In our last blog, Storage Switzerland discussed how a hybrid array, with a properly sized flash tier, could provide similar performance to an all-flash array while substantially reducing costs. The next step in storage architecture design is to use the “deep math” discussed in blog 1 to offload capacity from the hybrid array.
It’s Not About Costs
A hybrid array filled with high capacity hard disk drives, in addition to flash drives, is very cost competitive. While an order of magnitude less expensive, an object storage system’s primary role is to help an organization meet long term data retention requirements.
Data Management is King
Data of all kinds is growing at prolific rates and it is necessary to manage all of this data. Organizations want to retain data so they can analyze it years in the future. Governments and regulating bodies want organizations to retain data to be able to meet the demands of future discovery requests and privacy requests. Object storage systems are better suited for these tasks.
Any discussion of data retention must begin with data durability. While a primary storage system’s durability story typically begins and ends with RAID, object storage systems leverage replication and erasure coding to meet durability requirements. The first advantage is these systems have a more granular understanding of the files they store and many can set data protection policies on specific files or types of files.
Erasure coding, common in object storage systems, enables the object storage system to survive multiple storage media or storage node failures without losing data or access to data. Some object storage systems also provide an out-of-band maintenance system, which performs tasks like data integrity verification, and self-healing of stored data.
Some object storage systems can even guard against data loss from silent data corruption, by performing CRC checksums. The system performs background scans of the storage pool for checksum mismatches. If it finds a mismatch, the system will delete the corrupted block and generate a new correct check block in a location on another disk.
Another key aspect of data retention is scalability. The object storage system is the offload point for old data from the hybrid array and it will likely retain this information for a long time, potentially forever. It may also store backup copies of production data, eliminating the need for a dedicated backup storage system.
The object storage system needs to scale. Most object storage systems are a cluster of commodity storage servers each with storage capacity. As more data accumulates. IT just adds another node, which provides additional processing power for cluster management and capacity for the data.
Data management is a math equation and an object storage’s place is a key variable in that equation. The hybrid system is the storage place for active data (most of which is stored on flash). It is also the storage area for dormant data that has a good chance of becoming active again, and therefore retrieval times for that inactive data matter. The object storage system is the storage location for data that *might* become active and where slower retrieval times (by a few minutes) are acceptable.
The combination of object storage with an all-flash array or a hybrid array reduces the need for those systems to scale. The result is a simpler and less expensive primary storage purchase and a more secure, more cost-effective secondary tier.