Report iodyne Pro Data

Note: This is a live document and is constantly updated, as soon as we have new information.

Introduction

The purpose of this report is to document the testing of a device that was sent to our team for evaluation. The device in question is a storage solution utilizing Thunderbolt 3 technology, called Pro Data, by the company iodyne LLC based in the USA.

Methods

To test the device, the following steps were taken:

• All necessary kext files were installed and checked for proper functioning using the terminal command kextstat | grep -v com.apple
• The device was set up with one container across the entire storage pool, using a name with no special characters and RAID-0, APFS.
• Speed tests were conducted using Blackmagic Disk Speed, DiskMark, AJA System Test Lite and WrangleBot SpeedMark as a control.

Results

According to calculations, Thunderbolt 3 has a transfer rate of 40 Gbits, with 20 Gbits being data. This translates to 2.5 Gigabytes per second. However, the measured write speeds were found to be half of this expected rate.

• Blackmagic Disk Speed reported write speeds of 1200 MB/s
• Amorphous DiskMark reported write speeds of 690 MB/s, which seemed to slow and read speeds of 3400 MB/s, which would exceed the theoretical limit of one TB3 interface. These results were suspected to be incorrect and were initially discarded from the findings, however the manufacturer recommended this tool for testing so we will reflect the values in our test, for both read and write.
• WrangleBot SpeedMark, used as a control measurement, showed results that matched more closely with Blackmagic's values, which was expected.

Comparison with Contemporary Storage Solutions

Background on Benchmarking Issues

There are several reasons why a disk speed benchmark may return incorrect read and write values. Some possible causes include:

• The disk is not properly connected to the computer or the benchmarking software is not configured correctly, resulting in inaccurate test results.
• The disk is experiencing performance issues such as high disk usage, fragmentation, or hardware failure, which can also affect the benchmark results.
• The benchmarking software may be using a different file size or block size than the actual disk usage, which can also affect the results.
• The benchmarking software may be using a different test method than the actual disk usage, which can also affect the results.
• The computer may be running other programs or processes that can interfere with the benchmark results.

It's also important to keep in mind that benchmark results can vary depending on the specific software and hardware used, as well as the test conditions. It's always a good idea to run multiple benchmarks and compare the results to get a more accurate picture of the disk performance.

We've made sure to address these issues during our testing. By testing different block sizes, cycles, burst and sustained access and other techniques to test for possible false configurations.

Benchmarks

For reference, the following are the speeds reported by existing contemporary storage solutions.

Tools used: AJA System Test, Blackmagic Disk Speed Test, iStatsMenu Pro, Amorphous Disk Mark

Legend: ★ Values have been averaged over 3 or more measurements. ☆ Values are based on only 2 or less measurements.

Comparison to the enterprise solution

The Seagate Lyve Mobile Array is currently considered the fastest single-controller Thunderbolt 3 solid state disk storage device. It features a RAID of NVME sticks manufactured by Seagate and can reach speeds of 2.5GB/s for both read and write using a single Thunderbolt 3 cable. However, the iodyne Pro Data can only reach these speeds when reading data, with write speeds being half of the 2.5GB/s.

In terms of cost, the iodyne Pro Data has a starting price of $3,500 and can go up to $17,500, making it more affordable for owners and traditional rentals. The Seagate Lyve Mobile Array, on the other hand, has a price of $60,000 and is only available through specialized Seagate distributors, such as VAN ROTHE, at high rental periods for the film industry. This makes the iodyne more attractive for medium enterprises looking to purchase and own premium SSD storage. However, the Seagate Lyve Mobile Array offers a better performance and larger ecosystem that is based on modularity, which makes it more attractive for large scale enterprises looking for a tailored solution that fits with contemporary networking solutions and workflows.

