With the recent EMC Atmos announcement, it is great to see momentum behind using Information Dispersal Algorithms or Forward Error Correction to increase data protection. An emerging number of vendors are beginning to agree with us that RAID and replication aren’t going to work for massive storage systems – particularly for storage clouds. And, we’ve always said we can’t be the only vendor out there who understands there is a better approach.
Using IDAs or FEC is the “how” of these types of new storage platforms, and tablestakes to get started. We focused our first three years on getting the “how” to work, and launched our dsNet system using IDAs into the market in February of 2008.
We’ve then spent the next two years building resiliency into our system.
Resiliency that…
- Keeps your business running regardless of demands and risks
- Resists hackers’ attempt at stealing your data, even if it’s not encrypted
- The cost of threatening data is way too high
- Can run despite a series of catastrophes causing multiple simultaneous failures of drives, devices and locations
- Provides data security for data at rest or while in motion
- Guarantees data integrity
- Provides an “always on” architecture
- Optimizes multi-site performance
- Enables secure collaboration and communication
Let’s look at some of the features our product has today that deliver on resiliency, and discussion points on what to look for in these types of storage platforms.
SecureSlice™ Protection
Feature: Any group of slices less than the threshold is unrecognizable as data.
Benefits: Secures data in a multi-tenancy environment. Hardware breaches do not compromise data.
Discussion: Make sure data slices, fragments, or packets are completely transformed and don’t contain any original data to ensure data security.
PerfectBits™ Assurance
Feature: Integrity checks on both individual slices and data files.
Benefits: Guarentees bit-perfect data storage and delivery. Guarentees data cannot be corrupted during file updates even under adverse conditions.
Discussion: Don’t rely on hardware for data integrity – make sure storage system addresses data integrity in software intelligence.
Tunable data protection
Feature: Set the Dispersed Storage to a multitude of M of N combinations.
Benefits: Match reliability to application requirements and ensure acceptable fault tolerance of storage nodes and sites.
Discussion: Make sure your solution has data protection tunability. Without tunability, fixed configurations may not map well to existing infrastructures.
Rebuilder™ Process
Feature: A distributed rebuilding process works across all storage nodes.
Benefits: No single point of failure or choke point assures highly scalable reliability.
Discussion: Make sure your solution has a distributed rebuilding approach so that storage nodes can handle rebuilding versus taxing servers also responsible for access throughput.
SmartRead™ Performance
Feature: Predicts the optimal network routes and storage nodes for reading the minimal number of slices to most efficiently return data.
Benefits: High level of read performance guaranteed even with failure conditions. Minimizes network traffic.
Discussion: Look for solutions that optimize performance for distributed storage systems by ranking the highest performing storage nodes in real-time.
Access Software Client
Feature: SDK that enables reading and writing directly to storage nodes versus having to go through an additional server.
Benefits: Avoids choke points of a gateway, enabling massive parallel reads and writes. Can be embedded directly into devices.
Discussion: Look for solutions with storage nodes so that a software client can speak directly to the storage. Solutions with servers and disk enclosures will require access through the server preventing massive parallel reads and writes.
We figure six features is enough to get started with, and we look forward to sharing many more new features to deliver resiliency with our upcoming software announcement.
Thanks to all the emerging solutions that validate our approach.

