Talk tiered storage and you would be surprised if eyes didn’t roll and cloud over; to be expected really as it’s not one of the most interesting topics to discuss! Until now, that is… Take Storage magazine’s latest survey that revealed 60% of their respondents are not only using tiered storage systems, but they are getting excited about it! They’re talking flash-based tier-zero storage and automated processes to tier storage their data, keeping data that is on the costly tier to an absolute minimum. Now a tiered storage data model is the topic of conversation…
So, what is this new tiered storage? Flash storage, which are also known as SSDs or solid state, means that companies can develop tiered storage solutions that perform at optimum speeds when accessing the data. There are a range of benefits including higher throughput, reduced latency and access time, lower energy costs, and being able to align application workloads to exact capacity needs. Let’s take a look at the top 5 new developments in tiered storage.
1. Flash tier or tier zero? Well, they’re one and the same… ‘Hot data’ (your data) is temporarily stored on what are called flash SSDs (solid state drives) and written to a flash tier, also known as a tier zero for the purpose of rapid access. But avoid the temptation to build too many tier zeros or tier ones; not all your data is high response. For guidelines on creating storage tiers best practices, read Jon Toigo’s article.
2. Want automated tiering without and SSD? Depends… vendors tie the two together (for obvious reasons!) advising that SSDs in storage arrays delivers automated tiering. However, Amerijet International found that they didn’t get a faster performance on all their applications to which data is written; whilst Compellent got faster spinning drives with their automated tiering software and found that they didn’t necessarily need flash.
3. Different vendors, different answers. According to which vendor you ask, your tiered storage model can be automated in different ways within the storage controller – flash pools, flash cache, DRAM cache, or flash first! Some recommend starting at the bottom, whilst others go with starting at the top with ‘flash first’. Find out who recommends what and why, including improvements in performance and hit rates, before making a decision.
4. SSDs as cache. Ok, so you’re getting excited about automated storage; however, don’t forget that solid-state drives can also be used as a cache, with the same result at the end. Make sure the data you are going to access the most is stored on the fastest tier. One point to note though: caching creates a second copy of the data, whilst automated tiering removes data from the hard drive; if a cache fails, it’s a performance issue – you haven’t lost the data.
5. Don’t want tiers? Then there’s all-flash. There are some industry thought-leaders that believe tiered storage is going out of fashion, and they could be right despite the introduction of SSDs. If this is the case for you, then consider an all-flash system, such as IBM’s all-flash arrays, Flash System, launched in April. An all-flash system has just one tier, eliminating the need for any tiering software. Hewlett-Packard released their 3PAR StoreServe all-flash system in June this year; EMC and NetApp are also planning releases.
Tiered storage will probably be around for some time yet; not every company has the resources or capabilities to move to all-flash systems. But the development of SSDs and automated tiering has certainly brought life back to the practiced tiered storage model!