What you will learn in this tip: Discover the pros and cons of using RAID with SSDs, the basics of SSD RAID and...
the relevance of using RAID to improve data availability and increase the performance of HDD systems.
Twenty years ago, a business without a fax was unthinkable. Today, life without email defies the imagination and nobody faxes anymore. So, is it so crazy to wonder about storage – especially solid-state storage – without RAID?
RAID seems to be a storage fixture but it was only formally defined in 1987. The major objectives of RAID were always to address the lack of hard-disk-drive (HDD) reliability by improving data availability and to drive up the performance of HDD systems.
While RAID is still a de facto storage standard, there is still a question of whether it is an optimum approach – even for HDD systems – because almost all levels of RAID require an overhead to provide the protection.
For solid-state storage, there are even more questions about RAID’s relevance. This tip will explore the basics of using SSD and RAID together and offer key advice for RAID, SSD and your storage environment.
RAID with SSD: The basics
Solid-state storage gives users loads of performance, so RAID’s performance enhancements are moot. That puts the focus on data availability and protection. Many flash chips have basic RAID built-in to increase redundancy and longevity; the question is whether more system RAID on top of that helps.
Whether your solid-state storage is used as a tier or as a cache is a key consideration: Many vendors’ implementations require confirmation from a lower tier of spinning disks before confirming the write. And most cache – aside from read-only – is unlikely to offer immediate data protection.
This affects performance and adds to the cost – so check with your prospective vendors about how they address this.
Learn more about RAID with SSD
Discover more about RAID with SSDs in our tutorial on solid state disk
Learn about the pros and cons of RAID 6 vs. RAID 10
How RAID works with SSD
If SSDs merely replace some HDDs in a system, then the same RAID can be applied. You’ll invariably need the same RAID to be applied within and across RAID groups to allow tiering and movement flexibility. For most high-end systems, RAID 5 or RAID 10 are likely sufficient, but for added security, RAID 6 (double parity) is probably preferable. Of course, the raw reliability of solid-state is also relevant: Many solid-state vendors are now claiming “x” years for a given number of full daily writes that are often considerably better than equivalent figures for HDDs.
Why you might use RAID with SSD
There are many reasons why you might use RAID with SSDs. A PCIe solid-state storage card in a server is a popular approach to boost application and storage performance; however, it’s effectively a DAS model, which translates to a single point of failure. To help protect against losing data, a simple RAID 1 (using a mirrored flash card) model might be appropriate, albeit expensive. Otherwise, such cards are often implemented as read-only cache so the protection is performed more economically at the HDD level. This all comes back to knowing what you want to achieve regarding the balance of performance and data availability/protection.
Specialized solid-state RAID hardware/software
New and updated RAID controllers are emerging that allow storage systems to use more of the massive performance boost that solid-state storage can provide, as a few SSDs can quickly overrun the capabilities of traditional controllers. The main focus of these new controllers is simply to allow more SSD performance to flow to the server and applications rather than a revamping of RAID. However, it’s important to note that, with the new breed of purpose-built all-flash and hybrid flash/HDD arrays, some vendors are implementing a special version of RAID that’s optimized for their particular SSD implementations.
Economics and mixing HDDs and SSDs in RAID arrays
Having established that RAID is – by definition – all about redundancy, the simple truth is that some amount of copied data and/or parity data has to be stored somewhere. The extent and placement of that “extra” data depends ultimately on economics. If money was no object, you’d have multiple copies on multiple solid-state devices. Realistically, that’s not economically feasible, so a hybrid approach makes more sense for using RAID with SSD. Prospective users of solid-state storage should give careful considerations to the options offered by various vendors. A hybrid or mixed approach can make sense in the array (although it requires careful pool management) but it doesn’t make sense in a RAID group because the performance and reliability characteristics are so different that you’d lose the solid-state advantage in the process. Some of the emerging vendors have sophisticated firmware to fully use the solid state while placing the redundancy data on lower cost media, which can save in the range of 10% to 50% on overall costs (compared to traditional RAID on flash alone) for well-protected, highly available data.
Mark Peters is a senior analyst at the Enterprise Strategy Group, focused on storage systems. Mark’s particular emphases are block storage, virtualized storage and all types of solid state storage. He has more than 20 years of experience in IT. He can be reached at firstname.lastname@example.org, on Twitter, or on his The Business of Storage blog.