Enterprises of all sizes and types struggle to control data storage costs while accommodating increased performance requirements, triple-digit capacity growth rates, multiplying business applications, and skyrocketing energy costs. Imagine addressing all of these concerns with one practical solution, and lowering costs in the process. This is the promise of the “mixed” storage architecture.

Solid State and Hard Disks

Data storage at the enterprise level comes in two flavors, solid state and hard disks, with tape media used only for archiving. To appreciate the benefits offered by mixed storage solutions, a quick review of the options is helpful. Solid state drives (SSD) present a variety of systems that store data on Random Access Memory (RAM) and/or NAND Flash chips. RAM-based SSD has been available for years. It has appealed to high performance computing users, commercial enterprises such as equity trading firms or wireless providers, and DOD entities where the need for very low response times (latency) and very high inputs/outputs per second (IOPS) made it worth its cost.

Flash-based SSD, at least for the enterprise space, is a more recent arrival on the scene. Flash SSD has been used over the years in military and extreme environments because it’s much more rugged than spinning mechanical disks. But only recently have storage vendors made Flash SSD attractive to traditional enterprise users by solving Flash’s inherent problems with endurance and write performance. 

Mechanical spinning hard disk drives (HDD) have dominated the enterprise storage market for decades. HDD storage capacities have increased over 200,000 times since the 1950s, keeping HDD-based storage affordable. Unfortunately, the performance of HDD, measured either by its latency or by its IOPS capacity, is another story. Just in the past 20 years, one indicator of storage performance requirements, the speed of processors, has increased over five million times, while the basic read and write speed of HDD has increased an astonishing…three times!

No amount of HDD engineering has significantly narrowed such an enormous gap, but storage vendors keep trying. Individual hard drives normally accommodate 100-200 IOPS. Put one hundred together and access them in parallel and you have 10,000 IOPS. Put one thousand in Disk Enclosures, spread the data bits between them through various RAID (Random Array of Independent Disks) regimes to further enhance performance and protection, add some CPUs for management, and you have the typical enterprise storage array.

Pyramid of Performance

From the beginning, users understood that some data would be accessed frequently, such as database index tables, while other data – the records of a closed account, for example – might need to be archived or rarely accessed. Thus storage “tiers” or “mixed” storage architectures developed, with database managers increasing application responsiveness and lowering costs by moving frequently accessed data onto faster, more expensive storage and less frequently accessed data onto slower, less expensive media. Information Lifecycle Management emerged from this process, with data beginning its “life” in the enterprise as frequently accessed files, and over time moving to slower storage as it became less useful and accessed less frequently.

The fastest RAM SSD occupies the highest storage tiers, including Tier 0. Enterprise-grade Flash SSD uses RAM as a front-end data “pool” or cache for management and performance purposes, so is called Cached Flash. Mission-critical applications in industries ranging from financial services and telecom to e-commerce and video rendering benefit from Tier 0 storage.

On the border between solid state and hard disks, and also between Tier 0 and Tier 1, lie the new Flash RAID storage arrays. Enterprises who value a “one box” solution more than low costs and extremely high performance will consider these internally mixed architecture solutions. Solidly in Tier 1 and below lie the storage arrays that place some amount of cache before their HDDs to increase performance. These are the most likely to fade away as very fast SSD is paired with slow and inexpensive HDD arrays residing at the bottom of this enterprise storage pyramid.

An Inevitable Mixed Message

Tiered storage developed for the best of reasons: it helps IT managers increase application performance and lower overall data storage costs. You’ll notice that a pyramid makes a good analogy not only to represent the various storage solutions by their performance, but also by their costs, and interestingly, by their data volumes allocated to each solution. In many enterprises it makes sense to store most of the data on much slower and less expensive HDD RAID arrays because the majority of the dataset is accessed infrequently and thus does not affect application performance. A smaller portion of the data is accessed frequently, does affect system performance, and therefore deserves to be moved to very fast SSD storage; doing so improves overall storage performance and increases the productivity of mission critical applications, all while actually driving down costs. 

Tiered storage solutions are not new, but adding SSD to the mix has recently become much more attractive in the average data center for a number of important reasons:

In many cases, SSD can be deployed in a few minutes and appear to the network as just another disk, making SSD easy to implement in a heterogeneous storage environment.

Data center electricity costs worldwide will soon grow into the billions of dollars. Now consider: Flash-based SSD performs around 1,000 operations per Watt, depending on how it’s deployed and configured, while HDD performs about five per Watt.

The price of RAM SSD has fallen by half in the past 12 months; the price of Flash SSD coming on the market is four times less again.

The benefits of mixed storage architectures ripple up and down. Thanks to the SSD component, above the storage hierarchy servers no longer need cumbersome and costly extra memory that IT managers are constantly trying to manage as large caches. Instead, servers can return to doing what they do best, process data. And below the SSD in the storage hierarchy, hard disks are no longer stressed by performance requirements. They can also return to what they do best, store a lot of data relatively slowly and cheaply.

The Mixed Message: Who’s Listening?

Small or medium-sized companies with one application that must go fast, such as e-commerce or e-government portals or video editing or software development toolkits with many concurrent users and much riding on the users’ successes, can connect SSD directly to those specific servers and see startling acceleration, while using low cost HDD for all their less performance-hungry HR, invoicing, and inventory management applications. The same is true with very large enterprises. Think of telecommunications companies or financial institutions where profits are directly tied to the number of transactions processed per second. Think of pharmaceutical labs, scientific installations, or geophysical explorers who make better medicines or more key discoveries when they process enormous batches of data in minutes instead of hours or days. Think of SSD helping enable all this, in concert with higher density less expensive HDD, each used to proper advantage.

The Mixed Message Is Clear

Increased interoperability between available storage technologies and falling SSD prices are leading to new options in storage architectures for forward-thinking IT executives. When deployed together in multi-tiered architectures, SSD and HDD can be powerful allies, each addressing the other’s weaknesses, and each bringing some valuable strengthens. In the near and longer terms, when referring to storage solutions, the message is clear – and it’s mixed.

Neal Ekker is a vice president at Texas Memory Systems, makers of RamSan solid state disk systems and the World’s Fastest Storage, http://www.texmemsys.com. Neal can be reached at This e-mail address is being protected from spambots. You need JavaScript enabled to view it .