Cloud storage: Amazon vs Azure vs Google vs IBM Softlayer

Differences in cloud storage offers

There are some key similarities between the biggest four cloud providers in terms of their storage offerings.

They all offer an object storage platform, for example, which is accessible from within the provider’s cloud or the customer datacentre; file storage, but Google’s can only be accessed from compute instances within Google; and block storage, but generally only to applications and virtual machines (VMs) that run in the provider’s cloud.

However, big differences arise when it comes to the range of storage services available.

Amazon is probably the most developed, with offerings that span object, block and file, along with long-term Glacier cold storage for archives and as a tape replacement.

It is worth noting that the storage services on offer are not limited to those delivered under a provider’s brand. There are many third-party storage options available, with the likes of Dell EMC and NetApp making available iterations of their products in the cloud. Once again, Amazon has the greatest number of options available.

This could be a key decision point because the range of products and services available from a provider may be important to what you’re trying to build, especially as multi-cloud operation may not be a simple matter. In its Simple Storage Service (S3) product, Amazon has effectively produced a de facto industry standard cloud object storage protocol.

When it comes to compliance, all providers’ storage services need to fit in with key US and European regulatory frameworks, such as PCI DSS and HIPAA, and European Union (EU) regulations.

Another factor that may affect the choice of provider is geographical availability. Amazon has 55 “availability zones”, which group datacentres together in a region to build resiliency. Meanwhile, Azure has 52 “regions” and Google has 49 “zones”.

Amazon

Amazon Elastic Block Storage (EBS) provides highly available low-latency block storage for applications that run on Amazon Elastic Compute (EC2). It is aimed at relational and NoSQL databases, big data analytics and data warehousing applications.

It offers the choice of flash storage or spinning disk, and features include encryption, snapshots and access control.

Amazon Elastic File System (EFS) provides a file system interface aimed at workloads that include analytics, media processing, content management, web services and container storage. It allows file storage for EC2 compute instances – multiple if required – and can work in tandem with on-premise servers to allow cloud bursting.

Amazon Simple Storage Service (S3) is object storage in the cloud, and like object storage in general is useful for unstructured data in less performance-hungry applications. The service offers 11 nines of durability and scalability to trillions of objects.

Key targeted use cases are backup, archive, internet of things (IoT) data, analytics, data warehousing, user content such as images and videos, and website content. S3 also supports multiple methods of encryption and compliance with key US and EU data protection regulations.

Amazon Glacier is an archive service for infrequently accessed data that’s targeted as a replacement for on-premise tape, with costs claimed to be as low as $0.004/GB per month.

Data is effectively nearline at best, with three classes of recovery offered that range from minutes to several hours for bulk restores. You can query data in place, however, and only retrieve the subset you require. Data is distributed across geographical zones and compliance is offered to key US and EU data protection regulations.

AWS Storage Gateway is a software appliance that offers local storage and the ability to migrate, tier or interoperate with data held in Amazon cloud storage. It is aimed at use cases including migration, tape replacement, cloud bursting, tiering, backup and disaster recovery (DR).

Microsoft

Microsoft Azure Blob Storage is Azure’s object storage offering, aimed at file serving to browsers or clients, audio and video streaming, backup, archiving, DR and analytics (on-premise or Azure hosted).

The scheme for Blob storage is that data is held as a “blob”, of which there can be an unlimited number inside a container, which is, in turn, held inside the user’s storage account. Access can be through URLs or an Azure Rest application programming interface (API), PowerShell, command line interface (CLI) or Azure client library.

The Azure Files storage offering allows access through server message block (SMB), the Rest interface or from Azure storage clients. It can provide access to multiple applications and VMs per volume and can be accessed from Microsoft Windows, Linux and Mac clients from anywhere, through a URL.

Access control through Active Directory is planned, but is not yet available. Currently, shared access signature (SAS) tokens can be generated to allow private access to specified files for specified periods.

Azure Disk Storage offers three tiers of performance – Premium SSD, Standard SSD and Standard HDD – and can be managed or unmanaged. With unmanaged disks creation, management and scaling are done for you; while with unmanaged disks, you need to keep tabs on, for example, the use of storage accounts so as not to restrict VM performance.

Azure disks are virtual hard drives dedicated to a VM and aimed at providing persistent data to VMs that is not needed to be accessed from elsewhere.

Google

Google Cloud Storage is Google’s cloud object storage offering, which it offers in a range of four classes from high-frequency access to low. It is aimed at use cases that include media streaming, analytics, and backup and archive data.

Google Cloud Filestore is Google’s file access storage, which is currently in beta. It provides network-attached storage (NAS) access for Google Compute Engine instances, with capacity and input/output operations per second (IOPS) tuneable to customer requirements. Workloads targeted are media processing, web content and home directories.

Google Persistent Disk is Google’s high-performance cloud block storage offering, aimed at VM and container storage for instances running in Google Compute Engine or Google Kubernetes Engine.

Use is only chargeable for capacity and not for IOPS. Flash storage and spinning disk hard disk drive (HDD) is offered, with volumes of up to 64TB (terabytes). Multi-reader mounts allow many VMs to access a single Persistent Disk and data protection is through snapshots.

Cloud Memorystore is an in-memory data store for use with applications that need sub-millisecond latency access to application caches. It is currently optimised to work with the Redis open source database management system (DBMS). The service is offered in storage tiers and is charged by the gigabyte per hour.

IBM

IBM Softlayer Block Storage offers capacity and performance-optimised disk that can be provisioned in increments up to 12TB and 48,000 IOPS. Snapshots and replication are available, as is replication.

IBM File Storage offers network file system (NFS) access to NAS volumes from 25GB up to 12TB, with up to 48,000 IOPS of input/output (I/O) performance. File Storage can be purchased in four performance tiers, defined in terms of IOPS from 0.25 to 10.

IBM’s Cloud Object Storage offers 11 nines reliability, with data sliced and diced across multiple locations. For compliance reasons, you can choose to store data across regions or within a region, or even a single datacentre.

You can choose between three different classes of object storage – Standard, Vault and Cold Vault – depending on how frequently data needs to be accessed. Query-in-place functionality allows analytics to be run on data in IBM Cloud Object Storage.

IBM also has a cold storage archive service in preview, for which it claims costs as low as $0.002/GB per month.