This is the inside of the brochure, unfolded:

Don’t feel bad, it happens to all of us: eventually the diet and exercise that kept us lean and mean in our younger days doesn’t do the trick anymore. The same applies to enterprise IT, especially today. The amount of information flowing through your business is increasing exponentially, and users and applications alike demand faster, more personalized delivery and improved access to it. If you’re like most IT executives, you’ve noticed that your software-based middleware is bursting at the seams, and you know that companies don’t get love handles when they aren’t doing what it takes to stay fit—they lose customers and market share.

That’s why hardware-based messaging and content routing middleware is the future of enterprise information infrastructure. Wherever performance is critical and requirements repeatable, purpose-built hardware eventually rules the roost. The Internet wouldn’t work without high-capacity routers and switches, and high-power graphics cards are the foundation of the multi-billion dollar gaming industry. Middleware is at the same crossroads that networking was at in the 80’s when routing over LANs and WANs was performed in software until Cisco unlocked the commercial potential of the Internet with a hardware approach.

Why do I believe the time for hardware-based middleware has come? Financial institutions require the delivery of millions of messages a second, with latency frequently measured in microseconds, to satisfy customer demand and government regulations. Telcos capture and keep customers only by providing increasingly sophisticated services that disseminate real-time content based on highly-specific (and frequently changing) preferences. All kinds of enterprises are unlocking the value of their information infrastructure by finding ways to use real-time data to identify and address risks and opportunities as they unfold.

These requirements are exceeding the capabilities of software-based middleware, and architects and developers have squeezed so much performance out of software-based solutions that there’s just nowhere left to go. It’s a dead end road, because at some point the underlying operating systems and generic servers become the gating factor. Workarounds like server virtualization and increasingly massive datacenters help in some regards, but introduce their own challenges in terms of complexity, cost and scale.

So it’s not a matter of if hardware will take over as the technology we use to route and transform content as it flows over local and wide-area networks, but when it will do so—and I believe the time is now. I’m not talking about accelerating software or bundling software with a server, but embedding the data path in silicon to eliminate the unavoidable bottlenecks caused by software.

Hardware-based middleware improves information infrastructure in six areas:
1.Performance. By eliminating the interruptions and inconsistencies of OS tasks, application interfaces, and storage reads/writes, hardware can handle millions of messages per second with latency under 100 microseconds. It’s finally possible to sustain over 8 million messages a second without compromising latency. In large datacenters, often 20-30% of the CPU utilization is devoted to network IO and the TCP stack. TCP acceleration in hardware not only means better network throughput, it also means your software will now run more efficiently with fewer kernel interrupts. Your CPU can now be devoted to business processing, not network processing.

2.Reliability. Every firm suffers from multicast storms—one slow consumer kicks things off and by the time the storm has been diagnosed, the damage is done. Hardware-based routing lets consumers subscribe to only the data they’re interested in, eliminating the inefficiency and risk of having millions of packets flowing over the network to all consumers. Just as impressive is hardware’s ability to support truly guaranteed messaging—for applications likeorder processing and trade execution—at an astounding rate of 100,000 messages per second.

3.Availability. Most reliable messaging software can’t effectively recover from a catastrophic failure, but a pair of hardware devices can shadow each other and offer recovery and full client reconnectivity in just seconds.

4.Predictability. Software is susceptible to latency under load— usually exhibiting outliers of hundreds of milliseconds, and 5-10 second delays occur frequently. Since hardware isn’t subject to OS processes or context switching it can deliver the performance described above even at peak volume with outliers no more then 50-100 microseconds.

5.Flexibility. Hardware can accelerate the deployment of services by supporting many messaging requirements on one platform—everything from high fan-out messaging for content distribution to guaranteed messaging for the execution of transactions, along with intelligent routing and transformation for applications like complex event processing, fraud detection, and more.

6.Cost and complexity. Hardware-based middleware can reduce the datacenter footprint by 10:1 or more. Less servers, software licenses, maintenance, power and cooling—each reduction contributes to a less complicated and costly infrastructure.

On any dimension, hardware has compelling advantages over software, the kinds of advantages that can help you get lean and mean to not just survive, but thrive, in an environment of ubiquitous information and access.

This post is password protected. To view it please enter your password below:

Password: