Logo MacGurus and BurlyStorage

 

   1-208-983-9999
Shopping Cart

Shopping Cart 0 Items (Empty)

Store Search

Advanced Search

USB 3 and Thunderbolt

State of Storage - USB 3 and Thunderbolt - Mid-2013

Rick Stephens
7/23/13

  Planning storage migration on Macs, from eSATA and Firewire onto USB 3.0 and Thunderbolt.

  Every once in a while, everything seems to change all at once. 'Everything' is relative, of course, but lately the changes have been piling up. And for the professional, it is hard to plan where the storage path is going in the near future when all the options are different from what we are using today.

  What's changed:

  First to Apple came a USB 3.0 bus replacing both the old reliable Firewire800 and the long in the tooth USB 2.0. And then the much hyped Thunderbolt takes over from PCIe slots in the MacPro while adding unimaginable throughput power to the iMac, Mini and MacBook Pro lines. A new bus taking the place of an old one is not earth shattering. It has happened before and will happen many times again. But a replacement of two buses at once is a challenge for the user needing to migrate to new equipment. Happily most of us can get on board for significant performance gain if it can be done while maintaining sanity. The bad side is, both USB 3.0 and Thunderbolt have placed technical hurdles as well as financial ones in front of those who can't wait out the teething problems inherent in new technologies.

  USB 3.0.

  USB 3.0 bus specification is very fast, mature, and so common that the pricing is insanely competitive. USB 3.0 should be the simplest and least costly options path for those needing storage right now. The problem is, either Apple incorporated newer futuristic features that are beyond the capabilities of current products, or just because Apple had to do it their own way, instead of like anyone else would, USB 3.0 on a Mac has been a nightmare of incompatibilities and limited functionality with much of the existing USB 3.0 product base. Also a consideration, Apple's USB 3.0 performance is fairly lackluster. USB 3.0 specification allows for throughput performance up around 450-500 MB/sec. So far, the best we have seen from an Apple USB 3.0 port is 205 MB/sec. 205 is no slouch when compared to Firewire or lesser USB buses, but nowhere near what can be enjoyed from USB 3.0 ports on PC motherboards.

  USB 3.0 is not new, it has been around on PCs for a number of years. While USB has never been my favorite choice for mass data storage, and access to same, it is a common and widely used cross platform technology built into most computers today. For whatever reason, as mused on earlier, getting an Apple computer with a native USB 3.0 bus to work with most existing USB 3.0 to SATA bridges or RAID devices is a study in unpredictability. For a storage engineer and the professional photographer or videographer, unpredictability is the one thing we loath when it comes to our storage. And these problems can span the range from directory damaging spontaneous dismounts, to sleep problems or data corruption. Multi-drive storage is particularly afflicted, but some single drive enclosures have bridges that share some or all of these behaviors. Also weakened from what we could count on in past eSATA and Firewire storage is the ability to seamlessly hotswap drives. I have yet to see a USB 3.0 multi drive enclosure that can handle swapping out one drive without negatively effecting the other drives attached.

  That said, usually there are workarounds that allow mostly reliable access. Applications like Keep Drive Spinning, which places an applet in your Library folder that touches your drives regularly, keeping them online, will often remedy the spontaneous dismount problems. Or Jettison, used when Apple's USB bus drops drives offline when the computer sleeps. Jettison is a application available at the app store, can eject drives whenever the computer sleeps to prevent drive index corruptions. Turning off sleep or adjusting settings can allow drives to stay mounted. Having to carefully orchestrate our behavior in handling our storage and implementing workarounds for poor bus characteristics is not what we all have come to expect from connecting up to our Apple computers. What happened to the ease of plug and play?

  Developers are hard at fixes. And we have found a few developers that have come up with firmware fixes that get their products working well despite the problems Apple has introduced. While we can't possibly test every product out there, I would bet there are other devices that work reasonably well. For the dozens of multi-drive USB 3.0 devices we have tested there have always been issues to work through. And, so far, none of them have been of flawless plug and play reliability that I look for in a storage product. We'll keep looking and working on it.

  Thunderbolt.

