The usefulness and economic value of a storage system is governed by the initial assumptions made very early in the process by engineering teams. In my career, I have worked at a number of storage companies in which I was involved in the early design of products that were still being scoped.
Engineering, people, and ability to see into the future of hardware
The brilliant minds that give birth to a new storage system must possess outstanding engineering and people skills. However, it’s equally important to have a sense for the direction of technology . It is very difficult, if not impossible, to change fundamental assumptions after the fact if the initial assumptions were wrong about the direction of hardware technology.
Good assumptions lead to a system that can exploit just enough of the latest technology to increase performance and capacity, but keep costs flat to the end user. Poor assumptions will lock a customer into outdated hardware that comes at a premium cost.
Let’s examine one of the key assumptions that flash array vendors must make: the need to identify which types of disk drives are used now and in the future.
Most of the all-flash vendors on the market today have made the assumption that every drive in their controller, shelves, bricks, blocks, etc., will have identical attributes. Those attributes would include vendor, capacity, interface type, performance and cost. Next, flash vendors pooled their drives together and layered virtualization on top of the drives to result in a single pool of storage where their operating system could work its magic.
Creating a virtual address space from drives that vary in performance and capacity is too difficult; so the assumption was made that “they will all be the same.” One needs a uniform space in order to do system-level or OS-level garbage collection. The uniform space allows one to use reserve space from any drive on the system, and thus take the small reserve space on a consumer-grade device and aggregate them all together. One can then decide where to write data within the larger pool because all of the drives underneath have the same attributes.
The above approach is viable as long as the pros and cons are understood:
- The good news is that you have a system that can consume today’s drive technology efficiently enough to gain customers.
- The bad news is that you can’t use the latest technology on the market because that would require your system to use the latest drives in every slot in the system, drastically driving your price point up or slashing into your margins to the point that you can’t sustain the business.
Thus is the nature of the latest drives that have been released to the market: they offer great value; but the price is typically much higher than drives that are 1, 2 or 3 years old. Have you wondered why some vendors on the market today are still shipping all 256GB or 512GB SSDs in their system? You may have some insight into the “why”; but do you really want to pay the prices they ask for such old technology?
Over time, what does this look like graphically?
A more forward-looking assumption would be that your system may always contain different types of drives in the system. New drives of HDD and SSD are coming into the market so quickly that let’s assume we will always have a mix of drive types in the system. This mix of drive types allows the OS to quickly consume “some number” of drives into the system, while keeping the majority of the drives in the current generation.
In the above approach, one could let the enterprise-quality SSDs do the garbage collection:
- The good news is that you would have a system that can consume the latest technology in the right quantity to have compelling capability and economics. You would not have to massively increase prices for your customers. Customers would get a system that leverages the newest technology to keep their performance and density increasing, but their prices flat.
- The bad news is that your competitors would sling FUD (fear, uncertainty and doubt) at your product, stating you are simply throwing the latest hardware at the system.
What is our response at Tegile? We say, “yes we are using the latest hardware.” This prompts us to ask the orange guys, “Shouldn’t you be using the latest technology in your system, too?”
The jury is still out as to which are the right assumptions to make when creating a storage system. However, we have some strong votes in our favor. Tegile storage systems have earned this recognition recently:
When it comes to creating storage that serves a market today and into the future, you’re only as good as your assumptions. At Tegile, we think we started in the right place and our systems are built to serve today’s data demands – and the data demands of the future.