine vision to ensure that no problems occur during the surface mount assembly process
• Application of “SMART” tools and fixtures to POKA YOKE (mistake proof) potential mistakes and reduce process variability.
In summary, there are subtle but yet significant differences between a general –purpose SMT/ manufacturing facility that is engaged to build SSDs and a manufacturing facility that is optimized for SSD production. Key OEM customers and independent observers who have visited or audited our Penang facility have given feedback leading us to believe that in we are progressing well along the path to being a world class manufacturer of SSDs.
From recruiting experienced manufacturing and engineering personnel, to developing processes, procedures and practices along with investing in state-of-the- art SMT, assembly and test equipment, tools and information systems, we have established a strong foundation to scale the challenges that are unique to SSD production. We truly believe the “fruits of our labor and investment” will pay back in predictable manufacturing yields as well as consistent high product quality and reliability.
At STEC, we are proud to acknowledge our Penang facility as one of the most advance showcases for manufacturing world-class SSDs. For those of you who are located or traveling in Asia, we invite you to visit our facility to get a first-hand glimpse of how our SSDs are made.
Author Information
Doug Finke
Senior Director
Product Marketing
Lipstick on a Pig
Nothing pleases a soldier more than getting his hands on a new piece of “gear,” such as the latest set of night vision goggles or ruggedized laptop for example. The sinister green glow of the goggles or the resilience of a water-proof keyboard introduces a certain cool-factor; and cool sells. If it works right out of the box, that’s all the better. Unfortunately, unscrupulous companies often pack their new products with substandard components. Essentially putting lipstick on a pig and putting our troops at risk.
Sure, advertisements that claim their systems can be dropped from a 3-story building, tolerate boiling hot temperatures and a range of other inhospitable conditions, are points that a good marketing campaign promotes. But unless these parts meet stringent industrial temperature, shock and vibration standards, the realities of combat will always expose the truth.
The solid-state drive (SSD) in that cool desert colored laptop must contain a sophisticated controller with adaptive flash management algorithms to actively manage NAND wear leveling; it should have error management and tuning functions that adjust throughout the life of the flash; as well as advanced memory management algorithms to improve flash endurance; and parity inside the drive that allows data to be recovered, rebuilt and accurately returned in cases of a read error due to bit, word line or even total die failure. You can’t get this kind of performance from SSDs bought at a consumer electronics store, but incredibly, some manufactures are doing just that.
One of the best ways to ensure our soldiers get the quality they deserve is to ask what’s under the hood. Sure, any new computer is fast right out of the box, but while under the constant punishment of combat conditions and without industrial-rated SSDs, they’re sure to fail when you need them most.
The next time you’re at a military expo, ask the guy behind the booth what kind of SSDs they use in their ruggedized computers. If he hesitates and changes the topic, compliment him on his lipstick.
Author Information
Joe Siedlarz
Senior Director
Government and Defense
Mark Flournoy, LtCol USMC (ret)
Vice President
Government and Defense
SSD Myths and Facts
I frequently answer questions and address myths from end users and data center managers about implementing SSD technology, such as:
• SSDs are too expensive.
• All SSDs are the same aren’t they?
• Do SSDs wear out when you write to them?
• Why do SSDs slow over time?
Here are a few of my thoughts on these topics.
Q. SSDs are too expensive.
If time is money, then the superior SSD IO performance is the clear winner. For example, even in the best case scenario an HDD IO rate will be cap around 300 IOPS, whereas an enterprise class SSD from STEC can deliver as many as 70,000 to 80,000 IOPS in the same working conditions.
While SSDs are more expensive than HDDs on a per-Gigabyte basis, thoughtful consideration should be predicated upon performance, not capacity. If the priority is mass storage, and access speed aren’t critical, HDDs hold a value-advantage based solely on cost. But, if you’re running applications where accelerated data access is vital SSDs will quickly pay for themselves.
Remember, the value of an SSD should be measured by dollars-per-IO not dollars-per-Gigabyte ($/IO vs. $/GB).
Q. All SSDs are the same aren’t they?
As there are different types of HDDs (mobile, nearline and 15K), there are different SSD designs (mobile, industrial, enterprise), with each variant serving a distinct purpose. For mobile SSDs, lower cost and power consumption are driving factors, whereas enterprise SSDs focus on endurance, data reliability and sustainable performance.
