Blade Servers: The Cutting Edge in Consolidation

New technologies and capabilities give blades a new role in the enterprise data center. By Steve Campbell, Vice President, Marketing and Solutions, Server Systems Group, Hitachi Americas, Ltd.

It’s human nature. The girl who babysat your kids could grow up to become the CEO of a Fortune 500 company and to you she’ll always be the girl who babysat your kids.

Blade servers have done a lot of growing up in a very short period of time. Today they’re capable of a whole lot more than most data center managers think—or even believe. And that’s a shame, because it could translate to a missed opportunity for cutting cost, complexity, and power usage in enterprise data center. This article takes a closer look at the new capabilities of blade servers and examines their new role: as an outstanding consolidation platform for mission-critical applications as well as edge-tier workloads.

Consolidation of Power
Consolidation and blade servers came into vogue at around the same time: five or six years ago. They both became immensely popular as a means of cutting cost and improving efficiency.

Consolidation was the antidote to the dot-com-era “server sprawl” that many companies were experiencing. When you can reduce the number of systems crowding your data center floor tiles, you can save a lot of money in a lot of different ways. You reduce the complexity of the environment, so you reduce the administrative burden. You minimize the cost of managing resources. You reduce power consumption. Your server utilization rates can increase and you can unlock capacity, so you can delay or eliminate the need to make additional hardware purchases. And you’re also improving the level of service delivered by IT, so business agility and productivity increase.

Originally, blade servers were intended to provide consolidation in a different form—a consolidation of power. They emerged in 2001 as a novel way to pack more compute power into less space. A single blade enclosure could provide power, cooling, networking, and various interconnects, and individual blades could be added as needed to run applications and balance workloads. Simple, efficient, practical.

It seems obvious that the efficiencies of blade servers would be an excellent match for the objectives of data center consolidation. Yet blade servers have not been widely used for consolidation in the enterprise data center. Instead they have been deployed at the edge or the Web tier and used for file-and-print or other non-critical applications. Why?

Over the past few years, the limitations of blade servers became apparent. Inadequate scalability, questions about reliability, compromises in I/O and other capabilities, and excessive heat generation and cooling costs caused many data center managers to seek other alternatives. They felt they could not really rely on blade servers for mission-critical applications.

Today there is a new breed of blade systems on the market capable of handling mission-critical application workloads and ready to deliver the full benefits of enterprise-class consolidation. At the same time, blade server systems are getting even more power-packed, lightweight, scalable and energy-efficient, making them well suited to consolidation at the edge and application tiers of the enterprise data center. Here’s a brief overview of the new technologies and capabilities that make all of this possible.

The Advent of the “Enterprise-class” Blade Server
Data center managers have taken two very different approaches to consolidating mission-critical applications. One approach is to go out and buy the most powerful, massively scalable, sophisticated SMP system on the market and load it up with applications that used to run on multiple servers, thereby simplifying the environment. Another approach is to buy many smaller server units that can be racked and stacked right up to the ceiling; and if one fails you simply swap it out and there’s no disruption. This is “consolidation” in the sense that you can standardize the server and operating system platform and actually cut the total floor space needed for your compute power.

The new breed of blade servers combines the best aspects of both of these approaches.

Hitachi, for example, has introduced the Hitachi BladeSymphony 1000 blade server, the first system that combines enterprise-class reliability and scalability with built-in virtualization technology. This system was introduced in Japan three years ago as the first system in the industry capable of delivering the level of performance, scalability, reliability, and flexibility needed to run mission-critical applications with total confidence, and it has been extremely well received in the marketplace.

The built-in virtualization feature, which Hitachi calls Virtage embedded virtualization, is an excellent example of how the new breed of blade systems can meet the stringent requirements of the enterprise data center.

Virtualization solutions allow computing resources to be pooled. Virtualization makes it possible to allocate processing power on demand, and it can help boost the utilization rates of current systems, so IT doesn’t need to buy as many new systems. That makes it a good option for consolidation. However, there are problems with traditional virtualization solutions. One is that they’re typically third-party solutions, so IT management has to evaluate multiple products to find the best one, configure it for their environment, learn how to install and use it, and troubleshoot problems that come up. The other issue is that virtualization software often slows down applications.

What Hitachi has done is embed mainframe-class virtualization right into BladeSymphony, so there’s nothing required of administrators to make it work right. It just runs, right out of the box, at no extra cost, thereby reducing total acquisition costs. Virtage is based on Hitachi’s years of experience with virtualization in the mainframe world, and has proven to be both simple and reliable. It is true “Hypervisor” virtualization, not an emulation technology, so it’s naturally faster and does not slow applications down. It is implemented in the system’s firmware and reduces the overhead associated with software virtualization, increasing performance.

