HP EDW Compute Storage Node Options
When ordering any SQL Server Parallel Data Warehouse appliance, you only have a few options. One of those options for PDW is the disk configuration for your compute storage nodes. I’ve described them in detail in the past, please but one of the constants about appliances is that over time, specs will change. New hardware will demand it. Additionally, we will come up with new best practices and will implement them in future versions. Currently, the HP EDW has three choices for each Compute Nodes storage:
1. Twenty-four (24) HP 300GB 6G SAS 10K SFF (2.5 inch) DP ENT HDD
Known as the High Performance option, these Small Form Factor disks allows for many more spindles in a HP P2000. The advantage here is that the each distributed data LUN essentially has its own RAID 1 array to itself. It will share these spindles with the Replicated Tables LUN which spans all the data disks. Additionally, the transaction logs have triple the spindles available to them when compared to the High Capacity option. You can also see that there are two hot spares for each P2000 in this configuration.
2: Eleven (11) HP P2000 1TB 6G SAS 7.2K LFF (3.5 inch) DP MDL HDD
This is known as the High Capacity option. For more information on this configuration, please take a look at my previous posts on Storage Node Configuration and its Part II.
3: Eleven (11) HP P2000 300GB 6G SAS 15K LFF (3.5 inch) DP ENT HDD
For those who want the high performance of 15K drives, but don’t need the capacity of of the 1 TB disks, there is the option of having Large Form Factor disks in the size of 300 GB. I honestly see this as the least likely of the options being chosen by our customers. If you are going to pay for an appliance, you want to get at least a lot of capacity or a lot of performance. This configuration looks to do neither especially well. Option 1 is going to give you much better performance with each distribution having its own RAID 1 array for their LUN and Option 2 is going to give you much more than triple the storage space.
SQL Server PDW uses SQL Server page compression by default, so the above choices will give you the following rack capacity based on 2.5x compression of your data. This is a relatively conservative compression estimate, so you may get better compression rates and thus be able to store more data than described.