Engadget Primed: SSDs and also you

Primed goes in-depth at the technobabble you hear on Engadget each day — we dig deep into each topic’s history and the way it benefits our lives. Trying to suggest a bit of technology for us to collapse? Drop us a line at primed *at* engadget *dawt* com.

Table of Contents

Initially…
The long dominance of magnetic-drive storage
a short history of SSDs
Early consumer drives and a maturing technology
Why SSDs?
The challenges of SSDs
Not all are created equal
0 Wrap-up 0

To higher understand where of SSDs in today’s storage landscape, it’s worth recounting some history. Cast your mind back to a time when computers weighed tons and were delivered by forklifts. One such contraption, the 305 IBM RAMAC, debuted in September 1956. That sometimes cuddly acronym stood for “Random Access Approach to Accounting and Control,” and massive Blue’s system leased for $3,200 a month. For which you got a console, processing unit, printer, card punch and large power supply — all delivered by cargo plane, so long as you had the 30- by 50-foot air-conditioned room had to house it.

Most crucial to our story, though, RAMAC shipped with the IBM 350 Disk Storage Unit. Back then, drives didn’t need fearsome names like VelociRaptor and Scorpio to differentiate themselves; really, IBM’s was the primary hard drive, marking a revolutionary moment in computer science.

So what was this mechanical marvel? Reminiscent of two contemporary technologies, tape and drum storage, it trusted a moving, magnetically charged medium: 50 aluminum disks, or platters, each 24 inches in diameter. Stacked in a cylinder, they spun at 1,200RPMs while a couple of read heads moved vertically to the suitable platter, then horizontally to the proper track. IBM saw this random access capability because the system’s greatest selling point.

Here’s the way it worked: imagine a stack of fifty vinyl records, each separated by an area thin enough for a phonograph needle to pass between them. (If records are as foreign to you as papyrus scrolls, you’re able to substitute CDs.) To listen to a specific song, you simply want to find the best record and right track; you do not, as with a tape, should fast forward or rewind through all those unrelated songs. Random access, the facility to start reading from any point at the medium, dramatically reduces the time it takes to discover data; the seek time at the 350 was about 600 milliseconds.

The 350 stored about 4.4MB. The tale goes that it is able to have held more — in any case, you can always add more platters — however the marketing department couldn’t determine methods to sell to any extent further MBs, thereby beginning the long tradition of “it is all the gap you’ll ever need!” Having said that, it soon came with an optional second drive. For the following twenty years access times and capacity continued to enhance, and in 1973 Big Blue introduced a more recognizable precursor to trendy harddisk drives (HDDs). The IBM 3348 Data Module was a sealed cartridge containing the platters, spindle and head-arm assembly. The 1970′s version of removable storage, it came in 35MB and 70MB versions.

The magnetic platter concept pioneered and refined by IBM laid the groundwork for many years of fast, cheap and reliable data storage. Honed and miniaturized, it is the same basic technology present in HDDs world wide today.

IBM continued apace, increasing the scale and speed of its drives. In 1980, it introduced the primary 1GB model, as big as a refrigerator and weighing about 550 pounds. Oh, and it cost $40,000. (In 1980 dollars: we’ll will let you do the maths.) The corporate made business machines at business prices, but a sea change was coming.

That very same year, Seagate Technology introduced the primary 5.25-inch harddrive, the ST-506, pictured above. Founded by former IBMers, including the legendary Al Shugart, who’d helped develop the RAMAC, Seagate targeted the nascent PC market with smaller, cheaper drives. The ST-506 offered 5MB for $1500; as a result of the endorsement of massive Blue (which had entered the microcomputer based on Apple’s early success), its interface soon became the de facto standard.

With a growing marketplace for personal computers, innovation flourished. Companies which include Western Digital, Quantum, Maxtor, Connor Peripherals, HP and Compaq all competed to develop the subsequent bigger, faster drive. Through the 1980s, capacity increased by up to 30% every year; within the next decade the number hit 60%. By 1999 storage capacity was doubling every 9 months.

These kinds of gains, though, came from refining the underlying technology, not fundamentally altering it. Rodime introduced the primary 3.5-inch hard disk drive in 1983, establishing the brand new standard form for desktop storage. Since then, manufacturers have sought to squeeze a growing number of data into that space, or into later 2.5-inch disks. Even smaller sizes — that 5 1.8-inch Toshiba 5 on your right, as an instance — still rely upon spinning platters. IBM’s 1-inch Microdrive shrunk the tech much more, and for it slow competed with CompactFlash by offering greater capacity.

