With the conclusion of this year's re:Invent conference -- followed by a massive outage in AWS' US-EAST-1 region -- I decided it was time for a deep dive on a few of the cloud provider's recent announcements, particularly from an IaaS and PaaS perspective.
For many of the new services covered below, success will be measured in one of two ways: how much new data will land in AWS and whether the service generates a positive margin. But, as has been well-documented, once a service is launched in AWS, it truly never goes away -- just look in US-EAST and you can start the original M1 EC2 instance type.
Here's a brief re:Invent recap of the data services announcements I found most interesting.
AWS Marketplace expansion
AWS continues to build a marketplace that is more "choose your own adventure" than it is buying a service or product. The addition of AWS Marketplace for Containers Anywhere is interesting for on-premises container package deployments because it is not just about Amazon EKS Anywhere, but any Kubernetes cluster.
Think of an open Apple App store for Kubernetes, boasting more than 500 vetted container applications in the marketplace. For comparison, Google Cloud Platform has approximately 112 available applications, focused mainly on Google Anthos, Google Kubernetes Engine and GKE on-premises deployments. AWS Marketplace for Containers will likely be the most popular container service on AWS.
Amazon S3 Glacier options
The next announcement worth mentioning is the renaming of Amazon S3 Glacier storage class to Amazon S3 Glacier Flexible Retrieval, and the addition of Amazon S3 Glacier Instant Retrieval.
AWS claims the S3 Glacier Instant Retrieval storage class is the "lowest-cost archive storage with milliseconds retrieval" for rarely accessed data. However, enterprises need to perform an ROI on this storage option because "rarely accessed" may not mean the same to all organizations. The suggested rule of thumb is that data is accessed less than once a quarter.
The existing S3 Standard-Infrequent Access (S3 Standard-IA) storage class is $0.0125 per GB with a 128 KB minimum billable object, which is 128 KB base size, plus an additional 8 KB of index and metadata. Now, compare that to S3 Glacier Instant Retrieval at $0.004 per GB with 160 KB minimum object size, which is 128 KB base size, plus an additional 32 KB of index and metadata. Both options have instant retrieval in milliseconds with 99.9% availability per year.
So, why use S3 Standard-IA instead of S3 Glacier Instant Retrieval? It seems like a mathematical problem, and I'm sure there is an arbitrage in there somewhere.
If your head hurts after reading that, you're not alone. During the "Driving innovation and insight with cloud data on AWS Storage" keynote, Mai-Lan Tomsen Bukovec, AWS vice president, block and object storage, explained how AWS was "working backward from the customer" who had asked for a cost estimate for a "bulk" retrieval of five to 12 hours from Amazon S3 Glacier. The product team found it too hard to figure out how to provide the cheapest retrieval estimate. Instead, they decided to offer free bulk retrievals on Amazon S3 Glacier Flexible Retrieval.
It sounds like the technology stack is so complicated under the hood of S3 Glacier that computing the price estimate for retrieval is more difficult than just giving the retrieval away for free.
AWS Backup support for VMware and VMware Cloud on AWS
The next interesting item, on which I had input while working at AWS, is the extension of AWS Backup to VMs that run on on-premises VMware environments, as well as on VMware Cloud on AWS. This enables AWS users to deploy an AWS Gateway Open Virtualization Format, developed by the AWS Storage Gateway team, to connect to VMware vCenter. After a few minutes, users can see all the VMs managed by vCenter in your AWS Console, apply a backup plan and perform ad-hoc backups. Once the VMs are backed up, they can be restored to any VMware cluster or VMware Cloud on AWS.
The pricing will be very hard to compare if you currently use a vendor such as Veeam or Commvault. AWS Backup is priced on warm backup storage of at least $0.05 per GB per month, restores of at least $0.02 per GB and cross-region data transfer -- but I'm unsure how this applies to VMs. This is one place where ISV partners will once again suggest that AWS is trying to swallow the industry. My view is that the current backup vendors have enough differentiation in VM backup to fight this fight.
Additional file systems
Amazon unveiled FSx for OpenZFS earlier this month, bringing the number of its FSx file systems to four, alongside NetApp OnTap, Windows File Server and Lustre. You could call FSx the file system buffet, but why so many file systems? As it turns out, file systems are tough.
While AWS provides a page on how to choose an Amazon FSx file system, it will not make the decision that easy. For example, the page is a bit simplistic, suggesting that organizations use OpenZFS if they currently use ZFS or other Linux-based file servers. But if they have Linux connected to NetApp OnTap on premises does that mean FSx for OpenZFS isn't an option?
FSx OpenZFS provides NFS version 3 and 4 support, lower latency than all the other FSx filesystems, higher throughput per file system, higher max IOPS and lower max file system size. Also, deduplication or auto-tiering doesn't factor into users' decision. It will be very interesting to see where Amazon Elastic File System goes now that both FSx for NetApp OnTap and OpenZFS are in the portfolio.
Editor's note: Enterprise Strategy Group is a division of TechTarget.