There is a lot of industry talk – vendor and user – about data protection these days. Most of it focuses on backup, restore and DR. These are good and necessary conversations to have, but wouldn’t it be better to first have a conversation about how to reduce your chance of needing your backup or DR solutions?
Data protection begins on disk, and conventional disk-based data protection on disk is becoming riskier. As drive sizes follow Kryder’s Law, roughly doubling in capacity every year, drive speeds are almost flat. You know what’s coming next… Why is there so little discussion over the real problems this trend poses for RAID-based systems?
Have you heard a good answer to the question ‘how are you going to rebuild 3 TB drives?’ How about 8TB drives? They’re coming. Triple parity will save us, you say? Ok, then what? At what point are you just paying so much performance and capacity overhead that it’s just not worth it? This is not a problem that can be solved simply. We’re going to see all sorts of strange and unnatural attempts to solve traditional hardware RAID scaling, and that’s too bad – because there is a better way.
Isilon’s OneFS® does not rely on hardware-based RAID for data protection. OneFS can use all the resources – IO, compute and memory – in every node in the scale-out storage cluster to efficiently stripe data over all available devices. Because this capability is independent of individual RAID controllers, performance just scales – it actually gets better the more the cluster scales because each node adds not just capacity, but IO, memory, and CPU to drive it. It is also worth mentioning OneFS restripes data automatically to optimize performance and take advantage of any new nodes which are added – without any administrative action needed.
With OneFS you can specify your level of data protection at N+1 through N+4 (i.e. higher than standard hardware RAID offers today), and because OneFS controls data placement, you can specify data protection at levels RAID cannot; you can set the same protection level for the entire environment, or get as granular as different settings on a per-file basis. Because this is all integrated with SmartPools™ – the OneFS tiering solution – you can set rules based on data type, age, owner, etc., which allow you to control how data protection and access changes with the value of your data.
So, while no one is ready to say that traditional hardware-based RAID is dead or even dying, its inherent performance, capacity and management overhead is only going to grow with drive sizes and there’s no solution in sight. Is it time to start looking at alternatives?
What do you think?