Ep. 107 Distributed Data Show
Donnie Roberson of the DataStax Partner team joins the show to talk about the amazing performance results observed in running DataStax Enterprise 6 on Intel's latest generation hardware including the Xeon processors and Optane DCPMM, and when and where you might be able to get your hands on this technology.
0:00 - Jeff welcomes Donnie to the show and gets distracted by the drone races in the Accelerate expo hall.
1:54 - Donnie came to Cassandra from the Hadoop world and worked on the DataStax support team before making a move into into partner management and working on cool projects like the DataStax Docker images.
2:57 - The DSE 6.0 release in 2018 introduced a Thread Per Core architecture to allocate token ranges to specific cores within each Cassandra node.
4:17 - Intel has recently come out with the newest generation Xeon processors and Optane DCPMM. These two technologies are "a match made in performance heaven".
5:08 - Intel gave DataStax access to a private lab with 4 machines with this new hardware - 364GB RAM, 80 core Intel 2nd generation scalable processors, 2 Optane DCPMM 1.2 GB drives per machine.
5:51 - The workloads tested included 90% reads / 10% writes, 50% reads / 50% writes, 10% reads / 90% writes. On the write heavy test they observed 448 K ops/s (112 K ops/s per node). This is 2x performance over current NVMe drives. Read heavy workloads saw up to 5x performance impovement, with sub millisecond latencies.
7:18 - This hardware is available from Intel now, it will also be coming to Google Cloud Platform within the next couple of quarters.
8:11 - Jeff starts making up architectures on the fly involving Intel, GCP and DataStax Constellation.
9:10 - The highlight of the Accelerate conference for Donnie is the opportunity to meet Cassandra users and DataStax customers.
Developer Relations at DataStax