Azure Archive Storage

Microsoft defines Azure Archive Storage as “highly available secure cloud storage for rarely accessed data with flexible latency requirements,” which makes this service a direct competitor of Amazon S3 Glacier. In essence, however, Archive Storage is another tier of Blob Storage, in which the price for storing data is significantly cheaper per 1 GB stored ($0.00099 per GB, compared to $0.0184 per GB in the Hot storage tier above). However, data retrieval is significantly more expensive ($5 per 10,000 read operations, compared to $0.004 in the Hot storage tier). The use case for this cloud storage type is quite clear: You can store large amounts of data you don’t plan on retrieving often—if ever—and simply delete the data when it’s no longer needed.

Azure Managed Disks

For non-production VMs, like those running in development and testing environments, you can use Page Blobs to store a VM operating system and its accompanying data. In fact, this was  the only way to provision a VM on Azure until recently, when Microsoft introduced Managed Disks. Microsoft now offers four types of managed disks: Standard (SSD and HDD variants), Premium, and Ultra.

• • • Standard: These disks come in SSD and HDD variants and are designed for general purpose VMs. 
      • • • SSD: These serve in production workloads that don’t require high IOPS and/or throughput. The smallest option comes with 4 GB and can be increased to up to 32 TB (or, as Microsoft calls these disks, from E1 to E80, where capacity doubles with each level). The maximum IOPS and throughput with this disk is 6,000 and 750 MB/s. 
          • HDD: The size of these disks varies from 32 GB to 32 TB (S4 to S80, in Microsoft’s language), with the maximum IOPS and throughput capped at 2,000 and 500 MB/s, respectively.
    • Premium: This tier offers only SSD disks, which have the same capacity as the Standard tier, but an increased IOPS (up to 20,000) and throughput (capped at 900 MB/s). This, of course, impacts price. However, Microsoft offers reservations for this disk type—the only downside is that reservations apply to P30–P80 types (1 TB drives and higher).
    • Ultra: This tier offers maximum configurability to end users, so when picking a specific disk size (from 4 GB to 32 TB), there’s a range of available IOPS and throughput values. This way, users can tailor their disk to the workloads running on top of these volumes. 

Azure Files

Azure Files is Microsoft’s implementation of cloud file storage, with support for SMBv3 and NFSv4. There are four different tiers: Premium, Transaction Optimized, Hot, and Cold. 

• • • Premium: Naturally, this tier is the most expensive and offers the best performance, since the underlying shared network storage is based on SSD disks tiered together. 
    • Transaction Optimized: The latency offered in this tier is significantly slower than in the Premium tier. However, this tier is suitable for workloads with a lot of transactions, and can be used with a plethora of database engines, even for production needs.
    • Hot and Cold: These tiers are mostly used for instances that need shared file storage, but are not under heavy load, like user home directories or traditional enterprise user file shares. When choosing between the two, keep the following in mind: 
      • • • Hot has higher storage prices than Cold, but when it comes to the operations you perform on the files (read, write, delete), Cold comes with a higher price tag.
          • You should therefore use Hot for shared network file systems with lots of operations. Use Cold for bulk-storing data that’s not accessed or changed frequently, but still needs to be available from multiple servers (like in back-up scenarios).

Cloud Storage on GCP

Last but not least, I will cover storage options on GCP. The most important is probably Google Cloud Storage (a direct competitor to AWS’ most popular service, Amazon S3), but we shouldn’t neglect Google Persistent Disks or Google Filestore.

Google Cloud Storage

Google Cloud Storage (GCP) is following in the footsteps of Amazon S3, as it’s an object cloud storage where data is organized into buckets. The difference is that Google offers dual-region and multi-regional buckets (where data is stored in several regions, like the entire United States, the EU, and Asia), whereas with AWS, buckets can only exist in one region. 

GCP offers four storage classes:

• • • Standard: For frequently used data (e.g., to be used as a file share for your enterprise)
    • Nearline: For backups and data accessed a couple of times a month
    • Codline: For disaster recovery needs (once a quarter, as Google suggests)
    • Archival: Equivalent to Amazon S3 Glacier and Azure Archive Storage 

Also, GCP supports hosting static websites inside buckets, as well as versioning, and can be accessed via a web browser, Google’s gcloud CLI, or Google Cloud SDK, which is available in all popular development languages.

Google Persistent Disk

Google Compute Engine or Kubernetes Engine instances use Google Persistent Disk to run underlying operating systems or store application and user data. This block storage service can provision volumes on SSD or HDD drives up to 64 TB in size, and IOPS and throughput are directly related to the size you choose. This means that if you want improved performance, you need to provision large volumes. By default, all Persistent Disks are encrypted, and users can choose if they want Google to manage the encryption keys or prefer to import their own. 

One significant advantage over the competition is that Google offers regional Persistent Disks, which means that disks can be available in all of the selected region’s availability zones. This comes in particularly handy if you want to use containers with persistent storage in regional GKE clusters, a feature not currently available on AWS or Azure. 

Google Filestore

Google’s file cloud storage, Filestore, offers two service tiers: Basic and the newly introduced High Scale, which is still in beta mode. 

• • • Basic: This tier allows you to provision instances with up to 64 TB of storage, backed up by HDD or SSD disks, depending on your needs. The minimum capacity is 2.5 TB, which will cost around $200 a month if you select HDD disks in U.S. regions. If you opt for SSDs, prepare for a monthly bill of more than $750. Obviously, this cloud storage type is for enterprises that require a large shared storage in the cloud.
    • High Scale: This tier offers from 60 TB to a stunning 320 TB of storage. With the minimum configuration of 60 TB, you get 90,000 read and 30,000 write IOPS, with 3,000 read and 660 write MB/s of throughput. This amazing performance comes with a very high price tag—over $18,000 in U.S. regions and more in the rest of the world. So, be careful when provisioning—accidentally misconfiguring your setup can really hit your cloud bill hard.