Lead, Follow, but Get Out of Data’s Way

Lead, Follow, but Get Out of Data’s Way

Posted in tech

You know you’re on to something when others start name-dropping you. As data explodes and innovative IT professionals start looking at how to overcome the complexity of yesterday’s storage silos, it’s no surprise to hear others comparing themselves to the modern data management platform we’ve pioneered at Primary Data. We believe DataSphere is unique, from its architecture to its capabilities – so let’s take a look at why we seem to be making those name-droppers so nervous about our vision.

David Flynn, the father of modern flash storage, ushered in a new era of application performance at Fusion-io. He founded Primary Data to fix the problems caused by decades of storage silos. Our CTO has never been short on vision, and with other great minds working with him at Primary Data, DataSphere has been architected to provide an intelligent, automated data management platform that overcomes the complexity caused by decades of storage evolution.

With an open architecture, DataSphere puts enterprises’ existing storage infrastructure to work, and makes it easy to adopt the cloud. It gives admins unprecedented insight into data activity, and the power to make strategic management decisions based on that information. The DataSphere namespace makes different storage simultaneously available to applications, and importantly, it doesn’t get in the way of data access.

Out-of-Band Architecture

DataSphere creates a global namespace by virtualizing data with an out-of-band architecture that separates the metadata path from the data path. This means that applications have direct (native) access to data, without passing through a gateway or agent. Individual and aggregate performance of all operations are accelerated, as data I/O can be distributed intelligently across multiple storage devices with file-level granularity. Metadata operations are performed on mission-critical ready DataSphere servers that deliver enterprise reliability, accessibility and serviceability without impact to ongoing I/O.

New storage is added simply by adding a mount point to the global namespace without any changes in workflow. Once storage is added to the namespace data is virtualized transparently. From the application’s viewpoint, everything works the same as it did before. Indeed, the data path continues to be direct between client and storage. The impact of data address (layout) lookups is negligible, thanks to the compound operations feature in NFS v4.2. With this feature, the typical number of round trips between client and DataSphere server is typically two. If an application is accessing data cached on the client—which is very common—the data is accessed directly, without the need to check with server at all. This far preferable to bump-in-the-wire gateway and caching products.

Based on Open Source Industry Standard Protocol

The DataSphere platform is built upon the open source industry standard protocol NFS, so customers don’t have to trust their data to a proprietary file system. NFS gives clients direct read/write access to storage, while enabling DataSphere to perform live data mobility and receive performance telemetry from clients it uses to ensure data meets IT-defined service level objectives. DataSphere’s vendor- and protocol-agnostic architecture enables customers to add storage of any type to the namespace, from flash-based servers to shared storage to cloud and object stores.

Metadata Analytics with Machine Learning

DataSphere gathers metadata intelligence in real-time to understand how applications experience storage (for example latency, IOPS and bandwidth). It also collects telemetry on the data that applications access, such as which files are open, closed, with modified dates and times, as well as any other metadata. Importantly, this information is collected from clients to finally make enterprises (and DataSphere) aware of how storage and network performance is affecting application workloads. DataSphere can remediate non-compliance to objectives automatically. DataSphere analytics even implements machine learning to optimize data placement over time, adapting to each enterprise’s unique needs. This machine learning-driven automation relieves admins of some of the most tedious yet frustrating daily tasks, such as figuring out why some critical app is slowing down, while handling complaints until things are working again.

Move Live Data without Disrupting Apps

The NFS Flex Files feature enables live files to be moved without impacting applications by non-disruptively replacing layouts. This enables data access and data integrity to be maintained, even as files are being copied, which enables customers to ensure service levels are met even in the face of unpredictable workload spikes and outages. This ability to tier data across storage in real time enables IT to solve the performance problems of mission-critical apps without requiring downtime.

Capacity-Free Software Subscription

Few customers want to be locked into software subscriptions based on how much data they have.  This is why DataSphere’s pricing model is based on how many instances are deployed to support a customer’s unique infrastructure needs. With this approach, customers can implement easy and predictable scale-out strategies without being penalized for having to manage a growing pool of cold or cooling data that may never be touched again.

Want to learn more about what makes DataSphere different? We can start by showing you what you’ve been missing using our new Data Profiler tool – no DataSphere installation required. To learn more about how DataSphere’s global namespace can improve data performance and availability while cutting storage costs, connect with us at deepdive@primarydata.com. And stay curious about why you keep hearing our name.


Contact Form

Channel Partner