  First off, what is it? Sometimes hard to figure with all the hype. But simply put, in use a Thunderbolt port is most like the PCIe slot on a MacPro where we plug different cards in and gain faster, maybe newer buses and hardware. That kind of plug in power was only found by virtue of the MacPro tower's PCIe slots, but now we get this throughput power in MacBook Pro, Mini, iPad, iMacs, and soon the new MacPro with its incredible SIX Rev 2 Thunderbolt ports! With a Thunderbolt port you can plug in monitors, storage systems, capture cards or any of many other devices that make it through Intel's Thunderbolt development process. The way these devices work is pretty much identical to how they worked in a PCIe slot. The only difference is that instead of plugging into the PCIe port, the card now sits inside the external device and connects to the PCIe bus through a Thunderbolt chip. The device works exactly the same as it did connected to the PCIe slot. Best of all, so far, unlike USB 3, Thunderbolt pretty much works as advertised. Devices that do make it through Intel's costly and time consuming approval process tend to work well and are typically fast and reliable. So there is a silver lining to the expensive process. Devices that don't meet Intel's Thunderbolt specification don't make it through the process. That isn't to say that the mechanical features of a given device are always good quality (believe me when I say there is a lot of cheap junk made that gains approval), only that all the products with approval meet the specification.

  Advertised speed versus actual speeds. Marketing seems to rule here with Thunderbolt being credited with speed, speed and more speed, and curing bad breath and acne. Thunderbolt IS fast and convenient, but magic it ain't. It is simply PCIe over a wire, where the PCIe card is now inside the device attached via Thunderbolt instead of a card being plugged into the PCIe slot of the computer. Most of the time, with anything except a very high end capture device or RAID controller, performance is pretty much the same as it was when we plugged the card into the PCIe slot. And for most simple devices, performance is just the same. For instance, a 2 hard drive storage enclosure isn't any faster attached to Thunderbolt than it was to an eSATA card. It is just attached with a different cable. The claims that Thunderbolt is faster than a PCIe slot is so far, just nonsense. At this point a Thunderbolt bus has less bandwidth than most PCIe slots. There are numerous PCIe based devices that are faster than the capability of a current Rev 1 Thunderbolt bus. Once the new MacPro comes out with its Rev 2 Thunderbolt bus, we will see new devices built that will take advantage of the greater speed promised by Rev 2. Of specific note, most devices cannot use more bandwidth than a Rev 1 Thunderbolt bus. Only a high end RAID, capture card or similar very bandwidth needful device will have any need for Rev 2 bandwidth. But with Rev 2 we will enjoy the availability of greater bandwidth. Also note, Rev 1 Thunderbolt devices will not have a problem running on a Rev 2 Thunderbolt bus.

  Downside of Thunderbolt. Currently the biggest issue to lots of diverse products at great prices is the high cost of approval and the long process it takes to get a device through Intel. It is not uncommon to spend ,000 to ,000 to get final approval on a device. And that high initial cost takes the component manufacturers out of the game. We see virtually no simple inexpensive Thunderbolt bridge boards, like a Thunderbolt to SATA port multiplier bridge board, because no one can afford to invest k on a board that retails for 0 unless they have a very large market indeed. Hence, we only see complete devices that have the likelihood of selling enough quantities to pay back the very high initial costs.

  That means today, adapting an eSATA port multiplier enclosure to run on Thunderbolt is costly and requires the purchase of multiple parts, like the combination of a Thunderbolt to Expresscard34 adapter and an eSATA Exp34 host card. We can also use a Thunderbolt to PCIe adapter, and install an eSATA port multiplier capable host card. But because so few PCIe eSATA cards have Thunderbolt drivers written for them, we have a very limited set of card choices. Currently the best card choice for this task has MiniSAS type connections on the card which add extra cost by virtue of requiring new external cables as well as card and the Thunderbolt PCIe expansion box.

  Upside, these solutions are very convenient to use, reliable and often extremely fast, even if costly. As the cost of development goes down, more of the component manufacturers will get in the game, which will end up decreasing the end cost and increasing our options in how we retrofit existing storage. I expect this to happen in the relatively near future as market pressure is forcing anyone who wants into the Mac market to produce Thunderbolt products.

  New products and fixes for older devices are coming out on a daily basis. I expect this article to be obsolete in no time. So stay tuned for updates.

 Rick Stephens
7/23/13

Kryptronic Internet Software Solutions