Amazon Glacier Pricing Changes and Retrieval Tiers
Amazon Web Services announced several changes on 21 December, 2016. They’ve reduced prices, they’ve expanded from 1 retrieval tier (3-5 hour delay) to 3 tiers, and they’ve replaced their confusing retrieval pricing with much simpler per-GB pricing.
Prices for Glacier storage were reduced in most AWS regions. Prices dropped about 23% in most US regions, 24% in EU regions, and 16-28% in Asia Pacific regions. Full pricing details are at the AWS Glacier pricing page.
Before, there was only 1 way to get your data out of Glacier. Request an object be made downloadable, wait 3-5 hours for AWS to make it downloadable, and then download it. AWS still offers that option — they call it “Standard” retrieval tier. But there are 2 other retrieval tiers now — Expedited and Bulk. If you request an object using Expedited tier, AWS makes the object downloadable in 1-5 minutes! As you might expect, the faster you want it made downloadable, the more you’re going to pay.
No More “Peak Hourly Request Fee”
You may have seen some of the blog posts warning about Glacier retrieval and the dreaded “peak hourly request fee” which AWS charged based on the peak rate at which you were making objects downloadable during a given month (how’s that for confusing?). No more! AWS has eliminated that fee. Instead, they charge a straightforward per-GB fee for the data you download. The price depends on the retrieval tier you choose. There’s no longer any danger of running up an enormous fee by requesting data too quickly.
Arq supports all 3 Glacier retrieval tiers
If you’re using Arq Backup to back up your files to Glacier, when you go to restore a folder Arq will ask you whether you want to use Expedited, Standard or Bulk retrieval tier. It’ll estimate the costs as well.
It’s still a 3-step process to get data from Glacier though. Arq has to request that objects be made downloadable, wait until they’re downloadable, and then download them. Because of this, Arq requests only enough data as it can download in the given timeframe. For instance, if you choose Standard retrieval tier, Arq requests approximately 4 hours’ worth of downloadable data. Four hours later, it requests the next 4 hours’ worth while simultaneously downloading the first 4 hours’ worth as it becomes available. It repeats this process until all the data you’ve asked Arq to restore has been restored. In this way, the requesting part stays 4 hours “ahead” of the downloading part. The same thing happens in Bulk mode except the 4-hour interval is 6 hours instead. For the Expedited tier the interval is 1 minute. In each case, Arq makes sure it doesn’t get too far ahead of itself in requesting objects.
As before, Arq will request data at a rate you choose, so choose a rate approximately equal to the bandwidth you have available (Arq tries to estimate this for you).
UPDATE 20 March 2017: We’ve since disabled Expedited-tier support in Arq because it turns out Glacier’s per-request pricing for Expedited is bonkers.
For a full explanation of Glacier pricing check out our AWS Glacier Pricing Explained article.
If you’ve got any questions or comments, send us mail at email@example.com or write a tweet to @arqbackup. We strive to respond promptly and help you any way we can!