Comparison to most cost effective solution

The OWC Thunderbay 4 Mini is a solid state disk storage device that reaches speeds of around 1.4GB/s for both read and write. Given its low price of only 1300 €s, it is a cost-effective option. The price includes 300 €s for the chassis and around 4x250 €s for 4TB 2.5" SSDs. The maximum setup with RAID 0 will provide 16TB of storage. Additionally, the Thunderbay 4 Mini is capable of daisy chaining, which allows for adding another 16TB of storage at only 1300 €s per chassis, with a total of 6 devices per Thunderbolt 3 bus for a total cost of 7800 €s, resulting in 96TB of usable storage. However, this setup would require six power bricks, which would take up a lot of space and consume a significant amount of energy. The total footprint would also be larger than that of two 48TB iodyne Pro Datas. Nevertheless, at around 7800 €s, the OWC Thunderbay 4 Mini's solution is 78% cheaper, which may outweigh any other factors in a purchase decision.

OWC also offers a Thunderbay 8, but this offers little in cost effectives and would actually be more expensive than two docks, especially considering the size benefit, from having 2,5 inch bays instead of 3,5 inch for HDDs.

Comparison to cheapest solution

The Synology 12 Bay NAS device is an affordable option for providing up to 220 TB of storage using HDDs. However, it is important to note that NAS systems are capable of hosting much larger amounts of storage, up to Petabytes. This statement is focusing on comparing storage solutions for clients who may not have an in-house IT team with an existing workflow.

HDDs are currently the cheapest option for hot storage, with fast read and write speeds of up to 1GB/s. However, they may not be the best option for certain types of data streams. For example, using HDDs in a RAID configuration may cause slow response times, making it difficult to view footage larger than 1080p. 4K footage may only be possible with compressed data streams, such as HVEC, H264 or AVC Intra LongG.

Using a 4TB cache with NVME or SSD can mitigate some of these issues, but it may not be sufficient for larger projects or higher resolution footage. The cache would need to be constantly refreshed for 8K or higher resolution footage, so it would only be effective if hot data is smaller than or equal to the cache size.

Multipathing

Testing for multipathing was conducted, but no initial improvement in speed was detected.

Turns out that it is important to note that multipathing requires an individual thunderbolt 3 pair per thunderbolt connection to a Mac. On non-M1 Macs, two thunderbolt 3 ports are present per bus, so it is essential to use one cable per bus.

Following the misconfiguration, a minimal increase in Write Speeds (20-50MB/s) was observed, with a notable rise in Read Speeds with up to 2500MB/s for a total of 5000MB/s. We were unable to test Multipathing with multiple Pro Data's, which reportedly can achieve up to 5GB/s in write and up to 10GB/s in Read. If so it would vastly outperform any comparable device in this price range, even the Seagate Lyve Mobile Array. We hope to verify these claims soon.

Strange Behaviour on Apple M1 Silicon Chips

When two TB3 cables were plugged into a MacBook Air M1 or Mac mini M1, a strange behavior was observed. Kernel_task started using 1/4 of the CPU cycles to communicate large amounts of data, as long as the device was connected with two ports. This issue could only be resolved by disconnecting the device entirely and then reconnecting with only one port.

After further discussion with iodyne, we confirmed that this was a previously discovered M1 Silicon Thunderbolt Bus bug and the fix lies in the hand of Apple. With Intel boards these issues do not arise, as Intel Thunderbold infrastructure only allows a single Host on both Down and Up Stream. Whereas Apple allows any device to self negotiate host profiles. This would create a feedback loop as the Bus continues to try and negotiate a ever increasing payload of signatures.

Conclusion

We observed that the device met the advertised speeds of 5GB/s when employing multipathing, however, the write speed was slightly below the appraised values. We were unable to reproduce the manufacturer's benchmarks using Amorphous Disk Mark or any of our other resources and tools. Thus, further investigation and testing is necessary to pinpoint the cause of the discrepancy and find a remedy to unlock the device's full potential. All disk drives were reported to be in good health according to the accompanying application. The team is looking forward to receiving feedback from the manufacturer on how to address the issues encountered during testing