To continue increasing capacity, manufacturers ought to keep shrinking the magnetic grains on those platters. Smaller grains means more bits per square inch, usually called areal density; upping the areal density means you are able to store more data inside the same physical space, and they are still 6 finding how one can try this 6 . Hitachi’s 7 perpendicular recording 7 offered another technique to boosting areal density, one soon taken up by other manufacturers.

Eventually, though, magnetic storage runs into fundamental laws of physics. To that end, those immutable rules are represented by the superparamagnetic effect (SPE). After we shink magnetic grains below a undeniable threshold, they become vulnerable to random thermal variations which can flip their direction. What exactly does that mean? Writing to an HDD means changing the magnetization of grains, marking them as ones or zeroes. So long as that magnetization remains ordered, the grains could be read — the info may be retrieved. But when they begin randomly flipping directions, you not have ones or zeroes. Coherent, readable information dissolves right into a bunch of magnetized grains. Perpendicular recording is a method to stave off the SPE limit, as is 8 heat-assisted magnetic recording 8 utilized in conjunction with bit-patterning. More 9 exotic solutions 9 are within the pipeline besides, and a few manufacturers just keep adding platters.

Given all of the recent attention, you would possibly think solid-state disks just suddenly appeared. In reality, like their electromechanical brethren, they date back to the 1950s. Indeed, the ancestors to today’s SSDs predate platter-based drives. Magnetic core memory, seen above, is one form of early storage that required no moving parts. It too emerged from IBM labs and regularly served as main memory inside the company’s mainframes. But partly due to the cost — it can only be handmade, with workers using microscopes to determine the tiny filaments they were threading — magnetic core memory was largely replaced by drum storage, which, you’ll recall, eventually ended in HDDs.

Still, solid-state memory had a spot in lots of niche markets, especially where high durability was required. NASA spacecraft depended on it, and in 1978 Texas Memory Systems began selling oil companies a 16KB RAM SSD as a part of a seismic data acquisition system. That was also the year that StorageTek introduced the primary modern SSD; with a maximum capacity of 90MB, it cost $8,800 / MB. The high price ticket made it and similar RAM-based disks appealing to just a select few. Equally important, DRAM’s speed came at a price: it was volatile memory, requiring constant power to retain its data. That worked for prime-speed, always-on applications, but not for home users. Today, DRAM still fills its role as main memory, but serves as storage in just a small choice of cases.

It took the discovery of flash memory to truly push SSDs toward the mainstream. Dr. Fujio Masuoka developed it in 1980 while at Toshiba. Much to its later chagrin, Tosh didn’t capitalize on his work, leaving it to Intel to commercialize. Chipzilla positioned flash as a an option for BIOSes and other firmware, but soon saw another application: removable storage. Intel’s MiniCard joined a proliferation of sizes and formats, including Toshiba’s SmartMedia (generically known as a high-quality-state floppy-disk card, and frequently sold with the adapter seen here), CompactFlash, Secure Digital (and its later variations) and Sony’s Memory Stick. All trusted flash.

So how does flash work, and what makes it different from traditional magnetic drives? The quick answer is that as opposed to storing data magnetically, flash uses electrons to suggest ones and zeroes. You would already recognize why it is a plus: no moving parts. Which means no noise, no head crashes, and bigger energy efficiency because you would not have to maneuver a mechanical arm. And in contrast to DRAM, it’s non-volatile — it doesn’t need constant power to retain information. These advantages are obvious, but within the early going, when placed next to cheap and capacious hard drives, flash still appeared like a gap product, useful mainly in digital cameras and other consumer electronics.

For a more in-depth introduction to the physical properties of flash, consider this 12-minute (!) video from SanDisk’s SSD Academy. Do not be concerned if you would like to skip it, though; we’ll address essentially the mostsome of the most salient details when we start taking a look at flash’s rise instead to the standard HDD.

7 7 Why SSDs?

8 Return to top 8

0

Here is a good place to take a break from our story and think about more closely the appeal of SSDs. As we’ve alluded, they promise greater speed, power savings and quiet operation. Today we sometimes ponder the performance boost as most persuasive, but for early laptop users, energy efficiency mattered just as much, if no more. (Ask someone who’s replaced his or her laptop HDD with an SSD in regards to the benefits to battery life.) But surely for Samsung to charge a $900 premium — and for somebody to pay it — the impact needs to be great indeed. In truth, skeptics continue to invite that query on message boards inside the internet: how good could the technology be, to justify the smaller capacities and better prices? In a single of his typically comprehensive and informative reviews, Anand Lai Shimpi also emphasized performance, responding, “You do not think they’re fast, until I take one far from you.” Since we won’t have an enormous, Oprah-style SSD giveaway (and later cruelly snatch them back), we’ll walk through how they work, and why they outperform HDDs.

