Benchmark of Cassandra, HBase, and PNUTS with MySQL as a reference point.

Presentation

Research Document

This is the most profound youtube video I have ever watched. It’s 90 minutes, and it’s filled with things I never knew. For example, it explains what the Atkins Diet and the typical Japanese diet have in common that helps you stay thin. If you don’t want to watch all 90 minutes, the section from 45:00 to 1:09:00 explains how fructose causes all sorts of health problems. If you still don’t want to watch this, here are a few quick facts and teasers:

• There is an epidemic of obese 6-month olds, so it’s not just laziness that causes weight gain.
• Fructose gives you calories without making you feel full.
• Fructose creates almost as many health problems as ethanol.
• Fructose is converted into much more fat that glucose, since fructose does not trigger insulin production to use the sugar, so it just gets stored. You aren’t going to succeed in a low fat diet that is high in fructose.
• High Fructose Corn Syrup (HFCS) and white sugar (sucrose) are equally as bad for you.
  • HFCS is a mixture of approximately 50% glucose and 50% fructose.
  • Sucrose is a molecule made up of a glucose molecule and a fructose molecule bound together, so it is exactly 50% fructose.

This presentation is done by Robert H. Lustig, MD, Professor of Clinical Pediatrics, in the Division of Endocrinology Director of the Weight Assessment for Teen and Child Health (WATCH) Program at the University of California San Francisco.

Since it such a long video, you may want to download it.

Chrome appears to have some fairly serious bugs that can lead to deleted bookmarks. I tried deleting a single bookmark in the bookmark manager, and it ended up deleting an entire folder. Fortunately, there is a Bookmarks.bak available until the next time you quit chrome. This wouldn’t have been such a big problem if I hadn’t already been fighting bugs with the Xmarks bookmark syncing plugin for Chrome. I have now turned on Chrome’s built-in bookmark syncing, but I still use Firefox quite a bit, so I wish Xmarks and Chrome weren’t so buggy. This is why I normally avoid beta software like the plague.

http://www.google.com/support/forum/p/Chrome/thread?tid=56d0b36c2e503bd6&hl=en

http://www.osnews.com/story/22872/Linux_Not_Fully_Prepared_for_4096-Byte_Sector_Hard_Drives

This is a like a CSI episode on flash memory.

http://www.bunniestudios.com/blog/?p=918

Currently, PostgreSQL replicates a database WAL (change log) files to the slave databases, which made it more difficult to get slave databases to have less than a minute of lag. This wasn’t a big deal, since before 9.0, which is in development, you couldn’t query the slave databases until you failover, which stops replication, and that’s a rare event. With streaming replication, it will be possible for the slave databases to be as little as one second out of date, which makes them much more useful for offloading read-only queries from the master database.

http://www.depesz.com/index.php/2010/02/01/waiting-for-9-0-streaming-replication/

It’s still not clear what the root cause of the sudden unexpected acceleration in Toyotas is, but this article at least gives us an idea of how widespread the problem is. Toyotas for 2008 are between 10 and 16 times as likely to experience this problem as GM, which had the best ratio of number of cars sold to the number of complaints. The reason for the range between 10 and 16 is that Toyota received a significant jump in complaints due to media attention. Of course, Toyotas are generally rated as having higher reliability than GM cars. Also, not all of these incidents result in accidents. Therefore, it would be nice if there was a comparison of how often defects in cars from a given automaker cause an accident.

http://blogs.consumerreports.org/cars/2009/12/sudden-unintended-acceleration-sua-analysis-2008-toyota-lexus-ford-gm.html

This is a very insightful article on the very different issues that you will encounter when you put a large application in the cloud and how common solutions to these problems can’t be applied.

http://gojko.net/2010/01/25/designing-applications-for-cloud-deployment/

While PostgreSQL has had trigger based replication for a long time, it can be difficult to set up and maintain. Since 8.1 or 8.2, it has had warm standby, where you could fail over to a slave but not query it while replicating. The code for hot standby has been committed to the 8.5 development branch. From what I can gather from the mailing lists, 8.5 will probably be released around June.

http://www.depesz.com/index.php/2010/01/08/waiting-for-8-5-hot-standby/