For a long time now, Total Cost of Ownership (TCO) has been widely held as the most important aspect of evaluating an IT purchase. Many academics and industry consultants have endorsed its consideration of the ongoing financial commitments involved in owning and operating equipment as the most beneficial in the long term.
TCO evaluates everything from power supply and cooling options to service contracts for both hardware and software, floor space value and even the help desk for the application. TCO also considers how much a product will help IT and end users become more productive.
In other words, TCO considers pretty much everything that needs to be considered beyond the initial cost. And I have learnt over the years that it is indeed very important to compare these factors before choosing which vendor or solution to go with.
The reality in most organisations is that it is rare for most of the people that operate infrastructure to have any responsibility or control over operations or the purchasing process. The facilities manager, not the IT manager, pays the electricity bills. Budgeting is often looked after by a different business unit as well. On top of this, most people are reluctant to put their careers and reputation on the line to back TCO figures in front of a purchasing committee.
However you look at it, organisations have to deal with upfront purchasing costs. TCO is nice but in most cases, most of the battle is fought at the point of deciding how much organisations will need to pay to acquire the equipment they need. The long term costs often fall under the category of things that can be discussed later. This process also involves many different players and the evidence to justify these hard costs is often lacking or difficult to outline to procurement managers whose main priority is not speed or performance.
When it comes to SSDs and other flash-based solutions, I think a lot of vendors often do themselves a disservice by falling into the trap of comparing just the storage device costs. This gives an inaccurate impression that flash-based storage is ultimately more expensive than HDD-based alternatives. We have to remember that organisations don’t just buy drives. They also buy all the surrounding gear that is required for a fully operational infrastructure. These often include enclosures, circuit boards, backplanes and sometimes even power supplies. To enable drives to communicate effectively in and out, they require intelligent connections between arrays and servers. This means organisations also need to invest in RAID controllers, host adapters and even cables. All of these purchases are part of the acquisition stage and can be compared in a quantifiable manner.
In the drive for wider adoption of flash, we need to get more involved in the discussion around Total Cost of Acquisition (TCA) to make the in-roads we need. The TCO discussions can wait.
The speed and performance benefits of flash-based infrastructure are hard to argue against. SSDs have lower failure rates, are more than 100x faster and take up less physical space when compared with HDDs. When it comes to workload, fewer SSDs are also required to cope with a comparative workload than HDDs. Even HDD strongholds such as streaming of sequential is now conspicuously slower than SSDs.
If we consider the requirements for running a 50TB database with traditional spinning rust and with flash, as detailed in the image above, the first thing you will notice is that less hardware is needed. It takes 288 300GB hard drives to do the same job as 24 4TB solid state drives. That is a 91 per cent decrease in devices when flash-based solutions are deployed.
This also means a 66 per cent reduction in enclosures. If you multiply that over a hyperscale data center, it turns into substantial savings in real estate and maintenance. You also reduce energy consumption by 86 per cent and this figure includes both power for the systems and power for cooling.
On top of all of these advantages, flash also gives you 1 million IOPS versus 51,000 for hard drives. That is a 20x performance boost with a tenth of the hardware. This adds up to an overall TCO savings of 27 per cent and TCA is also reduced by 23 per cent compared to HDD-based solutions.
However you look at it, the acquisition cost of HDDs is fundamentally higher than SSDs. When the numbers are crunched and the scale of investment needed to get HDD based infrastructure off the ground is considered, flash-based solutions win the TCA argument, as well as the TCO argument. IT managers and flash advocates need to remember this as they try to convince the procurement manager that flash-based solutions are a wise choice.
Steve Wharton, Office of the CTO for Enterprise Solutions EMEA, SanDisk