Most data centers have at least five to six different storage systems from different vendors, increasing storage costs and management complexity. One potential solution is the all-flash data center, where all data is moved to a single all-flash array. While technically possible, the all-flash data center, at least for right now, is not practical. The overwhelming amount of data (as much as 85%) has not been accessed in the last year and since there are less expensive storage options than flash, that data shouldn’t be stored on flash. Instead of five or six types of storage systems or a single storage system, IT planners should look to create a two tier data center, where all-flash is for the active data set and object storage is for the in-active data set.
Data Movement vs. Data Placement
Most data management solutions focus on constant moving data to and from various tiers of storage. The problem is each time data is moved there is a chance for a failure. Also, to make this movement seamless, data movement solutions will use stub-files or metadata management controllers to make sure data can be transparently recalled when accessed.
An alternative is a data placement strategy, where data is moved but less frequently. Data placement is a simpler architecture that requires more from the storage systems but is ideal for a two-tier all-flash plus object strategy. With a data placement strategy, IT manually moves data only when it has to or at scheduled periodic intervals. Users are trained that if the data isn’t in location A (the all-flash array), then it is in location B (the object storage system).
Requirements for Tier 1 – Optimized for Performance
Tier 1 in the two-tier data center is more than likely an all-flash array. The all-flash array may have sub-tiers within it that it manages. For example, the flash system might have a low latency NVMe flash tier combined with a high capacity flash tier. Or it could be a hybrid array and have a flash tier coupled with hard disk drives. In either case, this flash system needs to have or scale to enough capacity to cost effectively store the year’s most active data. With that kind of scaling, IT only occasionally needs to move data from the all-flash tier to the archive tier.
Requirements for Tier 2 – Optimized for Endurance and Economics
The second tier needs to have two primary capabilities. First, it needs to ensure data authenticity and protection because it may store the last known good copy of data. Second, the tier needs to be very cost effective, ideally using high capacity hard disk drives and commodity storage hardware. The second tier may also serve as a backup to tier 1, replacing the need for another tier of storage – protection storage. Many backup solutions can send data directly to an object storage system today, and for those that can’t, many object storage systems can provide NFS or SMB access to their stores. Object Storage is an ideal storage system for the tier 2 use case.
How to Move Data Between Tiers?
The final question is how to move data between tiers. An increasing number of organizations are finding an occasional manual movement of data between tiers is both cost effective and simpler. While the manual method does mean IT may need to get involved in the occasional restore the frequency of that occurring is relatively small. Remember, the data on this tier wasn’t accessed in over a year. The chances of it being accessed again are very small.
Want to Learn More?
In our on demand webinar, “How to Create A Two Tier Enterprise With All-Flash and Object Storage”, we discuss the two-tier strategy and how to implement it in your data center.
Register for the webinar and learn:
- A Practical Strategy for Data Placement
- What the Two Tier Data Center Needs From an All-Flash Array
- What the Two Tier Data Center Needs From an Object Storage System
- How to Move Data Between the Two Tiers