Water is a very weird substance. Melting ice at 0C to water at 0C takes a fantastic amount of energy: Water has a latent heat of 384J/gram. Compared to other materials that is pretty much at the upper end of the spectrum.

Warming up 1g of liquid water by 1C is the original definition of 1 Calorie, or 4.184 Joules in todays units.

The water on our planet is a buffer that can store energy if there is more energy incoming that the earth can dissipate, or release energy if the bilancing is negative. Since some time around 1990 the energy bilancing of our planet is positive and the oceans are capturing energy at an alarming rate:

Note to self: Ignoring the Catalina Nagscreens is done like this:

$ sudo softwareupdate --ignore "macOS Catalina"
$ defaults write com.apple.systempreferences AttentionPrefBundleIDs 0 && killall Dock

And that should make MacOS ignore Catalina until further notice. Further notice does look like this:

$ /usr/sbin/softwareupdate --reset-ignored

Catalina is here. It fixes many things, among them iTunes. On the other hand, it drops support for 32-Bit MacOS applications.

Consider the upgrade carefully.

$ system_profiler SPApplicationsDataType
…
    Steam:

      Version: 1.5
      Obtained from: Identified Developer
      Last Modified: 23.08.16, 20:32
      Kind: Intel
      64-Bit (Intel): No
      Signed by: Developer ID Application: Valve Corporation, Developer ID Certification Authority, Apple Root CA
      Location: /Applications/Steam.app
…

Anything listed with 64-Bit (Intel): No will stop working after the upgrade. Check before you upgrade, and consider what you will be losing access to if you perform the upgrade.

Deutsche Welle is shocked: Generation Greta is watching Netflix (Article in German Language), Netflix runs on computers, and apparently computers are using power.

Let’s have a look.

EDIT: Second part Streaming and Energy now available.

End-User Device.

It’s 2019. End user devices are using Wi-Fi, and are running on Batteries. They are Cellphones, Tablets or Laptops.

Devices that are not connected to grid are more usable if they are trying to save energy, and so all modern devices are full of special purpose hardware that allows them to fulfill their main functions more energy efficient. In particular for the use-case of watching video modern hardware, even cellphones, have special ASICs that can do video and audio decompression and compression on the fly faster and with much less energy usage than a CPU could do.

We had a discussion at work about MySQL doing Disk I/O to a NVME disk, and if it is valid to turn off the doublewrite buffer when using XFS.

TL;DR: It’s not, you can turn off the doublewrite buffer only on filesystems that never do in-place updates (ZFS, btrfs), or that have their own doublewrite buffer (ext4 with journal=data). A flash layer underneath the actual filesystem is likely not going to help you without additional measures.

So I have been playing this Internet Spaceships Game. No, the other one . It also has NPC Backstory, and on 13-Sep-3305 , a multiplanetary food crisis was predicted, as a prep for the introduction of a new tradeable good on 18-Sep-3005 .

Distribution of the fertiliser has started at Marshall Dock in the Riedquat system. We encourage traders to take advantage of our introductory prices.”

That went wrong.

So some people flew to Marshall Dock and looked at the prices for Rockforth Fertilizer . What they found was a single station buying and selling the good at the same time, and buying it at about 9000 CR higher than they sold it.

The last time I saw a MySQL server operating at a performance limit was in 2012. Back them we had a production master on (then) current hardware, running stable at about 21000 QPS. At 24000 QPS it tended to become unstable and fall over, dying in global locks on the InnoDB Adaptive Hash Index or other global locks.

I need to better understand how MySQL works today, and what the limits are on a box that is considered large in 2019.

Ein Thread über Konsenssysteme aus Twitter

Mehr als ein Key-Value Store

Heise schreibt: Verteilter Key-Value-Store: etcd erreicht Version 3.4 . etcd ist auch ein Key-Value-Store, aber das ist nur ein Nebendetail. Die Beschreibung der neuen Funktionen im Artikel macht auch schon keinen Sinn für einen KV-Store.

etcd ist ein Konsenssystem. Es realisiert Clustermitgliedschaft, verteilte Locks, und darauf aufbauende Dienste wie Service Discovery und Loadbalancing.

Es gibt drei von Kyle Kingsbury getestete Implementierungen von Konsenssystemen, die funktionieren: Zookeeper, etcd und consul. Es handelt sich um verteile Datenspeicher, die im CAP-Dreieck Consistency über Availability präferieren.

Saving an older twitter thread on Ceph .

I just have finished reading Part 4: RHCS 3.2 Bluestore Advanced Performance Investigation and now I do not know what to say.

Ceph set out with the idea to make storage a commodity by using regular PCs + a lot of local hard disks + software to make inexpensive storage. For that, you ran a number of control processes and an OSD process per disk (“object storage demon”).

While I was testing away on all the SSD and NVME toys, I got my hand on a test box with even stranger that usual equipment. It was a Dual Platinum-8280 box with a really weird amount of memory: 7.5 TB.

“8280 ”. This is a machine with a “2” in the second digit, indicating Cascade Lake , CLX.

And one thing that is new with CLX is “Intel Optane Persistent Memory Technology”. That is actually 3D XPoint (“Crosspoint”), a big invention in applied materials science. It is a thing that Intel and Micron have been co-developing, Intel sells it under the trade name of Optane (and internally it was named Apache Pass) and Micron has it as QuantX. So far I have only seen Optane.