Blog‎ > ‎

Loop, HTTPS, UPS, WAL, Updates, R2S, DC, Dadada, Hack Attack, LTE, Follow, Object Storage

posted Jun 30, 2016, 10:57 PM by Sami Lehtinen   [ updated Jun 30, 2016, 10:58 PM ]
  • One more reason to have fast SSD servers? One hanged process used about 75000 IOPS 24/7 for months. Nothing new there. It happens. But in these cases automatically scalable cloud services could become very expensive. Because that kind of failed process uses 'infinite resources'. If you'll just give more, that will be used.
  • Is HTTPS privacy illusion? - Sure, but there's nothing new with that. It's just like encrypted email. It hides content, but doesn't hide traffic patterns. "The green lock doesn’'t mean you’re safe." - Safe from / for what? That's the question always.
  • Checked out bunch of APC and Eaton UPS devices. Online / Offline, etc.
  • Finally confirmed that the one project doesn't suffer anymore from exploding WAL files because there's ever lingering open read queries. - Phew - Finally confirmed fix. It took a couple of months to get this finally and totally fixed. But now it's all good.
  • Business Cluster in down town Oslo is pretty neat. Let's hope we'll get one in Pasila, Helsinki, Finland.
  • There's a warning that some programs do load updates over HTTP... No sh*t Sherlock! That's the norm and has been for ages. If someone starts to raise alarm about this every time, they'll be having a very long list of programs to alert about. Earlier there was a warning that some ASUS laptops do that too. Well, not surprised at all.
  • Telegram Raise-to-Speak functionality is quite nice. Now it technically allows "asynchronous phone calls". Neat.
  • Sonera's (Telia) Data Center in Helsinki Pitäjänmäki Finland will be biggest open data center in Finland. "Sonera’s future datacentre will house up to 200,000 servers, with a capacity of 24 MW and according to the company  will also be the most modern and energy-efficient datacentre in Finland." - Quote from Data Center News. It will be probably LEED and CEEDA certified.
  • Zuckerberg 'dadada' news were interesting. They mentioned in news article that because passwords were unsalted it took just three days to crack entire database. I really don't believe that. Salting doesn't help a lot if users use proper passwords like said here earlier. 40 charas random, if it's salted or not, it's sill about 128 bits strong. Which makes it highly unlikely to be cracked in three days. Sure, it's easy to hash once and check against all passwords, so salting is usually beneficial. Especially if users do use bad passwords. But in this case my comment is that there weren't actually ANY strong passwords in that dataset or they used hashing algorithm which was extremely weak and therefore it doesn't really matter if those are salted or not. If I use crc32 for passwords, sure it works like any hash, even if it isn't hash function. Finding a collision is quite trivial, if passwords are salted or not, doesn't really matter. - Actually this is the point when I started to think about this challenge. Which I just posted out of order to my blog.
  • Funny article about reality of data security? Sophisticated hack attack? Don't believe the hype. - I guess I'll pretty much agree with them.
  • Reminded my self about LTE Cellular Network related indicators. RSSI, RSRP, RSRQ, SINR, SNR, dBm vs dB.
  • Tinkered with Twitter & Facebook share & follow functions, without using JavaScript. The old school way. I don't really like the idea of using JavaScript buttons which spy on people. Only if you decide to share something the site of course needs to know about it.
  • Played a bit with OVH Object Storage - So far it feels much better than the BackBlaze B2 which was major disappointment with it's poor performance. But I'm pretty sure they're still working on it.