Another key element of the “enterprise-class” blade server is robust reliability, availability, and serviceability features. The BladeSymphony chassis components are hot-swappable. The system uses Hitachi’s mainframe-class memory management; there are redundant switch and management modules; the backplane and I/O are reliable; power supply modules are N +1 or fully redundant; and the event of a failure the N+M cold stand-by feature automatically detects the fault and provides fast recovery.

Of course, an “enterprise-class” system also has to deliver the performance users expect on 64-bit applications; and it has to scale to meet unexpected spikes in the workload. Hitachi has addressed these needs by providing support for up to eight Intel Dual-Core Itanium 2 blades or Intel Dual-Core or Quad-Core Xeon blades in the same 10U chassis. It is possible to mix and match the processor modules, and scale up or scale out the system.

For example, BladeSymphony 1000 scales out to eight Intel Dual-Core Itanium 2 processor-based server blades in the same chassis, and it scales up to single or multiple 16-way SMP with 4 or 8 Intel Dual-Core Itanium 2 processor-based server blades.

To be enterprise-class also requires a higher level of flexibility than traditional blade servers have delivered. For example, the system should support industry-standard operating systems like Windows and Linux, and it should allow customers to use their existing, standard PCI-X and PCI Express I/O cards. The system should allow you to consolidate at the database tier, the application tier, the edge tier, or all three. BladeSymphony is the only blade server that does that.

By fully delivering on all of these requirements, the new generation of blade servers, exemplified by Hitachi’s BladeSymphony, can now serve as a consolidation platform for mission-critical applications. Rather than purchase a massively expensive SMP server, IT can now use this system and replace rack-mount servers, cutting costs and saving space.

Pushing Consolidation to the Edge
Large enterprises are not only looking to consolidate their mission-critical workloads; they want to push the benefits of consolidation right to the edge—meaning they want to pack even more power into ever-smaller spaces while cutting total power consumption, and they want a system that’s simpler to deploy at the edge and application tiers of the enterprise data center.

Blade systems are now emerging to address both of those needs. Hitachi’s entry in this marketplace is a 10-blade, 6U system that combines dense compute power with a level of simplicity that makes it well suited for edge- or application-tier consolidation. Hitachi’s system, the BladeSymphony 320, comes with a 110-volt power option, so it plugs right in to standard power outlets with no special equipment or adapters. And it draws significantly less power than any rack-mount server that delivers comparable performance—and far less power than fully loaded traditional blade systems—so it saves money.

BladeSymphony 320 also exemplifies the new-generation of blade systems by packing extreme levels of compute power into extremely small spaces: up to 10 two-socket, quad-core servers in a single 6U chassis—with 60% space savings compared to rack-mount server solutions. For large configurations the system can pack 70 two-socket, quad-core servers in a single standard 42U rack—that’s up to 560 cores in a single rack.

We’re also seeing significant improvements in reliability, availability, and serviceability features in the smaller 6U blade servers systems--from hot-swap components and multi-configurable power supplies to enterprise-class “cold stand-by” capabilities for automated system failover. Hitachi’s 6U system also includes two on-board, hot-swappable Serial-Attached SCSI (SAS) drives, providing high performance and application flexibility; four Gigabit Ethernet ports per blade for high network throughput; and up to 16 GB of fully buffered memory.

And the trend toward ever-simpler administration continues: fir example, a BladeSymphony 320 server can be installed and brought up in a few minutes rather than the few hours it would take for a rack-mount server. A wide range of optional capabilities and components are available, such as a LAN pass-through option, Fibre Channel ports and switch modules for connecting to a SAN, and remote IP KVM. These components are all built in, so there’s less cabling, less complexity, and less work required of administrators. And in the case of the Hitachi 6U system, there are no space-consuming external devices to purchase and configure, no compatibility issues to deal with, and no worries about integrating with the management system.

About Hitachi and Hitachi Americas, Ltd., Server Systems Group
The success of the Hitachi BladeSymphony line in Japan led to the launch of Hitachi’s U.S.-based server systems group, which now serves the entire North American market. Hitachi is already a major player in the worldwide server market, having introduced its first computer in 1958, the first mainframe in 1961, and the first supercomputer in 1982.

Hitachi is an $80 billion dollar, globally diversified company, and now has a very strong presence in North America, with more than 15,000 total employees in the U.S. alone. The group is well connected with Hitachi Data Systems, which makes some of the most powerful and sophisticated storage products in the world, as well as Hitachi Global Storage Technologies, Hitachi Consulting, and many other Hitachi businesses within the United States.

Steve Campbell is Vice President, Marketing and Solutions, of the Server Systems Group at Hitachi Americas, Ltd. Before joining Hitachi in 2006, Steve served as vice president of marketing for Sun's Enterprise System products, where he was responsible for leading Sun’s mid-range and high-end platforms together with data center initiatives focused on consolidation, mainframe migration and high performance computing. Before joining Sun, Campbell had extensive experience in mission critical and supercomputing.