There isn’t a data center manager in the world that would place a laptop HDD into their mission-critical data center server or SAN environment, would you? If not, why would you ever consider deploying a consumer-grade SSD?
Q. Do SSDs wear out when you write to them?
It’s true… SSDs wear when they’re written to… but this is a good thing because it’s measurable! Drive life can be closely monitored in order to prepare for and predict maintenance. Even though SSDs wear with use, under normal heavy use cases, the life of an enterprise SSD can easily last well beyond the warranty period, depending on capacity and duty cycles.
STEC solutions such as CellCare™ Technology are built into our proprietary controllers to minimize the effect of wear out and extend drive life to the maximum extent possible. Even in worst case conditions, wear is managed to meet and exceed the warranty life of the drive.
Q. Why do SSDs slow down over time?
As consumer-class drives fill with data they slow down as the wear leveling and scatter/gather algorithms kick in to find space for the next write operation. This slow-down phenomenon is well know and understood, as discussed in multiple website and blogs measuring SSD performance have demonstrated.
Enterprise SSDs from STEC have proven, mature and advanced technology with years of firmware development built into their designs, and as a result much of these issues with slow down are intelligently managed by the STEC controller and firmware, resulting in far better steady state performance over time that’s free from the excessive slowing down experienced by other vendors.
Author Information
Scott Stetzer
Vice President
Technical Marketing
Hard Drives and Earthquakes and Hurricanes, Oh My!
Last week, while attending a conference in Tampa, Florida, I received a short phone text from my wife that read “5.8 on the r scale”. It took a few seconds to sink in that she was referring to an earthquake that hit near our home in Virginia. People around me quickly started receiving similar texts and calls. From past experience I have to believe an equal number of calls went out to data center managers throughout the affected areas from supervisors and CEOs checking the status of their networks and business-critical data systems. As if the quake wasn’t enough, Mother Nature topped off the week with Hurricane Irene reminding business and government leaders of the physical vulnerabilities of their data centers.
While computer hacking typically gets top billing when it comes to bringing down a network, shock, heat and other environmental factors are equally effective. During my last two tours of military service in Florida, Hurricanes Ivan, Katrina, and Wilma also posed considerable risks to our server farms. As a result, we spent long days in preparation of each hurricane developing contingency plans and assigning ride-out teams to remain on base in an effort to minimize the impact of the storm. The loss of just one cooling unit could cause an entire data center’s temperature to spike above 90 degrees requiring planned shutdowns to prevent system crashes. Our military missions depend on 24/7 access to information, so a data center outage in the U.S. could seriously degrade combat operations in Iraq or Afghanistan.
The risk of natural disasters makes redundant power systems and alternate backup locations necessary considerations when building data centers. Additional assurances can be gained by upgrading your servers with solid-state drives (SSDs). Their indifference to vibration and temperature are obvious but you also gain significantly higher rack space utilization, which in turn, lowers power consumption and associated cooling costs. In comparison to hard disk drives (HDDs), just one STEC ZeusIOPS® SSD uses less than half the power of 15 HDDs and provides data persistence and instant backup and recovery in the event of an unplanned power outage. If cooling failures are a concern, STEC MACH16™ SSDs are available in an industrial temperature variant that performs consistently at temperatures up to 185 degrees Fahrenheit.
Disaster planning doesn’t mean that earthquakes and hurricanes won’t occur, but it will lessen their impact on your operations. A tape backup solution may get you through an outage, but if you’re like me, this option will probably lead to shorter fingernails and anxious nights. SSDs are a much more reliable sleep aid.
Author Information
Mark Flournoy, LtCol USMC (ret)
Vice President
Government and Defense
Deploying Enterprise SSDs in Servers
Let’s face it, for servers, SSDs are viewed simply as expensive HDDs.
A glaring reason for this is that server vendors aren’t delivering hardware + software integrated solutions; they are generally only supplying the hardware. Most data centers rely on installing their own software onto the new hardware they have just purchased. The server vendor hardware costs are the measure of value, thus the costs of adding SSDs are seen as expensive in comparison to an HDD.
