StoreMI: The Way To A Faster JBOD

When AMD launched the Ryzen APUs earlier this year, one of the minor side-announcements was that AMD was promoting software called FuzeDrive, from Enmotus. For an extra $20, this software could be yours. For the new Ryzen-2000 series desktop processors and X470 platform, this software is now branded as part of AMD’s software stack, and can be downloaded for free from the AMD website under the StoreMI branding.

The Benefits of a Cache

The principle for StoreMI is that a user can take a mid-range system, powered by a slow drive, and add a small but fast drive to increase the speed of the most frequently accessed files. The software creates a storage ‘tier’ between the fast drive and the slow drive, giving the user a single drive with the combined capacity of the two drives, while the software implements pattern recognition to understand which files can be placed on the fast drive and help with acceleration.

Tiered storage is not new – it is used in many storage-focused enterprise systems backed by very complex software. Fast storage is small expensive, while content is typically large: content delivery networks (CDNs), like Netflix or Steam, will use tiered storage and caching such that the most frequency accessed films or games will come from storage that is both near to the user and from the fastest possible location.

In a modern computer, the fastest storage is the onboard memory / DRAM. This is where files and programs are loaded to when the processor needs to access the data or keep data close for the lifetime of the software. In recent years certain enthusiasts have used RAMDisks, creating a drive using the onboard memory, to act as a fast repository. The only downside is that the data is lost on restart as the data held in DRAM is volatile (or not persistent).

A modern enthusiast machine is likely to have some sort of solid-state drive (SSD) based on NAND flash – either a drive connected via the PCIe lanes as super-fast storage, or connected through a SATA port. These users often have a mechanical hard-drive, based on spinning platters of rust, as a backup for bulk storage, commonly referred to as a hard-disk drive (HDD), and the speed is limited to how the system reads from a drive that spins at 7200 or 5400 times per second. Many machines still ship with HDDs as their primary and bulk storage, much to the dismay of users that notice the immediate user experience benefit of an SSD.

With StoreMI, a user can take any configuration of PCIe SSD, SATA SSD, or HDD, and implement them into a tiered storage drive. The combined single drive will have the capacity of all the drives in the tier, and the software will manage which data should be moved around into the fast storage drive. This process is gradual, and the software will take time to learn which files are the most important – this will mean that the effect will not be immediately noticeable, but after the third or fourth time that software or a game has been loaded, the system should have a good idea.

The StoreMI tool also allows the user to add up to 2GB of onboard memory into the storage tier. This process does not add additional capacity to the tier, but the DRAM acts as the fastest cache and will hold copies of the data held on other drives such that data is not lost. As mentioned above, as onboard memory is volatile, the speed up information will be lost on restart. In our discussions with AMD, they felt that 2GB was a good amount of memory for this DRAM cache: due to the way the software works, the company said that a large cache showed no tangible benefit. It also allows systems with only 8GB of total system memory to take advantage of the software.

Configurations

Unlike the caching technology behind Intel’s RST (which has only recently supported caching on non-Boot drives), AMD’s StoreMI can be used at any time in the lifecycle of the system. For any user that wants to delay the purchase of an SSD NVMe or SSD SATA storage drive, or delay to buy a bigger drive, they can do so and implement it into the tiered storage at a later date.

StoreMI can support almost any configuration requested, either on a Boot drive or on a data drive. For a Boot drive, AMD recommends installing the operating system on the higher capacity slower drive first, such that the HDD, and adding a blank SSD as the fast tier, although the reverse is also possible for users that want to add the larger drive later (there may be additional steps to the process). The only difference is that the software is likely to move a lot of data around at the beginning.

Configurations that are suggested for StoreMI are:

  1. HDD + DRAM
  2. HDD + SATA SSD
  3. HDD + SATA SSD + DRAM
  4. HDD + NVMe SSD
  5. HDD + NVMe SSD + DRAM
  6. SATA SSD + DRAM
  7. SATA SSD + NVMe SSD + DRAM
  8. NVMe SSD + DRAM

The biggest noticeable improvement should occur in configuration 4, when an NVMe SSD is paired with a mechanical HDD.

