blasdel

Read the book first, in 2010 a UK man built one from scratch as best he could

http://www.thetoasterproject.org/page2.htm

https://www.amazon.com/gp/product/1568989970/

If he were an American, that time would put your father in La Société Charly Miller: https://rusa.org/pages/CharlyMiller

I had fun the whole time!

Mentally I don't really suffer anything, though I did cry like a baby while riding through misty valleys at dawn the third day. I had realized I'd crossed into Normandy, and got emotional from working out from first principles why I new that detail of french geography. In the first town I passed through in daylight the locals were hanging up large banners for their 75th Anniversary of Liberation by the US Army.

I had a few serious physical setbacks from my ankle, losing a cleat bolt, a wrong turn — but it was easy to run into old friends out on the course to regroup and make new friends along the way helping each other out of predicaments.

I lost a lot of my endurance abilities from Covid, so when I do it again I will definitely be shooting for a different goal: https://adrianhandssociety.com/

Haha I just kept gallivanting around Paris afterwards, I spent 10 days in France and cumulatively slept maybe 30h total.

I hadn't done much riding that summer because of a broken ankle and actually re-fractured it mid-ride from vibration/stress while descending into Brest. The biggest trouble was that it was impossible to get any ice for it, but I found nice self-adhesive bandages to immobilize it the next day and was fine completing the ride.

My personal fastest 24 hour split was about 600km, while riding the first half of the Paris-Brest-Paris brevet in 2019. I had my first sleep after that in Brest and took much longer breaks for the return journey for a total ride of 1240km in 78:02:21 (with ~26 hours spent stopped).

I'd done many 600km brevets before that comfortably in 35-37 hours with a ~4hr sleep in the middle, but at PBP there were thousands of other strong riders and it was fun to work together.

https://en.wikipedia.org/wiki/Randonneuring is a lot more "normal" of a way to ride your bike all day than the Ultracycling approach in the article. Instead of grinding out laps of a racecourse solo, ride from town to town on pleasant backroads socially.

A programmer that doesn't make websites is like a glass artist that hasn't made bongs

Randomized admissions were the implicit thesis of the satire we get the word from: https://en.wikipedia.org/wiki/The_Rise_of_the_Meritocracy

AWS EC2 | Seattle, DC | Systems Development Engineers

EC2 Recovery Systems is seeking hands-on engineers with a passion for designing and developing highly scalable, distributed software. Our systems monitor the entire EC2 fleet and automatically take actions that directly improve customer experience by maximizing the availability and reliability of compute resources. The team’s problem space is diverse and interesting and you will have ample opportunity to work with a variety of technologies in support of maximizing the health of all compute workloads across the EC2 infrastructure.

https://www.amazon.jobs/en/jobs/1350796/software-development...

https://www.amazon.jobs/en/jobs/1383738/software-development...

If you've received emails, Cloudwatch events, or SNS messages where the body starts with the string "EC2 has detected degradation of the underlying hardware hosting your Amazon EC2 instance" and wondered what those sparing words are describing, you will find these roles compelling.

Don't forget that Rome also has a wildly nonuniform interconnect between the core complexes, and the system integrator gets much less control over it than Intel's UPI links. When you really need to end up with a very large single system image at the application layer, the bigger architecture works out to be much cheaper than 256gb DIMMs or HPC networking.

8-socket CLX nets you 1.75x the cores, and 3x as many memory channels vs. a 2-socket Rome system. It also scales to a single system image with 32 sockets if you use a fabric to connect smaller nodes:

* 4-socket nodes: https://www.hpe.com/us/en/servers/superdome.html

* 2-socket nodes: https://atos.net/en/solutions/enterprise-servers/bullsequana...

That's 48tb of DRAM with all 128gb DIMMs, or 12tb+128tb when using 512gb Optane PDIMMs.

Now Tayne, I can get into: https://www.youtube.com/watch?v=maAFcEU6atk

> And that if there are evidence or facts that overturns their current belief, they'll gladly accept it. There are no such equivalence in religion.