1

Remember, today’s hard drives are approaching two physical limits: data density, which defines how much information may well be written on a given area, and spindle speed, or how briskly the platter spins. Greater data density gives us higher capacity drives, but is restricted by the superparamagnetic effect, barring new approaches. Spindle speed is a technique of accelerating throughput, but that’s topped out at 15,000RPM; rumors of a 9 20,000RPM Velociraptor 9 appear to have come to naught, perhaps because WD rethought the market. Perpendicular recording increases data density and throughput, assuming an identical spindle speed, but as you will discover from the diagram at the left, the dependence on moving parts imposes other limitations. As an instance, if you have ever let your hard disk spin down, then tried to read from it, you’ve probably noticed a small but perceptible lag. That is the drive spinning up and the read-write head moving around the platter to locate your data.

Solid-state drives have not one of the limitations linked to moving parts. There isn’t any spin-up time, because there isn’t anything to spin; because there’s no read-write head and all parts of the drive are equally accessible, latency and seek times are constant and occasional. The dearth of moving parts means less power consumption, because the drives doesn’t move heads or spin platters. And the probability of mechanical failure inside the style of say, a head crash, is non-existent. There is no head to crash.

All of which makes SSDs sound radically different from their HDD relatives — and they’re. Those differences have both pros and cons, which we’ll discuss momentarily, but first let’s get a handle on how the flash memory underlying SSDs actually works.

2

Here is a basic illustration of a flash cell. Notice it shares no similarity with the harddisk diagram above: we’re on the lowest layer of storage, where the foremost pressing question is tips to represent ones and zeros. HDDs do that magnetically; flash does it using electrons. The collection of electrons stored within the cell affect its threshold voltage: when the edge voltage reaches four volts, that reads as zero. Anything less reads as a one. (In flash parlance, a nil is “programmed” and a one is “erased.”) And the electrons are trapped within the gate even supposing power’s lost, making this non-volatile memory.

3

On this basic configuration, each cell stores a single bit: it reads as either a nil or a one. That, logically enough, is named Single-Level Cell (SLC) flash. However, using the very same physical medium, we are able to store more bits via subdividing the edge voltage ranges.

4

Now we’re recognizing four distinct threshold ranges instead of two. Meaning we will store two bits instead of one. Remember, we’re using a similar flash as before, so this is able to appear to be a bonus — we’ve doubled the capacity without raising our cost per bit. As always, though, there is a trade-off. First, it will take longer to read and write to MLC flash: typically about twice as long to read and 3 times longer to jot down. However, we’re talking about microseconds, so the adaptation is negligible for many applications.

More important is the matter of memory wear. Unlike HDD platters, which in theory could be written and re-written an unlimited collection of times, flash memory can only be programmed/erased a limited variety of times before it’s now not writeable. It’s called a P/E cycle limit, and with the sooner 50nm chips that number was about 100,000 times when used as SLC memory. Because MLC degrades faster, after about 10,000 P/E cycles. And as NAND structures have got smaller, so too have the selection of P/E cycles they’re able to undergo before wearing out. Sure, the numbers are still so high that the typical, everyday user would consider the drive obsolete long before actually hitting the P/E cycle limit — at which point the drive becomes read-only, while preserving the present data. But if it involves design and manufacturing, the limitation is rather real, and wishes an answer. Or rather, several solutions.

0 0 The challenges of SSDs

1 Return to top 1

5

As we’ve shown, flash SSDs present unique challenges. That’s one reason established player Seagate 2 came to the party late 2 , and Western Digital eventually just 3 bought its way in 3 . (We also suspect uncertainty in regards to the viability of the market and concern about cannibalizing sales in their mechanical drives also had a bit something to do with it.) And it’s why a late-entering semiconductor giant named Intel — presumably with the time and expertise to be told from others’ mistakes — could release its first drives to just about universal acclaim, only to get hit later with claims of 4 drive slowdown 4 .

