Network Attached Storage (NAS) and file servers have long outlived their usefulness. They’re expensive and don’t scale easily, making them a poor fit for many use cases. Nonetheless, they’ve enjoyed a virtual monopoly in unstructured data storage. When 100TB was considered a large file environment, this was fine. But in the world of multi-petabyte storage pools, the market needs new alternatives. In performance-sensitive applications like transaction processing, traditional NAS still has a role. But most applications, like user home directories, require more moderate-performance access to large-capacity file data. This capacity-intensive market is something object storage vendors have had their eye on for many years, but most have failed to gain any significant share.
Replacing NAS with object storage makes a lot of sense at first glance. Object storage systems are more cost-effective and can scale further. And they integrate data protection, via erasure coding or replication, for built-in data durability. They also support multiple use cases like unstructured data storage, long-term data archive and native object storage for next-generation applications.
The New Demands of Unstructured Data
There is also almost no end to file data sources, justifying the need for cost-effective, scalable storage, most of which does not need high-performance access. Backup and archive, storage as a service and security/surveillance data are all examples of data sources that just need cost-effective long-term storage area with moderate access performance.
The problem facing object storage vendors is that their systems either do not support the native protocols of NAS (NFS and SMB) or they lack the table-stakes feature sets that IT managers expect, such as snapshots, WORM and true HA with active/passive controllers.
There is another option for customers seeking to store files on object storage: a separate gateway from a separate vendor. While many gateways do support greater functionality, this approach creates several challenges for the would-be customer. First, these gateway vendors have to carve out a market for themselves. Since basic protocol translation is relatively niche, most of them add high-performance options and act as an accelerator for existing NAS, in addition to converting between NAS protocols and object storage protocols. The result is a solution that is overkill for the primary use case; eliminating NAS from storing unstructured data that is not in need of high performance. These solutions also add considerable cost, erasing some if not all of the cost savings of going to object storage.
The second problem with a gateway separate from the storage is the two-vendor challenge; Upgrade of one component frequently means upgrading to another. The hassle of dealing with two vendors, as well as managing two separate support contracts, may cause the customer to go with a path of less resistance like upgrading their NAS.
The final problem is that once the data goes through the gateway to the object storage, it is often not accessible by anything else other than that gateway. That means other applications accessing the object store through S3 APIs can’t leverage the data. For example, if a video surveillance application stores the data via NFS or SMB, a facial recognition application that wants to access that data via S3 won’t be able to access it.
Introducing Cloudian HyperFile
Cloudian is an object storage vendor that provides a native S3-compatible object store. Their solution is available as software-only or as a turnkey appliance. In the past, they have partnered with NAS Gateway vendors to provide legacy protocol access to their system but notices many of the same challenges described above.
To help customers alleviate these concerns, Cloudian launched its own NAS controller. The Cloudian HyperFile is a scale-out NAS engine designed to lay onto its existing object storage architecture. It also provides converged data access, where one application can store the data via an NFS or SMB mount and then enable an object-based application to access it via the S3 API.
HyperFile has a scale-out architecture, and each node has an active/passive component so even within the nodes there is no single point of failure. It supports SMB, NFS and FTP access. For data protection and data management, it provides both snapshot capabilities and WORM volumes. For security, it provides client level encryption and authentication through LDAP and Active Directory.
For service providers, HyperFile is multi-tenant. The service provider can dedicate a NAS controller per namespace (think per customer). They can also have multiple namespaces per tenant. Once the controller is dedicated to a namespace, the service provider can guarantee their customers isolated and consistent performance. And, because the NAS controller can be deployed as a VM, it will be a simple matter to deploy additional controllers as new customers come on line.
For enterprises, the namespace can also be global, meaning that multiple controllers can access the same namespace at the same time. This multi-controller access to the same namespace provides scalable performance from a single namespace and increased redundancy.
The solution also supports geo-distribution. The organization can use a single namespace across multiple clusters at disparate sites. Data can be continuously replicated among sites (a built-in feature of Cloudian’s object store), allowing fast, concurrent access to files. The geo-distribution feature facilitates collaboration and provides an additional layer of protection from a disaster.
Another interesting feature is HyperFile’s data migration engine. One of the biggest challenges in replacing NAS with object storage is migrating the data from the NAS to the object store. Considering the capacity of some NAS infrastructures, there could be hundreds if not thousands of terabytes of files to copy to the object store.
HyperFile helps here by creating a seamless conversion. Hyperfile starts by mapping all mounted file systems to itself. Then, it performs a background migration of the data. New writes are sent to the HyperFile, reads redirected to the existing NAS if the data has not yet been copied to the HyperFile.
Object storage is the logical replacement for a significant portion of the NAS workload. It is cost effective, more scalable (ie, more manageable in large deployments) and better at retaining data for long periods of time. Storing data in an object store also better positions the organization for cloud and data center modernization initiatives.
The problem is how to cost-effectively provide NAS protocols without adding too much cost and complexity. HyperFile not only addresses those challenges it also helps with the migration process. And since Cloudian is S3 API native, it puts the organization in a position to accelerate its cloud and data center modernization initiatives.