Distributed vs Centralized – What’s Right for You?

Distributed vs Centralized – What’s Right for You?

If you are considering consolidating your unstructured data in the cloud, and you should be, then you have probably seen solutions from both Panzura and Nasuni. The rapid growth of data, combined with unrivaled economics, scalability, and durability, has made the cloud the only practical solution for unstructured data storage and management.

At a high level, solutions from Panzura and Nasuni may seem similar, but in reality, they are very different. Look under the hood of these companies’ products and you will see that each has taken an alternative approach to solving their customer’s challenges. Panzura and Nasuni are two companies looking to solve the same problem, but with each taking a fundamentally different architectural approach that yields very different results.

Both companies offer solutions that store unstructured data in the cloud. They both have a file system that they use to manage data, and both use cloud storage to house the authoritative copy of the data. It is the way in which each company does this that makes a world of difference.

Let’s take a look at what the fundamental architectural difference is, and what impact it has on those using it. We’ll start with Nasuni.

Understanding the Centralized Approach to a Global File System
Nasuni uses a global file system that they call UniFS®. According to the Nasuni white paper, UniFS – A True Global File System: “UniFS keeps the mapping between the volume structure of files and directories, associated metadata, and objects in the cloud.” The white paper also states, “The Global Locking Service runs in the cloud so locking is always available and scalable to the global enterprise without dependency on any device to maintain lock states.”

What this boils down to is that with the Nasuni UniFS global file system, the metadata, lock manager, and all of the other services run in the cloud as a part of a centralized architectural model. It makes sense that Nasuni would run their global file system in the cloud as it provides them with several benefits. Creating a global file system in the cloud is fast and easy, as a result, it offered Nasuni the shortest route to market. Plus, since everything is run in the cloud, it is protected by the cloud provider and enjoys the high availability that only the cloud can offer.

Impeding Performance
Of course, as with anything, there are also drawbacks to this model. The first drawback is that it is subject to any incurred latency between the user and the cloud. Data may be cached locally, but communication with the lock manager must occur over the network. There will be increased latency as the lock request goes from the user to the cloud, is processed in the cloud, and then sent back over the network. A centralized lock manager can also impact performance in larger deployments as it may have to work through a large number of locking requests coming in from individual controllers. This can ultimately limit the performance of the lock manager and impact the overall performance of the file system.

A Risk of Data Corruption
The next issue with using a centralized model is that an interruption of access to the cloud, whether due to a local network, cloud provider issue, or anything else, can introduce the opportunity for data corruption. If there is a network interruption and users can still read and write to data without connecting to the lock manager, then new versions will need to be created. When connectivity is restored, then the new version will either overwrite the existing version or have to somehow be unified. Anyway you look at it, creating multiple “versions” is just another way of saying “data corruption.”

The Potential for User Lock Out
Another big issue is that since all the services are running as a centralized function in the cloud, if there should be an issue with the service, users can be completely locked out. This has been an issue with the Nasuni solution. They had to publicly comment on at least two occasions in 2016 when there was an interruption to their services that impacted customers. This is an inescapable problem of the centralized model.

An Introduction to the Distributed Approach and the Panzura Cloud File System (PCFS)
In contrast to a centralized model, Panzura uses a distributed model in the Panzura Cloud File System (PCFS). With a distributed model, a complete copy of all the metadata is maintained on each individual filer as well as in the cloud. The single authoritative version of the data is located in the cloud so that users can take advantage of the cloud’s economics, scalability, and durability. Since all of the metadata, as well as the active data, are local to every user, Panzura is able to deliver true data center performance at every location. If there is a loss of connectivity to any location, for any reason, then all of that location’s data and metadata are safe and secure.

Again, as with anything, there are also challenges to this model. A true, distributed cloud file system can offer the highest performance and scalability, but it is also the most challenging solution to offer from a technical standpoint. The Panzura Cloud File System is the only hardened, mature file system that has consistently been proven to meet the performance and reliability needs for distributed, enterprise-class organizations.

Prevent Data Corruption
With a distributed file system, each filer within the global namespace must be in contact with all others to maintain consistency. Fortunately, Panzura has been able to overcome this challenge with our highly efficient, patented SmartSync™ technology. With SmartSync, Panzura Freedom Filers are able to directly communicate with each other, either in a peer-to-peer model or through the cloud. They communicate without impacting network performance, and share atomic metadata updates in real time. Every Panzura Freedom Filer is consistently up to date with the metadata of the entire file system. In the event of a failure in network connectivity, users will have access to a read-only version of the active data. This enables users to continue to access the data without ever causing versioning issues or data corruption.

Deliver Data Center Performance Worldwide
With the unique ability to offer consistent high performance across more users, in more sites, the Panzura Cloud File System is the purpose-built distributed file system to manage unstructured data in the cloud. Only the Panzura Cloud File System has been proven to consistently deliver the economics, scalability, and durability of the cloud while enabling data center performance for the enterprise.

Key Takeaways
While Panzura and Nasuni may seem to offer similar solutions, there are significant and fundamental architectural differences between them. The Panzura Cloud File System underpins the Panzura Freedom Family. The result is a mature, proven, and massively scalable enterprise-class solution ideal for consolidating unstructured data for large distributed organizations. As a centralized product, Nasuni is well suited to smaller deployments seeking shared cloud storage and backup.

Leave a Reply