Exploring Cache Performance
A typical application workload consists of a mixture of I/O sizes and access patterns spread over the surface of the drive. This data tends to congregate in areas referred to as "hot bands," denoted by the blue areas in the graphic above. Hot bands of data are areas accessed more often than other areas. The AMT algorithms identify hot data at the I/O level, and the AMT algorithms promote it to the NAND layer to speed access.
HDDs provide plenty of sequential speed, and nearly unlimited write endurance, so caching sequential data would not be a good use of the limited eMLC layer. A quick test with sequential read/write data, to only 1% of the LBA range, confirms the drive does not accelerate sequential access.
We are admittedly making it very easy for the AMT algorithms to identify hot data by only reading and writing from 1% of the drive in this test. We have also included the original Turbo SSHD model as a point of reference.
The original Turbo SSHD identifies the hot data, and peaks at roughly 8,600 IOPS. The TurboBoost model reaches a peak of 10,200 IOPS, representing a hefty 16% increase. Both SSHDs feature the same density and 512e format, so this is a sign of architecture maturation. We also test 4k write speed, but its speed remains constant. In our standard 100% LBA tests, we will examine the benefits of the write back caching enabled in DRAM with the TurboBoost model.
With the determination that random read data is the only type of access cached in the NAND buffer, we expand our tests to 5% of the LBA range. This mirrors the amount of eMLC NAND used for caching. The larger access space illustrates slower performance from the NAND when it is full, and the increased load on the LSI processor necessary to manage a larger LBA table. This also shows a huge increase in performance, which is twice that of the original Turbo SSHD. The TurboBoost model tops out at 5,000 IOPS, and exhibits a much tighter range of performance than the previous model.
We experience a drastic increase in speed when we test with small LBA ranges and only one stream of data, so the task falls to us to attempt to trick the AMT algorithms, and simulate a more distributed workload with several data streams.
We created a complex multi-segmented test pattern with multiple data streams to test the efficiency of the AMT algorithms. We test 4K random read data with three data streams. The first addresses the same 5% of the drive we tested above, but only receives 80% of the workload. The second data stream reads from a larger 30% chunk of the LBA range with 10% of the workload, and finally, the third data stream reads from the entire capacity of the drive with the remaining 10% of the workload. This effectively forces the drive to ignore the "less desirable" read access, and cache only the most relevant data by simulating a hot band of data. The AMT algorithms still manage to pick out the hot band of data, and provide enormous acceleration.
We conduct our testing outside of the file system for numerous reasons. File systems are inefficient and bring forces beyond our control into the equation; such forces include metadata, buffers, and caches. However, for the purposes of the TurboBoost review, the file system can also introduce locality from file system metadata. In many cases, metadata is a primary bottleneck during typical use.
We test the same 5% LBA range tested above, where we topped out at 5,000 IOPS. We top out at 6,000 IOPS with a file system. Here we observe some of the negative results of testing with a file system with this level of granularity. The errata at the beginning of the test reminds us of the system caching and buffers brought on by NTFS, and these results should be taken with a grain of salt. However, they do exhibit the benefits of accelerating metadata.
PRICING: You can find products similar to this one for sale below.
United States: Find other tech and computer products like this over at Amazon's website.
United Kingdom: Find other tech and computer products like this over at Amazon UK's website.
Canada: Find other tech and computer products like this over at Amazon Canada's website.
Recommended for You
- We at TweakTown openly invite the companies who provide us with review samples / who are mentioned or discussed to express their opinion of our content. If any company representative wishes to respond, we will publish the response here.
Latest News Posts
- Intel's b2b shop could revolutionize PC game reselling
- GeIL reveal ASUS ROG certified EVO X ROG RGB RAM
- Seasonic reveal their PRIME Ultra power supplies
- Noctua to release two AM4-compatible low-profile coolers
- Battlerite to leave Steam's Early Access November 8
- Intel Core i9-7980XE and i9-7960X CPU Review
- Guardians of the Galaxy Vol. 2 4K Blu-ray Review
- Upgrading USB ports on top of case
- Areca ARC-8050T3 12-Bay Thunderbolt 3 RAID DAS Review
- GA-P67A-UD3P-B3 can't change multiplier past 38, can't change turbo ratio with i5 3570k
- AOC announces retail availability of AGON curved QHD gaming monitor
- Seasonic presents the PRIME Ultra power supplies
- EVGA announces GeForce GTX 1080 Ti FTW3 ELITE video card
- G.Skill releases AMD Ryzen-optimized Trident Z RGB DDR4 memory
- Hear the difference feel the beat of the DRUM