It is explicitly the role of a Prophet, Messiah, Bodhisattva, etc. to do such things — introduce new evidence that overturns or convolutes current beliefs

The day to day practice of all established religions involves reasoning through a historical lineage of commentaries on existing evidence and continually appending to it with new instances. Religion exists at the intersection of jurisprudence (what should be?) and the scientific method (what is true?). Both Law and Science are human philosophies that directly descended from religious primitives.

When done with video it looks like a real-world raytracing demo: https://www.stratusleds.com/aerial-leds

The /mysql/tmp/greetings.txt trick was cute

but do kids these days not know about https://linux.die.net/man/1/wall ?

More reasonable VMMs use the word "hypercall" for these paravirtualized interfaces

In a previous life I wrote code that only ran in Cygwin, it depended on a dummy networking driver to talk to the WinCE core of a sonogram machine.

Running Windows VMs as kernels for Cygwin environments felt less weird than most modern container architectures.

Meltdown goes much further back than that, potentially all the way to the Pentium Pro (1995)

It was also independently implemented in POWER and ARM A75

5 years ago after a princess tea party I ended up with some tall silver tiaras full of sparkly gems, and I started wearing them on my bike helmets. A tiara fits perfectly into the spot for roadies to park their sunglasses.

The constant feeling of "hey my eyes are down here buddy" in public proves that it works!

A1 is launched under ARM's new "Neoverse" branding that might help to distinguish this class of products from Raspberry Pi expectations: https://www.arm.com/company/news/2018/10/announcing-arm-neov...

Yes T2 performs quite admirably! If your CPU-bound workload doesn't use much memory, smaller sizes of T2 Unlimited can be a much better deal than C4.

T3 gives you both threads of your physical core and commensurably more CPU credits per hour to utilize them. You may notice that even T3.nano offers 2 vCPUs where T2.small has 1 vCPU.

Thanks for the reminder, ironically I put a lot of work into launching that platform!

The c5, m5, t3, c5d, m5d, z1d, r5, r5d, m5a, r5a, a1, c5n, p3dn instance families all use the Nitro Hypervisor, which is based on Linux KVM code but offloading a lot of functionality to the Nitro System instead of using the normative QEMU setup.

Firecracker runs as a process in a customer-provided Linux 4.14 or newer kernel using the upstream KVM apis, that can be on an EC2 Metal Instance or on your own hardware.

There's a Github Pages FAQ describing why it was made and how it fits with other solutions: https://firecracker-microvm.github.io/

and a high-level design document about how it works https://github.com/firecracker-microvm/firecracker/blob/mast...

There's a new "Elastic Fabric Adapter" in preview for 100Gbps networking instances: https://aws.amazon.com/blogs/aws/aws-previews-and-pre-announ...

That will allow applications to use a range of supercomputing techniques like https://ofiwg.github.io/libfabric/

You're asking for unikernels, which have had a recent renaissance: http://unikernel.org/projects/

Some of the density advantages depended on Paravirtualization, which is quite difficult to make safe against Meltdown: https://xenbits.xen.org/xsa/advisory-254.html

AWS uses virtualization domains to separate customer workloads, and Meltdown required that they all be HVM to provide strong guarantees: https://lists.xenproject.org/archives/html/xen-devel/2018-01...

Your account's Lambda functions won't share a kernel with anyone else's

The modern instance families support Turbo Boost, which is the X% guaranteed and up to %Y extra best effort you desire. On many of the largest instance sizes you get P-state control which lets you "disable that best effort extra bit" — but most operating systems will by default use it to strategically burst some cores even more at the expense of others.

Many workloads really just need a significant fixed amount of memory, and only periodically need high-performance CPU. T2 instances are much faster than you'd intuitively expect, if your workload is very bursty you can easily beat C4.

The newer Unlimited model "just" removes the edge case of being throttled if you run out of earned CPUCreditBalance, so you no longer have to reason about its interactions with other scaling factors.

It also allows a whole new usecase: if your workload doesn't need much memory or high I/O but will utilize as much CPU as it can get, you can save big by picking smaller burstable instances with Unlimited.