In this FAQ guide, Mark Peters, an analyst with consulting firm Enterprise Strategy Group, provides answers to the questions he is hearing most frequently from storage administrators related to solid state drives.
At Enterprise Strategy Group, Mark covers storage systems and storage in the data centers, and also writes a blog called Mark My Words, which he calls "an askance view of what's amazing, asinine or ahead in storage." In today's podcast, Mark is going to let share some of the questions he's hearing most often about solid state technology.
Below, you can read his answers to these frequently asked questions or download a podcast of the Q&A.
Table of contents:
What's amazing about solid state?
How will SSDs impact enterprise storage in 2009?
What applications will drive SSD adoption at the enterprise level?
Will SSDs be integrated into more storage arrays or servers?
Should flash be in the form of cache or as persistent storage?
What's the biggest myth about solid state?
What's amazing is, like many things in this industry, solid state technology is not new. People talk about solid state disk and solid state storage as if it's only turned up in the past year or two. But the first solid state device for the enterprise was produced in 1978, so it's been around in one form factor or another for a long time.
As to the asinine side, just to be controversial, I think it's the fact that you still hear some people saying it's not going to be for real and it's not going to happen.
Well, it's going to begin to have more of an impact. There's a very generic, glib phrase. But there are a number of things happening that make the statement true. There's a certain self-fulfilling element to all this in terms of the number of vendors and the amount of effort going into making this technology real and easily consumable. The rate of progress of hard disk drives has slowed to some degree, so that makes solid state more interesting.
Given that we're in such challenging economic times, that makes solid state more interesting. Obviously with my focus on the data center I look at the green aspect of computing as well, and it's hard to overlook solid state from that perspective. But the biggest single thing is the maturation of the technology, the greater availability of the technology and the better explanation of how it gets used.
You still hear people talking about the cost per gigabyte of solid state, but that's not really the point. The point is what solid state can do for you from an I/O perspective. So it should be about the cost per I/O. That's going to be the tipping point to speed the adoption.
I don't see solid state ever becoming the only form of storage we have in any reasonable timeframe, but it will certainly become significant over the next two to three years in terms of handling I/O. For general capacity that doesn't get hit very often, in terms of peoples' overall beta infrastructure, that will still be spinning disk, because the economics haven't changed that dramatically. But in handling I/O and handling performance, solid state makes great sense.
That depends on the particular industry and what they're trying to achieve. Generically, whatever is most important to a business or enterprise or organization in terms of getting throughput and I/O handled, wherever you need speed, wherever you need a great deal of performance in terms of throughput, then solid state will be great. Specific examples would include database acceleration, Web services and finance operations. But it can be applied to whatever is the most compelling area in terms of throughput and performance in any business.
Yes. I'll expand on that. It's interesting that we have a storage hierarchy in the first place. Frankly the only reason we have one is economics. There is no real reason to have anything in computing separated. The overall stack, from the application down to the computer platform, down through the network to the storage, all those things at the end of the day we get all twisted about which of those is most important, and where certain things are going to fit, but at the end of day, none of those really matter. Getting work done is what matters.
Sp where will solid state go in the infrastructure? Where it makes the most economic sense. Coming back to apps that people deal with, if it's something where it's long-term persistent storage, then obviously, with today's infrastructure stack, will go to solid state disk.
If you're looking to boost the overall performance of your infrastructure, it makes more sense to have a caching device or acceleration module further up the stack. Looking into the future with a massively optimistic crystal ball, we come up with something the size of a thumbnail that stores everything you could ever need stored and sits within the computing platform. We only put things where we have to because of economic impetus, and that economic impetus is determined by what we're trying to achieve.
Long term storage goes on a solid state disk? Certainly for overall performance improvement, that will go in cache, that could even go at the server level or in some specific cache module that sits on the network in front of the storage.
No, it really depends what you're trying to achieve. If you want everything persistently stored, then you're going to have to use a solid state disk. The advantage of cache is that you get a performance improvement over your whole storage infrastructure and this makes it more attractive to have cache in some areas, you don't need any special software or skills to make that happen.
You add the cache in, and in this case it's solid state cache and by definition your hot I/O will go in there. It's not applied to everything, if you have a huge budget then you can put everything on an SSD. But no one's going to do that in the short term. It doesn't really matter. It comes down to what people are trying to achieve and how much money they have. I keep coming back to that, but that's why we have a storage hierarchy. Money is the root of everything here. If money were no object, everything would be in solid state.
I'm going to give you three. The first is that solid state is not new. The second is that people talk about it being too expensive and I still hear this even from people in the industry and it appalls me. I keep hearing people say once we get the price down to that of a Fibre Channel 15K disk then we have a business for solid state.
I think that's complete hogwash. That's looking at it from the wrong perspective. Solid state for next few years is all about a relatively small amount of capacity that can save a tremendous amount of I/O. You can get 80% of your I/O taken care of with 20% of your capacity. Even in terms of today's pricing, cost per I/O or the I/O per watt for solid state are already very compelling.
The third thing is that people talk about the ruggedness. You hear a lot about write endurance and program disturbance and wear leveling and error correction codes, and so on and so forth. Yes I know there are issues in the short-term but they'll be overcome. There are already people finding fixes and ways to address them. Let's face it, that's what we've done with spinning disk over the years – we've found ways around the problem, whatever the problems may be.
That's why we have RAID, that's why we short-stroke and abuse the capacity of many of the spinning disks that we've got, to get the performance that we need. We find ways, some of them stupid, some of them forced upon us, some of them imaginative ways around things. So for the ruggedness, the security and the reliability of solid state, there are already ways around it and those will only be improved if you look at the technologies coming over the next few years.
This was first published in December 2008