Google surprised a lot of people with its very strong offering aimed directly at Amazon S3. It is clear the product managers offering the Google cloud storage product were very familiar with complaints people have about Amazon S3 and are attempting to address many of those complaints in this product. Let’s take a look.
First, let’s review Amazon’s four cloud storage offerings, partly because most people don’t realize they have four offerings. Most people know about Amazon S3 Standard, Standard Infrequent Access, and Glacier. The lesser-known product is S3 Reduced Redundancy. It’s very difficult to find a webpage on Amazon that discusses and compares all four of these products in one place. For some reason, the Reduced Redundancy product always seems to be on its own.
Before we continue, here’s a very brief object storage terminology lesson. A put is the S3 command for storing an object in an object storage system. A get is the command for retrieving an object from an object storage system.
The Standard Product is the original object storage offering that offers redundancy across multiple geographic regions. Infrequent access is essentially the same product, except you are agreeing to do a lot of puts and very few gets. In return, Amazon will charge you less per month and charge you more when you retrieve it. S3 reduced redundancy is also the same product as the S3 product, except that Amazon will not replicate your data to multiple regions, and the availability and durability SLAs reflect that.
Glacier however is a very different product that is meant simply as an archive. Puts in Glacier are just like puts in other object storage systems. Glacier gets, however, are very different. You must perform a two-step process that first specifies the object you want to retrieve and then gives Amazon a certain amount of time to retrieve that object from glacier and placed where the object can be downloaded. Then a second operation actually downloads it. Depending on how much you are willing to pay, the delay between the initial request and the beginning of the download can be a few minutes or as much is 12 hours. This means software interacting with Glacier must interface with it very differently than they interface with standard S3 offerings.
It can be very confusing on Amazon to figure out what you get and what you will pay when reviewing Amazon’s storage offerings. I discovered as much while preparing for a webinar that we did comparing the two offerings. Each offering can have a different level of durability, access time, and pricing. There can even be variations within the product of what you will pay based on what you do with the product. In contrast, Google’s offerings were much easier to decipher.
Google offers four products to match up with Amazon’s four products: Multiregional, Regional, Nearline, and Coldline. Unlike Amazon, Google discusses all four products on a single page, and each product has the exact same level of durability and access time, which is measured in milliseconds. There is a difference in availability between the different levels, with a high of 99.95 percent availability for Multiregional and a low of 99 percent for Nearline and Coldline.
Not only are Google’s products easier to understand, they are also easier to use. An object placed in Google storage is accessible the same way via the same get command, where Amazon uses different access methods for some of its tiers. Even if you place the data in cold storage (analogous to Amazon’s Glacier product), you can retrieve it instantly and via the same command that you would if it was in any of their other offerings.
This ease-of-use should also translate into ease of development, meaning products that wish to offer support for all different levels of the Google products only have to program to a single API with a single command set. People wishing to support both Amazon S3 Standard and Glacier, however, must write two different routines. This helps to explain why many vendors support either S3 or Glacier, but not both.
The comparisons aren’t always in favor of Google, however. Some of Google’s per gigabyte charges were significantly higher than Amazon’s rates, although it is important to note that as of this writing Amazon recently dropped its prices and Google has not had an opportunity to respond to that yet. Depending on how you use the products, these price differences could result in very different bills between the companies.
Google’s four cloud storage offerings match perfectly with Amazons four offerings, and are significantly easier to understand and use. The one concern that its pricing isn’t yet in line with Amazon’s pricing, but that should be an easy adjustment for Google to make. Products that are easier to understand and have bills that are easier to predict typically win out in a competitive market, so I would expect Google to make significant inroads here. The only thing that might hold Google back is some people’s concern about its history of creating and dropping products, so it has some work to do there. But then again, it wasn’t that long ago that Amazon just sold books.