Database index and ID uniqueness, How much contrast is enough, Ceph

Post date: Dec 11, 2016 5:41:31 PM

  • In on project I've been watching they've had interesting issues. They had composite database key build from A B C columns. Then there were a problem that they got some composite key collisions. To fix this clever engineers replaced A with locally generated running number. Usually it's static number generated by the data source. When the change was done, I had a feeling, this isn't a great thing at all. Because everything relies on ABC. Well months passed and then someone alerted about massive statistical anomalies. - Now if you're database guy, you should already guess what it is all about. - They checked the central database, secondary replication database and processing database replicas. All matched with flying colors. So what's wrong? Because the original data source wasn't easily available, checks against it was skipped for quite a while. One day they were checking that data again, but now it seemed strange. While checking the data for one day, it was like darn, I think I'm having déjà vu. Well, now you guess what had happened. Same data sets had gotten in the database for several times (partially). It was clearly evident when sorting data by B or C column. - Awesome. Now small clever workaround caused a major mess up. What should have been done? Well, the case where A B C collided there was D column which didn't collide. But it wasn't included with the composite key. It should have been. Then all the normal checks & processes would have worked. Yet of course it could have required fixing in multiple places. Changing one but very vital part quickly to remedy some problem, without thinking it completely through can and probably will lead to big problems. Also the question why A column was colliding in the first place was lack of coordination when deploying the data sources. Usually all systems are installed and configured so that A B C is enough, but in this case there was an exception and original A values were overlapping and therefore the column D became essential. - As being said, nothing new, business as usual. Minor things can lead to bad outcome and it just happens. But only thing I'm happy about was the fact that when the A renumber thing got enabled, I knew it deep down that it wasn't a good idea at all, even if I didn't exactly know what would follow.
  • It seems that the display technology is progressing and they're now working on HDR displays. This is exactly what I described earlier. Higher color and contrast range on displays. And I guess it's like with sound, at some point it's "more than enough". Let's see how long that will take. Yet in movies like The Chronicles of Riddick, it would have been nice to have dynamic range with massive upper limits. Instead of seeing the sun melting the rock (Moon Crematoria Prison), you could have almost felt it. It's like standing next to fresh lava stream. The heat of the sun, in the movie. It would have been awesome. You know the mid day feeling around equator on clear day, how the sun burns and it's almost impossible to adapt to the amount of light even if you're out for quite a while. It's just too much. That kind of feeling. If that's then combined with our military exorcises where you combat using star light alone under polar midnight conditions, it's going to be quite a shocker to move from scene to another. Getting from 0.0002 to 110000+ lux could be quite painful. But people can see in both conditions. Would the people love it? I think it would be rather painful and annoying. In case of scenes like the Riddick's, if technology allows it and director likes it, I might go as far as 4x - 10x that for short durations. Then it's intense. I can feel black plastic things burning and dark wool smells like burnt too. - Now we're talking! Actually the documentation I did read was only up to 10k cd/m^2 but sure it could be more in future. It's like the car stereo competition, who got the loudest one. Nicely the color production is going to be improved a lot to Rec. 2020, but that's not great yet. Which is nicely 75,8% color production, but there's still room for improvement. - There's a reason why some holograms, CD-disks, animals and lasers got so, ooh, wonderfully deep colors, which you don't see on displays. Maybe in future we do.
  • Ceph is being so slow that it causes operating systems to crash and fail. This is now being great at all. Aww, I really wish theres some kind of solution. Yet of course one easy solution is to abandon the platform and switch to another. But it's not easy solution to make either.