Hello
Any journalists present, or do the press releases just get published here automatically now?
An Apeiron Optane array has five times the throughput, more than 6 times the IOPS, and up to 38 times lower latency than an Aperiron flash array. In a technology demo Apeiron has compared the performance of its ADS1000 storage with NVMe flash drives and Intel Optane 3D XPoint drives. Global sales and marketing exec veep Jeff …
Someone did a 3 drive nas to guess what this would perform like if made into an SSD type solution, or used in servers or whatnot. They came up with the rough idea that it works great for workloads with very high queue depth and not much else.
It's a great tech. It's got great features. It just needs a little maturity to get into it's niche (Small cache between RAM and SSD/Flash) or get better performance (beat SSDs/flash and RAM hands down).
I can get that over to you...this particular customer was very interested latency for larger block I/O. We have the typical "4K" Read data as well. This is the block size we used to overlay internal Optane vs. Optane in our network. Networked Optane is actually faster through our network for a variety of reasons. The most critical point is that the Apeiron network is not adding any discernable latency whatsoever.
Actually it just shows that Optane is byte addressable, and therefore much better with files of any size. You'll notice the difference between the 2 nand bars is negligible w and w/out 256k files, but the difference between the Optane bars is huge. This is because nand has to address large blocks of data at a time, so hen nand gets smaller files, it's still having to erase and rewrite large portions, no speed bump. Optane can literally write 1 byte at at time. So, if you need to write a small file it's faster than if you have to write a larger file.
Congrats Apeiron on "breaking the benchmark" for storage, both the basic IO/sec and GB/sec measurements, which have been in place for about 50 years.
I would guess the next benchmark will be one of latency measured at the application (ie in user space, as in the way HPC communication is measured), and the winning number there will start at 1us for a storage read or commit-write. Any guesses who the competitors in this new age will be?
Meanwhile .. where folks buy systems with a set budget and are careful with RFP reviews, etc. I would think once we get to this point (and it is coming), there will still be demarcations. You won't be putting Petabyte level configs in Optane for a while. The nextgen Optane + NVMe IOPS per $ will be high but you won't get enough of it, etc. But yeah, SPC1 and SPC2 are pretty well trashed, eh? That or they will have to sort it based on tech and that would be ugly.