Despite Intel’s 5 initial denials 5 , the drives did see a performance drop through the years. Specifically, because the drives filled up, write speeds slowed, sometimes drastically. This wasn’t only a problem with Intel’s offerings, either. Once reviewers knew to see for it, they discovered it common to just about every SSD: as free space decreased, write performance took successful. Most drives were still faster than conventional HDDs, however the difference was noticeable between a brand-new SSD and a “used” one. In hindsight the explanation seems obvious, but it surely wasn’t immediately so in 2009.

Here’s where two aspects of flash memory converge to create a singular problem. The 1st, as we mentioned, is memory wear. No HDD controller has to account for the predictably finite lifespan of its underlying magnetic media. Flash SSDs do: they must limit the variety of P/E cycles that allows you to keep the drive in tip-top shape. Not just that, but remember the fact that our P/E limit is per cell. It’s roughly analagous to the issue of bad sectors, but hence it’s predictable and inevitable for each cell. Knowing this, controller designers want each cell to wear evenly — spread the P/E cycles over the drive, as opposed to programming and erasing an identical cells until they become unusable. This can be called wear-leveling, and it’s further complicated by our second aspect: the physical architecture of flash memory.

6 Remember the cells we introduced above. They’re the smallest part of storage: in SLC flash they store a single bit, and in MLC flash they store two. Those cells are grouped into pages, typically 4KB in size (see the illustration for your left). A page is the smallest structure that’s readable/writable in an SSD. Pages are grouped into blocks, that are the smallest erasable structure in an SSD. Now you are seeing a difficulty. Why read, write and erase? HDDs should not have a separate erase function on the physical level. Once you delete a file, that simply means removing a pointer. There isn’t a action taken at the harddisk, no “erase” function. Your data remain magnetically encoded at the drive, so that it will eventually overwrite the “free” space.

But flash doesn’t work that way. It is a different medium with different rules. SSD makers favor to play by these rules as the upside is vastly improved performance.

In the event you erase a file on an SSD, the method is initially the identical: nothing happens. At the least not on the physical level. No data disappears. Shall we say you deleted a 4KB file that got written to a brand new page. That page is now free, so far as your operating system’s concerned. Provided that you need to overwrite that page will the SSD need to do a little work. But if it does, it has to do greater than the HDD, and that is the key to understanding performance degradation.

As we said, the HDD simply overwrites the field with new information. The SSD, though, can’t just overwrite a page. It has to erase the page first. Now remember the asymmetry between readable/writeable and erasable structures. To erase a page, you might want to erase the complete block containing it. What in regards to the other pages within the block? Well, they must be read to a buffer, then written back after the block’s been erased.

You will see how this results in a drop in performance. You simply tried to jot down a page, but in reality you ended up reading a block, erasing it, then re-writing it with the hot data. What feels like an easy write operation is actually a 3 step process. SSDs attempt to avoid this by writing to open pages first, but as space fills up the controller has fewer options. There’s simply nowhere else to lay the information without doing this read-erase-write shuffle.

7 Now, manufacturers recognize this and take a look at to mitigate it in several ways. One is over-provisioning: including more flash at the drive than the user actually sees. Intel’s X-25M, for instance, shipped with 7.5-8% extra flash. The spare space means more open pages that may not require read-erase-write. But that just staves off the inevitable. In case you keep filling up the drive, you are going to hit the performance barrier. The question is how well your drive will cope.

Another helpful approach is the TRIM command, now implemented in most recent operating systems. This forces the SSD to administer deleted files quickly, instead of look forward to them to be overwritten. Delete a file and the OS tells the controller to duplicate the block to cache, erase it, and rewrite the remainder pages. Which means pages within a partially used block are freed up automatically, throughout the delete phase instead of during an overwrite. In fact, sometimes it’s good to overwrite a file, say once you save a brand new version of it. You’ll still suffer the read-erase-write penalty then, but TRIM can alleviate among the pain. Some drives feature other methods of 6 garbage collection 6 , all with a similar goal: unencumber deleted pages before it is necessary to overwrite them.

But wait a minute. Rewriting more blocks eats into our limited P/E cycles, right? That’s right. TRIM and other methods of garbage collection contribute to an issue called write amplification. That simply means you’re writing to the SSD more often it’s essential; ideally, the write amplification ratio will be 1, meaning the quantity of information written to the flash memory is strictly kind of like that written to the host. This remained a perfect, though Intel’s early drives came with reference to reaching it. This number seemed a threshold, too; of course, how could you write less data to the flash memory than to the host? One company discovered how, and we’ll pick up the tale there.