AMD states that if a tier reaches across from native chipset to controller based SATA ports, the software is likely to move files related to hibernation over to the drive on the native SATA ports for stability; this process might take up to 30 minutes.

Users can also remove drives from the tiered storage, if there is enough space to put all the data on the drive that stays in the tier. The removed drive will be left with zero data, and can be removed from the system or used for other things.

The Big Limitation: 256 GB on the Fast Tier

A point not mentioned in our initial briefings when FuzeDrive was launched alongside the APUs, but repeated sufficiently in the StoreMI User Guide, is that AMD’s bulk licensing deal with Enmotus means that the faster drive in the tier can only be up to 256GB in size.

When adding a drive larger than 256GB as the fast tier, the system will partition the blank drive automatically, offering the extra capacity as a separate drive letter on its own.

When adding a large HDD as the slow tier to an SSD boot drive, this is only a problem if the SSD is bigger than 256. Users in this circumstance will be required to migrate the operating system (using other software) from the SSD to the HDD first, then boot the system using the HDD and add the (now blank) SSD as a fast tier.

Some users might see this as a big deal – adding a 3TB HDD slow drive to a 512GB SSD-based boot drive shouldn’t have to be this complicated. However in this circumstance it might be suggested that the drives be kept separate, and items like the Steam folder are manually bifurcated into two locations with the favorite games on the sizeable SSD. However StoreMI is geared more towards systems that would naturally only have a large HDD in them to begin with – adding in a small fast SSD, say 64GB-128GB, is the intended use case here.

The Failure Rate: Downsides of a JBOD

Most options for combining drives in an array involve an element of speed (reading data across many drives at once) or protection against failures (data is replicated, or a parity bit is introduced), and usually storage array options combine both, trading failure protection for speed or vice-versa. For users involved in storing files, one of the options for combining drives is known as a JBOD, or ‘just a bunch of disks’. A JBOD offers neither speed nor failure protection.

A JBOD array does one thing: it combines the drives into the array to appear as one contiguous file space, and it treats it as such. Putting eight 10 TB drives into a JBOD will appear as an 80 TB drive in a system. However, it does not read from or write to the drives simultaneously – it will purely write data in a sequential manner and read the data from the drive that has it. As a result, it is still only as fast as a single drive, but if one of the drives in the JBOD array fails, the array is broken and the array is lost. Without specialist tools, all the data is lot from the whole array as well, and the data on that specific drive is almost certainly gone. If a single drive has an average failure rate, an eight drive array is expected to fail eight times as frequently as it relies on every drive being in workable condition.

What StoreMI does in this context is that it suffers from the same lack of protection from drive failures. There is no mechanism by which the data is protected if one drive in the tier fails – if one drive in the tier fails, the data across all the drives is lost. If the boot drive is a bargain SSD using low quality NAND, or the hard drive is old, then losing the data across both drives is a real possibility.

Initially I thought this was a significant issue. If a user was to put 10 drives into the tier, for example, it could spell disaster. The significance was muted however, when I learned of the 256 GB fast tier size limit, as it means that most users are likely to only pair two drives into a tier. AMD’s response to the threat of failure was to say that users should expect to keep backups regardless, and no specific comment was made based on the increased failure rate of a two-drive tier compared to a single drive. Because StoreMI moves data around from SSD to HDD, there could be additional concern as the software could write more data to the SSD over time than a casual user might do if it was just the boot drive, causing the NAND to wear out quicker. Neither AMD nor AnandTech see this as much of an issue, given that modern MLC and TLC SSDs are very good at managing bad data blocks and have overprovisioning built in.

Benchmarks!?

At present we haven’t had time to directly test StoreMI, having focused on other projects and upcoming events. If we get time, we will have a new article on StoreMI.

