4 SSD Strategies For Big Storage Vendors

Intel Puts Future On Exhibit

I moderated a panel on solid state disk (SSD) at the ExecEvent in Austin last week. The panel was populated mostly by emerging vendors, so of course we discussed the subject of big storage vendors and what their SSD strategies are.

At this point it is hard to understand most of the big vendors' plans, but I shared what I thought they should be doing. Big storage vendors need to begin to move beyond looking at SSD as just a faster hard drive and start developing an end-to-end strategy that leverages SSD to enable customers to meet the demands of the modern data center. Here are the four steps big storage vendors need to take:

Step one: Fix the storage system. SSD is not an upgrade to the 15K HDD. Simply putting the technology in the same system as you currently put hard drives in will push the performance envelope only so far. There will be improvement, but you won't be getting maximum value from your premium SSD investment.

The latency inherent in a storage system that goes unnoticed in an HDD-based system is fully exposed by SSD. As I discussed in What is a Storage Controller?, storage systems are a combination of software driven by compute power to deliver data and data services to the storage administrator. Big storage vendors need to rethink the way they connect their storage controllers to the storage devices and how they connect those controllers to the applications they provide storage to.

[ Adding SSDs to your network? Read How To Maximize Your SSD Investment. ]

This latency between the storage controller and the storage device is one of the reasons we continue to see standalone SSD appliances do so well in the market. They are designed specifically for SSD and these vendors work hard to remove latency between the controller and device. Big-storage vendors will either need to break from the shackles of their legacy designs while leveraging their software investment, or they will have to buy some of these emerging SSD-only storage systems.

Step two: Fix the storage network. One of the hidden costs of implementing an SSD solution is upgrading the storage network to support the performance of memory. There are a number of ways to fix this. One is to move the SSD closer to the server, an approach being popularized by PCIe SSD vendors. We also have seen two instances of edge types of devices that bring the network connection and the SSD storage directly to a rack of servers. This lowers the cost of the infrastructure to access the SSD device but retains shareability.

Step three: Build the bridge. SSD is going to be deployed by data centers in a variety of different ways. They might go all in with SSD as part of their next storage refresh, they might look as some sort of network cache to provide performance relief to a broad range of workloads, or they might leverage PCIe SSD in the server to fix a few specific point storage performance problems. Although emerging vendors can't typically afford to participate in each one of these markets, big storage vendors can and should.

Step four: Enable SSD everywhere. Eventually, the data center is going to consist of a variety of SSD solutions to address performance bottlenecks throughout its infrastructure. It is reasonable to expect that many data centers will deploy a combination of server, network, and storage-based flash solutions. The key component for the larger storage vendors is to enable the use of SSD no matter where it is. The vendors will need to provide a way to move data back and forth between these SSD locations--not just tiers--as it makes sense. Surprisingly, only a few of the vendors have the software projects underway to enable this type of data movement.

Big storage vendors don't need to buy their way into the first three steps outlined here if they don't want to. The key is to have partnerships in place. Number four, though, is critical. Vendors need to have software intelligence that can move data between these different SSD storage types and locations so that any performance problem can be addressed as close to its origination as possible.

Big data places heavy demands on storage infrastructure. In the new, all-digital Big Storage issue of InformationWeek Government, find out how federal agencies must adapt their architectures and policies to optimize it all. Also, we explain why tape storage continues to survive and thrive.