We developed Object Mount (formerly cunoFS) (https://www.storj.io/object-mount?hn=1) specifically to not rely on any metadata storage other than S3 AND preserve 1:1 mapping of objects to files AND support for POSIX. We have a direct mode that uses LD_PRELOAD to keep everything in userspace so no FUSE overhead.
This approach isn't right for every use case and juice might be a better fit for this sort of 'direct block store', but wanted to include it here for folks that might want something like Juice but without having to maintain a metadata store.
(Disclosure: I work at Storj that develops Object Mount)
When the cameras become a revenue stream for a city it is not a good thing.
Cameras have been installed to fine cars running red lights. The city then reduces the length of the yellow to catch more people and offset the high cost of the cameras. The shortened yellows cause increased crashes and fatalities.
Net-net the track record in the states is not great.
* I work in my unfinished basement, but often roam to recliners in the house because the basement is cold in the winter
* standing desk from IKEA
* 5K ultra wide LG monitor
* Thinkpad P1
* Microsoft ergonomic keyboard
* Regular run of the mill mouse
* Blue yeti microphone
* Currently using the laptop webcam for video
I get a lot of comments about being in an unfinished basement so I recently bought a green screen and some ring lights to try and improve my picture. The webcam on my laptop is too grainy for OBS to use the green screen, so looking to buy a new webcam.
Database indexes are definitely one of the most approachable concepts of databases for computer science majors. If you have studied data structures and Big O notation you have all the concepts needed.
While I agree that how a query planner works is one of the most ‘magic’ aspects, I think the output from the query planner in most databases is very approachable as well to regular common programmers and will get you quite far in solving performance issues. We know what full scans are (searching through every element of an array), etc.
The challenge is usually discovering that the jargon used in your database really maps to something you do already have a concept about and then reading the database documentation…
Matt Might was one of my favorite professors when I was in school. I hadn't learned about his son's passing until seeing this. Both Matt and his son are an inspiration to what can be done even in the face of the unknown.
I can't imagine the pain and anguish over the years that must have been felt as parents, so it was comforting to read about what kind of person Bertrand became.
I hope that if my son ever needs my skills, I would be up to the task so perfectly understated here: "Extending an artificial intelligence tool on the fly to analyze the data, Bertrand’s father found an unusual strain of pseudomonas as the likely culprit in Bertrand’s body."
Best of luck with the Bertrand Might Endowment for Hope!
S3 has been around for a long time and they had a lot of objects to transition when they upgraded, so I imagine that is why it took awhile.
There are other object storage systems that have strong consistency guarantees that came out after S3.
It greatly simplifies things that an object written to S3 it is immutable.
On a high level, all writes to your storage use some UUID. All reads use a consistent metadata storage (pick a modern database). After your write is complete and you are sure it is persisted, do the metadata update and return success. Everyone gets a consistent view of the operation.
Glacier can get pricey when you store/retrieve your data because of ops, retrieval fees, and egress (if going to the internet). We feel like immediate availability is a compelling advantage of our product.
How do you arrive at twice as expensive as AWS glacier?
AWS S3 Glacier is the same $0.004/GB/month and requires the same minimum of 90 days, but retrieval takes "from 1 minute to 12 hours". They also have large retrieval costs (on top of the regular egress to the internet) and API call charges.
AWS S3 Deep glacier is only $0.00099/GB/month, but requires it to be stored 180 days and also have operation/retrieval fees. "For long-term data archiving that is accessed once or twice in a year and can be restored within 12 hours"
This is great feedback, we'll take this into consideration. You can always reach out to talk about specifics about your use case and how we can better meet it (there is a Contact Us button near the bottom of the page).
Precisely. Although we also believe you shouldn't have to worry about going bankrupt if you ever need to restore all your data or want to migrate somewhere else, which is where the 100% egress of what you have stored comes from.
Interesting thought thread on our pricing structure... It saves money if you intend to send the backup to 10x places every month. If you only intend to do it once though you don't come out ahead $10 * 3=$30 versus $1 * 3+$22.50 = $25.50.