In the storage world, people have taken notice of all-flash arrays and what they are capable of. The switch in revenue between hybrid and all-flash arrays is a clear indicator of this paradigm shift. In fact, in a recent prediction Gartner stated that by the end of 2020, almost 50 percent of data centers will use all-flash arrays for major workloads, almost 40% more than today.
While it is good to be a part of this change, as a data center manager you cannot neglect the underlining question – where does all-flash fit in your data center environment and are you ready for this change?
Falling SSD prices and advances in data reduction technologies offered today are making flash more affordable with every passing day. In fact it could be a strong motivator to shift to an all-flash array: after all, who could say no to affordable performance from a storage array? But before you make a move, it is worth evaluating your current environment. Ask yourself:
- What are the data sets I am currently dealing with?
- What are the specific read / write ratios?
- What is the overall amount of duplicate data (if any)?
- Where would deduplication and / or compression will make sense for me?
- What are the applications where performance must not be compromised?
- Where can a certain level of latency be tolerated – and to what level?
- What are the data migration policies I should keep in mind?
- If I am to consider a Disaster Recovery scenario, what are the protocols around that?
- What will be the cost of that growth – in terms of new features, software, additional capacity, licensing, etc.?
With these answers, you can begin to build up an idea of where you are, where you want to be and what technology you need. It is then that you can evaluate what vendors have to offer. Every all-flash vendor has something unique to offer and every all-flash array is different, in its own right. This could be in terms of: room to grow; data reduction features offered; distribution of resources within the array; arrays ability to automatically deal with the I/O demand in case of low and high priority workloads; architectural advantages of the array and others.
No matter how different the features and capabilities offered by the all-flash arrays in the market today are, they have one driving force in common: performance. But performance numbers can be very difficult to compare. This is because performance numbers are impacted by the configuration used – the workloads, the data block sizes and so forth. This is not standardized across vendors and many not be relevant for your specific setup.
Price parameters are also difficult to establish for all-flash arrays. It all depends upon how the respective configuration is quoted. For example:
- How many SSD’s are quoted – to understand the shelf space they occupy
- The capacity of those SSD’s – to understand how many SSD’s will give the raw capacity needed
- Whether or not deduplication and/or compression is factored in, while arriving at the effective capacity.
- If the quote from the respective vendor includes support cost and the level of that support
All these aspects are sometimes very difficult to segment and follow through.
So, now the question is: How can you compare different all-flash arrays?
This is where SPC Benchmarks can be of great help. SPC is a vendor neutral, industry standard body that sets the criteria for storage performance improvement. These criteria are based on real life scenarios wherein the performance of storage array on both OLTP and sequential applications is measured. SPC Benchmarks are divided into SPC 1 and SPC 2:
- SPC-1 consists of a single workload, designed to demonstrate the performance of a storage subsystem while performing the typical functions of business critical applications.
- SPC-2 evaluates the performance of storage subsystems during the execution of business critical applications with intense I/O operations, initiated by three workloads: Large file processing, large database queries and video on demand.
SPC benchmarks are complex and take months to complete. However, they can offer an overall framework and could be your starting point when considering different vendor offerings.
That’s why I’m particularly pleased that the new ETERNUS AF all-flash arrays – available in APAC since October 2017 and EMEIA since January 2018 – performed well in the SPC benchmarks.
ETERNUS AF650 S2 can scale up to 3PB (raw) and delivers enterprise class functionality (automated quality of service, data reduction technologies, cross family replication amidst others), all at a price / performance of $0.10 SPC-1 IOPs.
With these benchmarks you can compare the latest generations of storage arrays, both all-flash and hybrid. If you are considering all-flash arrays, you can find out more about ETERNUS AF650 S2 and ETERNUS AF250 S2 all-flash arrays on our website. If not, we hope this write-up gave you some ‘food for thought’ and got you thinking about your options.