7 7 Not all are created equal

8 Return to top 8

8

That’s Intel’s X-25M, 9 launched 9 in late 2008. Sandisk, Toshiba and TDK had already 0 entered the market 0 , which really started to balloon in 2007. It recalled the early days of desktop hard drives, with players large and small attempting to outdo each other.

Unfortunately, that did not always mean an awesome experience for consumers. JMicron had begun offering its SSD controllers to smaller, independent vendors corresponding to OCZ, Super Talent and Patriot Memory. The controller let those companies use the cheaper MLC flash, while Samsung, for one, stuck with the dearer, lower-capacity SLC. But users of JMicron’s early controllers found serious problems. While fast in benchmarks for sequential reading and writing, under real-world conditions the drives stuttered unacceptably.

The issue revealed some narrow thinking at JMicron. Optimized for sequential reading and writing, its controllers choked badly when it came to random 4k writing. But most users don’t spend their days reading and writing sequential files. Nor do they buy storage based solely on benchmarks. Instead, most should use multi-tasking operating systems that — you guessed it — write a considerable number of small files. The JMicron hiccup occured when writing those files interfered with other applications. The difficulty must have been caught early, before drives shipped to consumers, but instead buyers became unwilling beta testers.

Chipzilla’s entrance made things interesting. The corporate had the expertise to construct an outstanding first controller and the industry pull to secure flash at bargain prices. In spite of this, many were surprised when it launched among the world’s fastest drives. A two-pronged attack — the X-25M for consumer use and the X-25E sporting SLC for enterprises — put Intel arguably on the top of the heap, performance-wise. 9

Here’s where 1 SandForce 1 is available in. The corporate 2 entered 2 the SSD controller ring in 2009, emerging from stealth mode with a promising cache of proprietary technologies. Their big breakthrough? A quartet of tweaks that allowed MLC to switch the dearer SLC without sacrificing durability or speed. Since MLC is twice as dense as SLC, that meant doubled capacity. It also let the smaller firms compete with SLC powerhouses like Samsung and Intel, who had privileged access to high-grade NAND chips.

OCZ shipped many of the first drives featuring a SandForce controller, and basically 3 hit the limit 3 on 3Gbps SATA, hitting 265MB/s on a 2MB sequential read. Granted, most other high-end SSDs also bumped up against that SATA ceiling: high speeds are an enormous portion of the appeal, and are virtually innate to the technology. Where SandForce really impressed, though, was within the sequential write tests, reaching 252MB/s and blowing even Intel’s enterprise offering out of the water.

How’d SandForce go toe-to-toe with Intel? First, they’d an intelligent data-monitoring system called DuraWrite. Remember, the issue with MLC is that doubling the NAND storage capacity shortens its P/E cycle limit by a couple of factor of ten. SandForce reasoned that in case you desired to make MLC competitive with SLC, you simply needed to reduce the volume of writing actually happening — considerably. DuraWrite does just that, through a mix of compression and deduplication. Which means less redundant data gets written, lowering the write amplication ratio to 0.5, SandForce claims. (That was enough to 4 catch IBM’s eye 4 .) Other enhancements targeted reliability, power consumption and, obviously, those lightning-quick read and write times.

SandForce has arguably struck a balance between price and function, one which permits them to serve both consumer and enterprise markets, and it kind of feels to be paying off. Despite making no drives of its own, the corporate is without doubt one of the most recognizable names in SSDs, and was just bought 5 just bought by LSI 5 for $370 million. At once, SandForce looks to be within the pole position for the cast-state innovation race.

Needless to say, which could change at any time. The SSD market remains just a little a Wild West: the technology hasn’t been perfected, and we’ve all heard a lot of horror stories about 8 recalls 8 , 9 faulty firmware 9 and 0 other problems 0 . For some users, though, the reward is definitely worth the risk. It also is still seen whether more than one companies will emerge victorious, as did Seagate and Western Digital with HDDs, or whether SDDs will continue to return from a wide selection of manufacturers. As everyone knows, today’s winners may be tomorrow’s losers. The only real thing we will be able to say with much certainty is that SSDs show loads of promise, and we’re just starting to tap it.

[Image credits: 1 Ed Thelen 1 , 2 IBM 2 , 3 Micron 3 , 4 Orion 8 4 , 5 Texas Memory Systems 5 , 6 OEMPCWorld 6 and 7 Anandtech 7 ]

Source

This post is tagged: conscious consumer, horror stories, magnetic drives, solid state disks, solid state storage