Search results

I recently dealt with this for my group at work. We ended up going with a 16-port Raritan Dominion LX II for KVM-over-IP goodness (though the required dongles were as much as the unit itself--$1200). For reboots, we went with 3 of Dataprobe's 8-socket iBoot switched PDUs. That gives us 24...

Oof--I was mistaken about ITX boards with 25GbE. As BlueFox suggests, I was thinking of a D-1700 board, the D17006U-4Q--but that's an mATX board, not ITX. Maybe that was just wishful thinking on my part. http://www.asrockrack.com/general/productdetail.asp?Model=D1700D6U-4Q#Specifications

At work, I have four cluster nodes with 25 GbE that I’m about to connect via Mellanox’s host-chaining feature. We have a 10-12 workstations (also with 25 GbE cards) that would be great to use as cluster nodes on occasion. I’d love to set this switch up on an unroutable subnet for HPC traffic...

I take your point about having a lot of NVME drives in one machine. And yes, risers and dedicated cards will get you four drives in an X16 slot, and that can be a lot of data. But a 10-gigabit Ethernet link has a real-world transfer rate of one gigabyte per second, or about one third of what a...

To go a bit further, I wouldn’t mind M.2 cards becoming a de facto PCIE expansion standard for SFF machines. Plenty of M-ITX motherboards have 3 M.2 slots…ditching the standard 16-lane PCIe slot on those boards (used mostly for graphics cards) would allow *seven* M.2 slots if the motherboard...

The only switches I’ve found that support ECN also support PFC, which means I might as well run lossless ROCE. Is anyone aware of a reasonably affordable switch (ideally 25GbE) that supports ECN?

None of those assumptions are right, but they're pretty close! We're using dual-port 25GbE CX5s to build a four-node cluster for running finite element simulations using ANSYS. ANSYS solving speeds are mostly sensitive to latency, not bandwidth, so 25Gb ROCE is what we want. We're starting off...

Is it worth a try? I sure hope so. But as far as I can tell, host chaining is entirely undocumented. Posts asking for help seem to be met universally with “oh, it’s not working for you? Call us directly!” The first time I saw that, I thought, “what great tech support! But I’ve seen it often...

I'm a 3-5 weeks away from implementing a four-node ROCE cluster using Mellanox X5 cards' host-chaining functionality, so I'm curious about this too. For the code I run, latency is the the principal concern, and in theory Roce should be great for that. But hard benchmark numbers are thing on the...

A tip from another FEM-focused mechanical engineer: if hyperthreading is enabled on that machine, disable it. Nearly all FEM codes run faster with hyperthreading disabled. Most hardware-savvy FEA folks do this as soon as they get their hands on a new machine, but with three out of four sockets...