WOA Issue 29

WOA Issue 29

In this issue

  • Works on Arm Office Hours
  • Supercomputing 2017 (SC17)
  • Red Hat Enterprise Linux on Arm
  • Debian 8 and Debian 9 on Packet’s Arm based Type 2A
  • Hardware: Cray XC50 with Cavium ThunderX2, Cray compilers
  • Hardware: HPE Apollo 70 with Cavium ThunderX2
  • Hardware: Bitscope Raspberry Pi cluster at LANL (Gary Grider)
  • Software: Ubuntu release of Docker 1.13
  • Patch: arm64: Add software workaround for Falkor erratum 1041
  • Opinion: AWS bare metal?
  • Colophon

Works on Arm Office Hours

Ed Vielmetti (Works on Arm project manager) is hosting Works on Arm office hours online on Wednesdays. The first event was on Wednesday, November 15, and saw participation from Ed Maste (FreeBSD Foundation), Jacob Smith (Packet), and Dieter Reuter (Hypriot). There is a dedicated Zoom channel at Packet available, and chat is available 24×7 through the #worksonarm channel on Packet’s Community Slack and on Freenode’s #worksonarm IRC channel.

Thanks to out to Mario Loria (Liquid Web) for showing the way with a great telecast for the Kubernetes #office-hours team – I’ll be learning the tech from the Kubernetes team which seems to have these real-time chat systems figured out.

Look on these channels for the details of the Wednesday November 22 event. Still trying to sort out best times to cover the world – there’s no one time that’s good for everyone.

Supercomputing 2017 (SC17)

Most of the news this week is coming out of the SC17 conference in Denver, which saw a huge amount of news related to Arm in high-performance computing. There was representation from Red Hat, SUSE, Cray, HPE, IBM, Cavium, and Qualcomm.

Arm’s Community weblog has a great roundup of all the news, which is worth a look.

If you are following the server or HPC space closely, then you probably have seen the volume of Arm server news go well beyond 11 in the past week. Most of the news is coming straight out of our server ecosystem which has never been more active and robust.

If you are a reporter or analyst and have any questions for Arm or need a contact for one of our ecosystem partners, please contact Allison PR.

Red Hat Enterprise Linux on Arm

Red Hat Enterprise Linux (RHEL) is now supported on arm64 systems with RHEL 7.4. The initial support is expect to target new systems from HPE. Jon Masters, Red Hat’s “Captain Arm”, made the announcement at SC17.

Today, after 7 years of work, Red Hat formally announced Red Hat Enterprise Linux for Arm. I am a proud father and very happy for the entire team: we did this together, all of us. I am forever grateful to you for your work, and to my execs for believing in another of my crazy ideas early on.

Debian 8 and Debian 9 on Packet’s Arm based Type 2A

As part of the Works on Arm project, Packet announced availability of Debian 8 and Debian 9 on their Cavium ThunderX based “Type 2A” servers.

Hardware: Cray XC50 with Cavium ThunderX2, Cray compilers

Cray customers will have an Arm-based supercomputer with Cray’s HPC software stack and optimized compilers. The XC50 targets scientific workloads.

Building on Cray’s adaptive supercomputing vision, the Cray® XC® series integrates extreme performance HPC interconnect capabilities with best-in-class processing technologies to produce a single, scalable architecture. Understanding that no single processor engine is ideal for every type of user application, the Cray XC50 system offers the addition of the 64-bit Cavium ThunderX2™ processor with the stability and performance of the Cray Linux® Environment, as well as a premium programming environment in a proven, scalable infrastructure, making the XC50 system the first production Arm® supercomputer available.

Hardware: HPE Apollo 70 with Cavium ThunderX2

At SC17, HPE announced a ThunderX2 based HPE Apollo 70 system.

The Apollo 70, using Cavium’s 64-bit ARMv8-A ThunderX2™ Server Processor, is purpose-built for memory intensive HPC workloads and delivers up to 33 percent more memory bandwidth than today’s industry standard servers. The Apollo 70 also provides access to HPE’s partnership ecosystem delivering key HPC components including Red Hat® Enterprise Linux®, SUSE® Linux Enterprise Server for ARM, and Mellanox® high-speed InfiniBand & Ethernet fabric solutions.

Hardware: Bitscope Raspberry Pi cluster at LANL (Gary Grider)

At SC17, Bitscope showed off a 750 node, 3000 core Raspberry Pi cluster, designed for Los Alamos National Labs as a development and training system for massively parallel workloads.

The BitScope Pi Cluster Modules system creates an affordable, scalable, highly parallel testbed for high-performance-computing system-software developers. The system comprises five rack-mounted BitScope Pi Cluster Modules consisting of 3,000 cores using Raspberry Pi ARM processor boards, fully integrated with network switching infrastructure.

LANL’s press release on the topic quotes Gary Grider:

“It’s not like you can keep a petascale machine around for R&D work in scalable systems software,” said Gary Grider, leader of the High Performance Computing Division at Los Alamos National Laboratory, home of the Trinity supercomputer. “The Raspberry Pi modules let developers figure out how to write this software and get it to work reliably without having a dedicated testbed of the same size, which would cost a quarter billion dollars and use 25 megawatts of electricity.”

Software: Ubuntu release of Docker 1.13

Canonical has released Docker 1.13 for 16.04 LTS “Xenial”. Of note is that this fixes a long-standing bug where the system Go 1.6 version guesses wrong when choosing a page size.

Thanks to Michael Hudson-Doyle (Canonical) and Patricia Gaughen (Canonical) for the work necessary to get this fix backported.

Patch: arm64: Add software workaround for Falkor erratum 1041

A patch from Shanker Donthineni (Qualcomm) addresses an issue with speculative fetch on the Falkor cores in Qualcomm’s Centriq 2400 processor. The erratum is QCOM_FALKOR_ERRATUM_1041.

The ARM architecture defines the memory locations that are permitted to be accessed as the result of a speculative instruction fetch from an exception level for which all stages of translation are disabled. Specifically, the core is permitted to speculatively fetch from the 4KB region containing the current program counter 4K and next 4K.

Opinion: AWS bare metal?

Jacob Smith from Packet asks whether Amazon’s AWS service will offer bare metal hosting to augment its current virtualized services, and what that means for Kubernetes and bare-metal hardware.

Other Architectures Get a Boost – While I’m not convinced that an AWS product would let you truly touch the hardware, I think promoting a bare metal offering exposes what’s under the hood: and increasingly that means x86 and something else – be it Arm, AMD, OpenPower, etc. This will help software like Kubernetes accelerate its early investments into knowing the hardware (e.g. device tags for GPU’s) – helping users to extract value from a massive variety of hardware.


Written in a hurry using vim and Github. Thanks to Zoe Allen (editor), Scott Burns (kernel work), Jacob Smith (leadership) and Jeff Underhill (project sponsorship) for all their help in putting this week’s issue together. Coffee from Roos Roast on Liberty Street in Ann Arbor. Coworking provided by Workantile.

You can unsubscribe at any time.

Explore Topics