Performance tip on 3D XPoint a.k.a. Optane memory

Kolen Cheung

I just read a comment by Phil Gons:

we strongly recommend installing Logos on an SSD (ideally an NVMe or even 3D XPoint). The performance improvements are well worth the investment.

That’s an interesting remark about using 3D XPoint (a.k.a. Optane memory.)

I’m a bit surprised by this. Comparing high end NVMe SSD to Optane memory, the NVMe SSD actually has more throughput when the queue depth is deep. Optane excels at random IO, among other things.

We know that Logos performs much better on SSD comparing to HDD, because HDD is a very poor random access device (comparing the IOPS) limited by the spinning speed which is limited by Physics.

But I’d be surprised that the kind of random IO is intensive enough to saturate a high end NVMe device such that Optane starts to make a difference.

Can someone explain when it is so intensive? And when is it the most intensive? (I imagine searching something from the whole library?)

I’d also love to see a standard benchmark tool within Logos. Say something you run, and it will give you the timing for different tasks and summaries it in different categories, and provide feedback to users as to which to upgrade and/or which features to disable to speed things up.

Scott E. Mahle

Running Logos on a convertible laptop with a 1 TB Intel® SSD + 32 GB Intel® Optane™ memory + 16 GB RAM. In general, this computer has been really fast at everything I’ve asked it to do. Well worth the investment! I completely understand why they would recommend it.