Did you ever want to create a lease for a specific node? Did you ever want to create a lease that does NOT include a specific node? Ignore a node that has been reserved? Reserve a whole rack perhaps? Or just a few nodes but on the same rack? Then look no farther; here are five tips and tricks for node selection and node avoidance!
CHI-in-a-Box beta release, a new MPICH appliance, CUDA10 support for Ubuntu, and more inside! See what we've been cooking up this summer.
This month: a new firewall security feature for bare metal nodes, updates to official disk images, and a webinar you should take a look at if you want to learn more about wide-area layer-2 network provisioning with Chameleon!
Chameleon's DirectStitch capabilities enable isolated direct OSI layer 2 connections between tenant networks and external facilities, including other Chameleon sites.
Learn about the upgrades to the testbed we released in the run-up to the summer, including increased storage on our GPU nodes at TACC, more user-friendly options for stitching layer-2 circuits, and fully operational 100G nodes at CHI@NU!
Motivated by the opportunity to optimize the architecture of data transfer infrastructure, we recently prototyped an elastic architecture for data transfer on Chameleon Cloud in which the DTNs expand and shrink based on the demand...
I was going to make a joke about how April showers bring May flowers, but that seems to imply that April was a bad month, even if we were making it rain (or snow, as the case may have been in Chicago.) Anyhow, here’s what we’ve been up to this past month: multiple networks now allowed for nodes, complex reservations, and some major foundational upgrades!
Exploring custom high-performance drivers in specialized operating systems with an aim to scale HPC applications in order to meet the future needs of exascale computing has motivated us to build a high-performance InfiniBand driver for Nautilus (Aero-Kernel) and evaluate its performance. In this blog, we would like to share our experimental framework and results achieved.
Many things can go wrong when deploying a large cluster of nodes. We thought it would be helpful for some of you to include some useful tips and tricks for deploying 20+ node clusters when some of the nodes may cause you problems.
Happy April 1st! Did we really replace all of our nodes with GPUs? I'm afraid that such news is unsubstantiated and we'll get to the bottom of who is spreading such rumors. However, we have some cool new features around our object storage and floating (public) IPs, and some information about upcoming big changes inside.