Chameleon Changelog for October 2021

Dear Chameleon users,


Your ship has come in! At least, the ship that had a few racks of new hardware destined for CHI@UC. It had quite an exciting journey, doing donuts in the Suez canal for a few weeks, then taking the scenic route around Cape Horn, before waiting for a good Tetris player to clear some space in the port of LA.


New hardware arrives at CHI@UC! A few days ago we announced the installation of the first batch of hardware to accompany this new phase of Chameleon's evolution. This adds a lot of horsepower to the CHI@UC site, in particular adding more GPU and general compute capacity, but also adding InfiniBand to the site, enabling experimentation with disaggregated hardware! Besides InfiniBand, the site's network topology will also allow for richer networking experiments: in the past only a subset of nodes had more than one NIC networked to the testbed; for all new nodes, there are two 25G interfaces you can connect to your testbed networks, and we plan on adding a third once we wrangle some more cables.


This new installation replaces the aging Haswell inventory, the nodes that have been with Chameleon since the beginning. Don't worry, they're not gone, but rather entering a new phase: we're working with several partners to bring them back online as a separate Chameleon site for additional capacity!


We're only at the beginning of the new hardware buildout: Xilinx FPGAs are coming soon, and we're investing in more storage as well. Check out the blog post for a lot more details. In particular, we want to know what kind of SSDs to buy:

The storage_nvme nodes are currently “empty”, but we’re planning on adding the following mix of enterprise: Samsung PM1733, Intel D7-P5600, Kioxia CM6-R, SK Hynix PE8010, and consumer: Corsair Force MP600, Samsung 980 PRO, WD SN850, Sabrent Rocket 4 Plus. Please, let us know if these look good to you and/or if you have any other suggestions leave a comment on this blog by the end of the month: we will try to accommodate as many options as we realistically can but time is of the essence.

CHI-in-a-Box is now ARM-ready. In the past, all CHI-in-a-Box sites have for the most part specialized in x86 nodes. In order to support upcoming ARM64 hardware installations, we've formally added support for provisioning ARM nodes as part of CHI-in-a-Box! Importantly, your site can run a mixed bag of node architectures, allowing you to provide both x86 and ARM machines to your end users. Read the release notes for more information. We also put together a configuration guide for enabling this capability.

Simpler SSH access to your nodes. As you probably know, in order to connect to a bare metal or KVM instance on Chameleon, you need to use a private key to authenticate. Over time, you may accrue several keys, which you use for separate purposes, or via separate interfaces (e.g., the GUI versus Jupyter/Python). This would often lead to an instance getting a key that you perhaps didn't expect, making it more difficult to get started. Now, all Chameleon base images will automatically configure any keypair you have configured under your account on the Chameleon site. This means that whether you launch your instance from the GUI, CLI, or Jupyter, and whether you access it from home, work, or… your car (?), you should be able to log in as long as all your keys are under your user account. Hope that simplifies things for you!

More capabilities for containers on CHI@Edge. Raspberry Pis on the CHI@Edge platform can now be equipped with a variety of peripherals via the GPIO interface, and you can launch containers that can directly interface with the GPIO pins. We have also made it possible to get CAP_NET_ADMIN inside your containers in the event you need more advanced configuration of net ifaces inside the container. There is also now support for CAP_PERFMON, which will utlimately allow more powerful instrumentation of e.g. power and load metrics, but the edge device kernels largely do not support this quite yet. Ask for more details on the chameleon-edge-users group!

Provide funding sources explicitly at allocation request/renewal. To better serve your future allocation requests, we've updated the request page to include an additional form that lets you provide your funding agency, award, and grant name.

Add a comment

No comments