Comments on: Backups /rants/backups/ sealed abstract class drew {} Sun, 27 Mar 2016 22:51:38 +0000 hourly 1 By: Stefan Reitshamer /rants/backups/comment-page-1/#comment-1893 Mon, 19 Jul 2010 01:11:38 +0000 /?p=636#comment-1893 Well, I did spend a lot of words pointing the features that S3 comes with, but I also mentioned it would be awesome if Amazon offered a bottomless RAID drive type of option for a lot less money. I still think it’s more than just dropping a NAS in a colo facility. Someone needs to monitor and repair the NAS, expand it, and probably accept disk drives full of data for quick initial backup.

There’s another problem as well: many people (such as some professional photographers) are producing bytes to be backed up faster than they can upload those bytes with their available bandwidth. I wouldn’t be surprised ifthe people with TB of data are generating data at that sort of rate. If you add in a large initial upload, you’ll never finish a backup. I think that’s where the emailing-a-disk feature becomes critical. Unfortunately disk space is still getting cheaper much faster than bandwidth (at least in the US).

By: admin /rants/backups/comment-page-1/#comment-1891 Sun, 18 Jul 2010 22:42:08 +0000 /?p=636#comment-1891 I don’t think 3TB is an outlier scenario, as I recently polled HN to assess exactly that question. I think the number of people who have 3TB is certainly small relative to, say, 20GB, but the type of people who have 3TB of data are much more likely to be interested in paying to back it up, because they have invested significant time into collecting that data. So I wouldn’t be surprised if the 3TB size is a more lucrative market, in spite of being smaller.

I fully agree with you that the reasons you’ve cited are why Amazon is a great deal more expensive than the alternatives. But I don’t think you’ve grasped the point: I don’t care about five nines of availability. I don’t care about redundancy. I don’t care about replication. We are not talking about serving content to customers here. If my backups go down a few minutes each day it’s no sweat off my nose. As long as they don’t fail at the precise time my machine fails, it doesn’t matter even if I have a 500MB data loss event per month, as long as it is caught quickly by the rsync script and written over with the right data. Amazon is selling apples; I want an orange.

By: Stefan Reitshamer /rants/backups/comment-page-1/#comment-1785 Wed, 07 Jul 2010 10:33:39 +0000 /?p=636#comment-1785 S3 gives you quite a bit more than a colo NAS. Amazon stores your files in multiple data centers, which means they replicate your files among data centers for you, and they monitor the storage at those data centers and replace failing hardware and re-replicate your files to the replacement hardware. Plus it expands automatically as your data set grows. Plus you can send a disk to Amazon for faster import (useful for the initial import). There are some interesting comments here about S3 vs colo:

Also, Amazon just announced a Reduced Redundancy option for $.10/GB per month (instead of $.15). It would be really cool if Amazon offered a single-data-center version that’s like a bottomless RAID drive, for a lot less money.

I was going to suggest that 3TB is an outlier scenario, but each person who takes lots of HD video (e.g. iPhone 4 users) proves me wrong.