NetApp has announced new version of their storage operating system, Ontap 9. One of the improved features is Advanced Drive Partition V2 (ADPv2).
With ADP you can slice SDDs or HDDs into two partitions, one for root aggregates and one for data aggregates. This means that you don’t have to reserve separate drives for root aggregates, instead you can use a small “slice” of disk for root aggregate and use remaining “slice” for data aggregates. By using ADP the overhead related to root aggregates is smaller and you will get more usable space for data.
ADP is supported in three use cases
- Entry level FAS systems, FAS2500 series. These usually ship with few disks and with traditional separate disk scheme for root aggregates, the usable vs marketing capacity was low
- Flashpool SSD drives can be put into a storage pool with ADP. Parity and spare drives can be shared and usable cache capacity is higher
- All Flash FAS (AFF). Similar usage as with Entry level FAS systems
In this blog entry I will concentrate on AFF and ADPv2. With ADPv1 there was a slight hick-up with entry level AFF configurations. Two root aggregates are needed in a HA-configuration and the required usable size per root aggregate is around 390 GiB. The smallest AFF ships with 12 x 400 GB SSD drives. The smaller the disk size and the smaller the number of disks, the higher percentage of disk space goes to root aggregate slices. Especially with 400GB drives there wasn’t much space left for data aggregates after root aggregates had been built.
With 12 SSD disks you had to make a decision between maximizing capacity and maximixing performance.
Either maximize capacity by using only one data aggregate (and less parity + spare drive slices). The downside with that decision, is that you are leaving about 50% of performance potential unused, as only one storage controller is serving data, while the other controller has only root aggregate, is waiting for failover event to takeover the one and only data aggregate. Out of marketing capacity of 4,8TB (12x400GB), you will get roughly 1,99TB of usable space in data aggregate or 41,44% usable space of marketing capacity.
Example of Asymmetric configuration with 12 x 400GB SSD drives
The other option is to maximize performance and make two data aggregates on top of two root aggregates. With two data aggregates both storage controllers are actively serving data and you will have more CPU and memory to produce performance. This decision comes with penalty in usable capacity as you will have to use more disk slices for parity and spare.
Example of Symmetric configuration with 12x400GB SSD drives
With this Symmetric configuration you will only get 1,33 TB of usable space for data, compared to 1,99 TB with Asymmetric configuration. Or measily 27,62% usable space of marketing raw capacity.
To get most out of their valuable investment, most of the customers using AFF with only 12 SSD drives, chose the Asymmetric version to maximize their usable capacity.
With Ontap 9 and ADPv2 the situation with small AFF configurations is much better. Instead of slicing the disk into two partitions, ADPv2 can now slice the disk now into three partitions, one for root aggregates and the remaining space is split into two equal sized partitions for data aggregates. This means that you won’t have to choose between maximizing your capacity or maximizing performance. With ADPv2 you will get the same usable capacity as with ADPv1 asymmetric configuration ( only one controller serving data), while having two data aggregates and both controllers serving data and producing performance.
Example of ADPv2 with 12x400GB SSD
In part 2 I will cover AFF 12 disks setups with larger SSD drives (3.8TB and 15.3TB)