The Hybrid RAID – Are You Maximizing Your Costs to Performance Ratio?
Traditionally, when looking to setup a RAID array, the goal is to mix performance and size while maintaining low enough overhead to remain profitable. With this in mind, we need to figure out how to meet these different goals.
A traditional SATA array will meet the requirements of size and low cost but will lack the performance needed by modern high I/O applications. By using SAS drives, you can make a move towards higher performance but the cost per GB is increased. The last solution in a traditional RAID setup will be using all SSDs. SSDs will provide both the highest performance and the highest cost per GB.
This is where the Hybrid RAID solution comes in.
Hybrid RAID arrays will allow you to leverage the high storage benefits of SATA drives along with the performance of SSDs. By implementing logic in the RAID controller, your server can utilize the SSDs for highly accessed data while still storing large amounts of static data on the SATA disks. In this configuration, the SSDs are utilized as a form of cache memory. Although it’s not as fast as the onboard RAID controller’s cache, it has the ability to store much more data. Testing of this technology has yielded gains of up to 1,200% increased I/O over traditional SATA arrays when used in high I/O situations.
What kind of applications do you regard as being high I/O?
The servers that will see the most benefit from this solution are those running SQL, Virtual Machines, and web servers. SQL servers tend to see the greatest benefit from the Hybrid RAID configuration. Often in these situations, the biggest bottleneck facing the server is accessing data as quickly as the clients request it. By allowing the RAID controller to learn what data is most requested, it can move that information to the SSDs for quick access. As I/O lowers on the server, the data is then synced back to the traditional SATA drives. This is similar to using Write Cache on the RAID controller but at a much larger magnitude. Most modern RAID controllers offer between 512MB – 2GB of onboard memory that can be used for caching data. By using an SSD, you can easily cache up to a maximum of 512GB worth of data. That is a theoretical gain of up to 100x the cache space.
If the Hybrid RAID solution is so great, why doesn’t everyone use it?
While it does provide advantages over traditional RAID configurations, there are also drawbacks. The need to add 1-2 SSDs into your RAID configuration will often mean the need for a larger server chassis to accommodate the extra hardware. This technology is also only available when an upgraded license has been purchased from the RAID controller manufacturer. These two factors will add cost to the server.
Is a Hybrid RAID right for me?
You need to look into the performance of your server to know the answer to that question. Every server will get to a point where one asset is becoming the bottleneck to performance. If you are not fully utilizing your CPU or RAM, then you may need to increase the rate at which you can access your data. Tools like “top” and “iostat” can help you figure this out in Linux. On Windows, you can look at your system performance under the Resource Manager. The key is that, if your disk access rate is high and the rest of the hardware utilization remains low, there is a good chance you would benefit from better performing disk speeds. In this case, the Hybrid RAID may be your solution.
Often times you may find that you can get better overall performance with the use of less servers once you remove this bottleneck, thus allowing you to offset the increased cost of the server hardware.
Updated: January 2019