File in the cloud

Faster network connections and higher-performance cloud storage have allowed organisations to replace local file servers and NAS volumes with cloud storage. And with improved integration between on-premise hardware and cloud file storage, organisations can use cloud file systems as additional on-demand capacity for workloads alongside on-premise NAS volumes.

Use cases here include bursting to the cloud for peaks in storage demand, as well as longer-term applications such as document archiving, and backup and recovery.

The cloud’s advantages for file storage largely mirror those for other areas of cloud computing. These include scalability in terms of speed and capacity, on-demand or usage-based pricing, and removing the need for upfront capital spending, on hardware and datacentre space.

Cloud file storage brings other benefits too, including sophisticated storage tiering to match performance and cost with application requirements; redundancy, including storing data across multiple zones; and the potential for improved automation.

Although the cloud cannot match a finely tuned NAS system for raw performance, this matters less for file-based storage than it does for block storage. And data is increasingly being stored and processed in the cloud.

Azure

Azure Files is Microsoft’s main, managed cloud file storage option. Described as “serverless file shares”, Azure Files supports SMB and NFS for Windows, Mac and Linux clients.

Microsoft provides Azure File Sync to cache cloud file shares on Windows servers, which it says provides on-premise performance levels. Azure also provides persistent shared storage containers using NFS or SMB, via the Azure Kubernetes Service.

Tiering allows Azure Files users to match performance requirements and budgets by mixing SSD (premium) and HDD (standard) storage. SSDs support SMB Multichannel for improved performance, with between 2x and 4x gains in IOPS, Microsoft says.

Azure Files Provisioned SSD V1 and V2 support up to 102,400 IOPS and a throughput of 10.340MiBps, against 50,000 IOPS and 5,120MiBps for HDD Provisioned V2.

AWS

AWS’s primary cloud file storage offering is Elastic File System (EFS). It’s designed to be “set and forget” for use with AWS compute instances, and comes with three storage classes: standard, infrequent access (IA) and archive.

EFS Standard uses SSDs to provide “sub millisecond” latency, whereas IA and archive run in the low double digit millisecond range. But AWS claims IA is 95% cheaper than standard, and archive is 50% of the cost of IA.

AWS also offers FSx, a managed cloud file system for use with servers. FSx supports Windows File Server, using Windows’s native file system. FSx for Lustre offers supports for high performance applications, including via cloud GPU instances, with “up to terabytes per second of throughput”, according to Amazon.

Linux workloads are supported through FSx for OpenZFS, while AWS also provides file storage using NetApp’s ONTAP operating system for data migration and hybrid operations.

Amazon File Cache provides high-speed storage for bursting, while AWS Storage Gateway gives users the option of on-premise access to hybrid storage. 

Google Cloud

Google Cloud’s Filestore offers file storage with a range of performance options, up to 25GBps throughput and 920,000 IOPS, as well as capacities up to 100TB.

In addition, the hyperscaler supports VMs with VMWare certified NFS storage, through Filestore Zonal and Filestore Enterprise. Container workloads are supported through Google Kubernetes Engine, again on NFS for stateful and stateless applications.

Filestore Basic comes in HDD and SSD versions. HDD provides a maximum sequential read and write of 180 MBps and 1,000 IOPS with SSD reaching 1,200 MBps and 60,000 IOPS. Filestore Regional and Zonal tiers provide a throughout of 26,000 MBps and 920,000 IOPS. Pricing is based on the service tier, instance capacity and performance, as well as the region.

IBM

IBM’s cloud file storage is part of its Virtual Private Cloud offering. It bases storage on cloud NFSv4.1, with file share capacity ranging from 10GB to 32,000GB.

IBM offers customisable IOPS, with users able to change capacity and performance in use. This, IBM says, allows users to adjust performance and costs according to workloads. IOPS options range from 0.25 IOPS/GB for low-intensity workloads, through 2 and 4 IOPS/GB up to 10 IOPS/GB for the most demanding workloads.

File Storage for VPC is also compatible with both virtual and bare metal servers, as well as IBM’s watsonx AI technology.