The value of server deployments for SSDs are realized within the system integrator. This is where the application is added to the hardware and the point in which the measurable benefit of SSD to the system is realized through the implementation of this new performance paradigm.
The SSD value proposition is not based on capacity. Users don’t buy an SSD for capacity, but rather, for speed. This means that server deployments using SSDs have to consider the overall application or use model. The first questions to ask are:
• Does this server do lots of IO?
• Must it boot fast?
• Will it manage lots of IO on or off of a SAN storage system?
• Is it used for business intelligence filtering and reporting?
If you answer yes to any of these questions then I suggest that SSD is in your future.
For boot servers: SSDs with 32-50GB of capacity make excellent boot devices with a significantly lower TCO than an HDD deployment, not only from boot time considerations but from the maintenance aspect as well. SSDs have no moving parts, thus are significantly more rugged than the rotating mechanics of a traditional HDD.
For virtual machines: SSDs make excellent mechanisms for improving VM performance giving the ability to boot multiple images much faster than traditional storage.
For OLTP and Databases: SSDs are clearly superior for transactional performance, accessing database records without the normal latencies associated with mechanical seeking to reach the data.
Email systems: have been big winners with SSD installations improving the performance of email servers and their ability to respond to user mail clients.
The fact is that almost any server can be improved with the installation of an SSD in the box and used as a caching store between the server and the SAN or NAS storage on the fabric (e.g., FC, iSCSI, FCoE, etc).
Author Information
Scott Stetzer
Vice President
Technical Marketing
I wish I had a Solid State Drive in Iraq
During my second tour in Iraq, we had set up a data network supporting more than 12,000 Marines spread across an area the size of North Carolina. A combination of satellite and microwave systems linked subordinate units to higher echelons, which in turn linked back to the United States. During the first few weeks, travel was easy and U.S. forces had the freedom of movement throughout the area of operations. If a network problem occurred, it was a simple fix to put a four man repair team on the road. Distances between Forward Operating Bases, or FOBs, were no less than 30 miles away and some units were a few hundred miles from the next closest base. A repair team would spend a day on site and then return once the job was complete.
Within a month, we saw an uptick of Improvised Explosive Devices (IEDs) and mortar activity. At first we thought these were isolated incidents, and then it became clear there was a calculated effort to deny U.S. forces freedom of movement along roads and bridges. As you can imagine, traveling between Forward Operating Bases (FOBs) became a significant event with a requirement for no less than three armored vehicles with four or more Marines per vehicle. Travel was slow and deliberate leaving many units to either wait several days for a communication repair team to arrive or take matters into their own hands and attempt repair of complex and fragile computer systems.
This loss of movement was exacerbated by heat and dust. The dust in Iraq is extremely fine grained, and easily passes through the dust screens on our Exchange and data servers. Canned air was a Server Administrator’s best friend, blowing out dust and hopefully preventing head crashes caused by sand grain hitting the hard disk drives’ (HDD) spinning platters. While initially effective, we later found that each blast of canned air mixed with the fine dust simply sand blasted the HDD platters.
Drives started failing and eventually resulted in a loss of over three months of classified emails for more than 100 users at a remote FOB some 40 miles from our main camp. A repair team was launched to fix the problem and while en route encountered an IED resulting in minor wounds to three Marines and a totaled vehicle, but fortunately no loss of life. After the incident, the Marines continued on and eventually replaced the Exchange Server and gingerly wrapped it in cheese cloth to mitigate future dustings.
More than six years later, this scenario repeats at least weekly as military units in Iraq and Afghanistan continue to rely upon spinning HDDs. A simple drop-in replacement of a solid-state drive (SSD) and these servers would be able to withstand high shock and temperature swings between well below freezing and up to 185 degrees Fahrenheit. Most importantly, the lack of moving parts in SSDs would eliminate the dust and dirt related failures experienced by HDDs. These are more than just features of convenience; implementing SSDs in place of traditional HDDs will reduce system crashes and, in-turn, reduce additional risk to servicemen and women who are already in harm’s way.
Author Information
Mark Flournoy, LtCol USMC (ret)
Vice President
Government and Defense