New X470 Chipset and Motherboards: A Focus on Power Benchmarking Setup and Power Analysis
POST A COMMENT

546 Comments

View All Comments

  • jjj - Thursday, April 19, 2018 - link

    I was wondering about gaming, so there is no mistake there as Ryzen 2 seems to top Intel.
    As of right now, I don't seem to find memory specs in the review yet, safe to assume you did as always, highest non-OC so Ryzen is using faster DRAM?
    Also yet to spot memory letency, any chance you have some numbers at 3600MHz vs Intel? Thanks.
    Reply
  • jjj - Thursday, April 19, 2018 - link

    And just between us, would be nice to have some Vega gaming results under DX12. Reply
  • aliquis - Thursday, April 19, 2018 - link

    Would be nice if any reviewer actually benchmarked storage devices maybe even virtualization because then we'd see meltdown and spectre mitigation performance. Then again do AMD have any for spectre v2 yet? If not who knows what that will do. Reply
  • HStewart - Thursday, April 19, 2018 - link

    I notice that that systems had higher memory, but for me I believe single threaded performance is more important that more cores. But it would be bias if one platform is OC more than another. Personally I don't over clock - except for what is provided with CPU like Turbo mode.

    One thing that I foresee in the future is Intel coming out with 8 core Coffee Lake

    But at least it appears one thing is over is this Meltdown/Spectre stuff
    Reply
  • Lolimaster - Thursday, April 19, 2018 - link

    Intel 8 core CL won't stop the bleeding, lose more profits making them "cheap" vs a new Ryzen 7nm with at least 10% more clocks and 10% more IPC, RIP. Reply
  • HStewart - Thursday, April 19, 2018 - link

    I just have to agree to disagree on that statement - especially on "cheap" statement Reply
  • ACE76 - Thursday, April 19, 2018 - link

    CL can't scale to 8 cores...not without done serious changes to it's architecture...Intel is in some trouble with this Ryzen refresh...also worth noting is that 7nm Ryzen 2 will likely bring a considerable performance jump while Intel isn't sitting on anything worthwhile at the moment. Reply
  • Alphasoldier - Friday, April 20, 2018 - link

    All Intel's 8cores in HEDT except SkylakeX are based on their year older architecture with a bigger cache and the quad channel.

    So if Intel have the need, they will simply make a CL 8core. 2700X is pretty hungry when OC'd, so Intel don't have to worry at all about its power consuption.
    Reply
  • moozooh - Sunday, April 22, 2018 - link

    > 2700X is pretty hungry when OC'd
    And Intel chips aren't? If Zen+ is already on Intel's heels for both performance per watt and raw frequency, a 7nm chip with improved IPC and/or cache is very likely going to have them pull ahead by a significant margin. And even if it won't, it's still going to eat into Intel's profit as their next tech is 10nm vs. AMD's 7nm, meaning more optimal wafer estate utilization for the latter.

    AMD has really climbed back at the top of their game; I've been in the Intel camp for the last 12 years or so, but the recent developments throw me way back to K7 and A64 days. Almost makes me sad that I won't have any reason to move to a different mobo in the next 6–8 years or so.
    Reply
  • mapesdhs - Friday, March 29, 2019 - link

    Amusing to look back given how things panned out. So yes, Intel released the 9900K, but it was 100% more expensive than the 2700X. :D A complete joke. And meanwhile tech reviewers raved about a peasly 5 to 5.2 oc, on a chip that already has a 4.7 max turbo (major yawn fest), focusing on specific 1080p gaming tests that gave silly high fps number favoured by a market segment that is a tiny minority. Then what happens, RTX comes out and pushes the PR focus right back down to 60Hz. :D

    I wish people to stop drinking the Intel/NVIDIA coolaid. AMD does it aswell sometimes, but it's bizarre how uncritical tech reviewers often are about these things. The 9900K dragged mainstream CPU pricing up to HEDT levels; epic fail. Some said oh but it's great for poorly optimised apps like Premiere, completely ignoring the "poorly optimised" part (ie. why the lack of pressure to make Adobe write better code? It's weird to justify an overpriced CPU on the back of a pro app that ought to run a lot better on far cheaper products).
    Reply

Log in

Don't have an account? Sign up now