Storage Hardware

After spending last week at SNW to present on 6Gbps SAS and virtualization, I left with a lot on mind. First, it was a great show and SNIA and Computerworld are to thank for that. Their professionalism in running an event are second to none.

Everyone has asked me two things. First, I get asked, “How was attendance?” Frankly, it was light. However, it seemed that there was a significant presence of smaller organizations from the region attending. It was refreshing to discuss their strategies as we often get consumed by thoughts of mega data centers and can easily forget about issues that impact smaller organizations.

For instance, one small IT shop was trying to develop a disaster recovery program for their one and only server farm. While we might envision a world in which that DR site would be hundred, if not thousands, of miles away, they simply wanted to replicate their server environment four miles away. Why? Because someone might steal their servers and walk away with the company’s entire database! Talk about needing some insight on the best buy to replicate data! SNW provided that type of information and made this end-user's time at the event purposeful.

The second question I get is “What was everyone talking about?” It seemed that the usual hot topics received the most attention: SSDs, virtualization, dedup, tiering and green. A few thoughts on each…

SSDs – the Lamborghini of storage. My colleague at Microsoft, Chris Lionetti, offered up this nugget…”if you’re not already short stroking your drives, why would you use SSDs to improve your performance?” Good point on performance. However, I would offer that many are concerned with power savings, but at the current price points that may require a long payback period.

Virtualization – the catch-all phrase for everything cool in high-tech. Many end users are confused by the term "virtualization." Why? Because every supplier in the world is using the term in different ways. For environments in which you've deployed typical virtual OS (VMware, Hyper-V or Zen), I recommend that you choose storage solutions that allow you to aggregate your storage into a common pool, is built upon a Unified Fabric of Ethernet and is highly optimized. Might I suggest EqualLogic? Sorry, couldn't help myself.

Dedup – a classic duh. Who wouldn’t want to dedup data whenever possible? Backup data is a great place to start.

Tiering – a great architecture allowing you to deliver high-performance SAS drives for applications and lower-cost, lower-performance SATA drives for less frequently accessed data.

Green – we’re all getting greener…more efficient power supplies, lower power SSD drives, deduping data, tiering to put less used data on lower power drives, virtualized environments that more fully utilize the hardware we deploy, energy efficient Ethernet. This is not only great for the environment, but it also reduces your power expenses, which aligns your goals more completely with your organization’s needs even if you don’t directly own the budget for that expense.

With President Obama’s cap-and-trade proposals it may have material impact to your organization’s bottom line in the years to come.

Next up…Interop. Hope to see you there!

Have you seen the latest?  Storage Magazine ranked Dell the #1 midrange array provider in service and reliability.  With our acquisition of EqualLogic, we turned last-year’s 9^th place ranking into a #1 ranking in product reliability and initial product quality.  Hopefully, you didn’t miss Dell’s other recent achievement where Gartner recognized Dell as a leader in Midrange Disk Arrays for their Magic Quadrant 2008.

Dell has made a tradition out of listening to you, our customers.  In turn, you’ve rewarded us with your support.  We’re excited to be plugged in, so we can keep building storage solutions you want.

Thanks for all of your feedback.

In my last post about blade servers, we focused on price, and the TCO savings blades can provide. The E-Guide on Blade Server Trends also tackles the myth that blades require too much power and cooling. Here’s a snippet from that guide:

[It’s true that] early blade systems generally used more power and cooling, even with lower speed processors and less memory. Today's blades have more efficient power supplies and improved overall thermal design. Many also have sophisticated software to help in managing heat and automating certain actions to address problems, including powering down components as needed when the temperature gets too high.

There is, however, a power and cooling issue with blades in terms of density. Even though one blade uses less than one rack server, since blade systems are designed for high density, the number of blades per footprint can be substantially higher than rack servers. (This is fact is one of their biggest advantages in space savings.) So it's important to pay attention to power and cooling issues from an overall data center planning perspective, and to ensure that you plan appropriately.

So, the benefits of blades won’t go against your business’ green goals. When planning your data center, how do you factor power and cooling factor into the discussion?

One of the primary ways to simplify IT infrastructure – and save in the process – is through server consolidation. Blade technology is a good way to approach this, but many IT managers are reluctant to take the blade route. There are several myths about blades that create the impression that blades aren’t the best choice for consolidation. In a recent E-Guide on Blade Server Trends sponsored by Dell and Intel, virtualization expert Barb Goldworm attacked these myths.

Her comments are worth reading:

Because blade systems require an up-front purchase of a blade chassis, one misconception is that blades are a more expensive solution than rack servers. In fact, if you are only implementing one or two servers in a single location, this is true. However, if you are implementing four, five or more servers, the total cost per server can actually be lower, because of the shared components within the blade chassis. Just calculating the numbers on hardware alone (without counting other savings such as power, cabling, and management), the costs for a blade server can be lower than a comparable rack server.

One blade customer did a very basic comparison using Dell blades, calculating the cost per server based on the blade cost plus 1/10 of the chassis cost. He estimates the blades saved him 20% over comparable rack servers. While these numbers vary by vendor and configuration, the bottom line is that unless the chassis is mostly empty, blades cost less.

Stay tuned to the blog next week for myth #2.

Storage area networks (SANs) have greatly simplified the management of storage resources. By creating a centralized storage pool that can be allocated as needed, SANs have enabled many organizations to reduce both management and capital costs. Now, an approach has emerged to take SANs to the next level, with further simplification and savings. It's called thin provisioning.

Thin provisioning helps IT departments attack the fundamental problem of storage under-utilization. In many businesses, a significant portion of purchased storage is not in active use, although it is kept constantly available -- taking up real estate in the data center, consuming power, and increasing the load on data center cooling equipment. In addition to wasting opex dollars, there is a capex penalty to this approach. By buying storage capacity "too soon," IT organizations fail to benefit from the constantly falling cost per GB that each new generation of storage technology provides.

In essence, thin provisioning changes the process of allocating space on a SAN so that physical storage resources are only committed to a volume when data is actually written to it, eliminating significant waste.

Currently, to allocate space on a SAN, an IT administrator defines a logical volume and makes it available to a server. The SAN controller then responds by reserving physical space on the array of disk drives that it manages. Every byte of storage determined by the volume size is backed up with a physical location, whether it is used or not.

With this approach, unused space is effectively wasted, because it cannot be reclaimed for other volumes or applications. With thin provisioning, this space is available, because each volume claims physical resources only when needed. Furthermore, the allocation process is automated, so that administrators don't have to spend extra time.

The net result is that IT organizations can come much closer to the goal of purchasing storage when it's actually needed, while reducing the complexity of the administration process.

You can find more detail in my recently published article in Dell Power Solutions.