Information Technology Architecture is a funny thing, especially when it's applied to the information technology sector. Delmarva Group believes in providing enterprise class service to all of our customers whether they're hosting a single website with us or oustourcing a major web facing application. The fundamentals of our architecture are the same.

The internet utilizes technology that allows traffic or packets to take any number of routes to get to their destination. Why then, would an organization provide only one place for those packets to go? This thinking lead Delmarva Group down the path of full systems redundancy that starts at the very beginning, when you type in a url that points to a website that's provided by Delmarva Group.

Like all other professions we draw on our past to enhance our future. Much has been learned and developed in the areas of systems redundancy and clustering. Let's examine some of that technology.

Redundant Hardware

One of the first things developed computer side of things was redundant hardware. The idea here is to take the most common failing elements of a computer and add more of them so if one fails, the other takes its place. For starters, anything with moving parts is a high risk point of failure. This primarily includes hard drives, power supplies and fans. Many servers today offer RAID drives, redundant power supplies and several fans in an effort to keep the server running in the event of a failure.

Does it work? Ultimately, yes. Experience has shown that these hardware systems do help keep servers available in the event of a hardware failure. But it's not enough.

Clustering Technology

Ditigal Equipment Corporation (DEC) was one of the first computer companies to mainstream clustering technology. Today, there are many different organizations and strategies to clustering technology.

The general idea of clustering is that you have several dedicated servers that can provide the same services running against the same data so if one server fails, the other(s) will pick up it's work hopefully without end user interruption.

Why is clustering so important when we have redundant equipment in our servers? This question has lead to many, many discussion in the IT planning rooms throughout the world. At Delmarva Group, the answer is a simple one. Even if you have a server when everything redundant, you still have a single server which equates to a single point of failure. In an enterprise class environment, there cannot be a single point of failure.

That thought extends to a cluster as well. For example, if you have 2 clustered servers, you still have a SINGLE cluster.. .thus a single point of failure.

Backing Up

So what when the system does go down. What now? Professional grade services are backed up on a regular basis. Frankly, backing everything up is the easy part. Restortation is the key to any disaster recovery strategy. Unfortunately, too many organizations overlook this very important feature.

In a case from the not so distant past, a company was backing up giga bytes of data to tape. This company implemented a single clustered pair of large servers to provide services to a 20,000 person company. Everything was okay until they had a failure that required them to restore from tape. Unfortunately, this took far too long. With the technology choices made, services could not be restored until the restoration process was complete. In short, the client was left without services for almost 30 hours.

This unfortunate incident resulted in much lost time for the client company. Ultimately the client left the company providing the service.

What would have saved this situation? We'll backup strategies need to be designed backwords. In other words, the goals of the restoration part of the process need to be defined first. Only when that's complete can a proper backup strategy be developed.

How Far Does This Go?

Now that we have clustering and servers with redundant components, what's next. Network redundance is crucial to being able to deliver enterprise class services to our clients.

Many organizations, especially small business, have the ability to connect to the internet in several ways. Cable/DSL modems are quite popular in this segment. Full T1 lines are often employed by larger organizations to provide internet access for their employees.

The problem is again, that many of these systems introduce a single point of failure. For example, a company could provide a fantastic architecure of all clustered servers with RAID drive systems... or even a Storage Area Network (SAN). But if those services are provided to the internet or clients over a single T1 line, all that great technology is still dependent on a single CSU/DSU and/or router connected to a local T1 provider.

Additionally, while many providers will tell you that they are redundant, they run out of a single facility. Okay, granted, the best ones run with multiple network providers, extensive power backup system and such, but isn't that still a single point of failure? At Delmarva Group our systems are setup at multiple facilities. This ensures no single point of failure from start to finish, top to bottom, the web browser to your website.

Delmarva Group Philosophy

We at Delmarva Group construct systems and services that are redundant from the very beginning of the process. Our design motto is simple: no single point of failure. This includes everything from Internet Access points to power to networks to servers.

In practice, all of our hosted webservices have several avenues from the internet into our serving facilities. Our goal is to have 2-4 access points for each and every site we host. This will provide service redindancy at the highest level. So, if we have a router fail, our customer services are still operational. Traffic is automatically routed to another facility in that event.

You may ask if we use RAID systems and serves with redundant power supplies. We do, when we feel it's necessary. Servers that require large amounts of disk space, such a email servers, are built with such technology. But the key is, the service must not go down.

When it does go down, the business MUST encur MINIMAL impact. All of our architecture decisions are made by asking this question: How can we minimalize business impact in the event of a failure.

This has lead us down a highly redundant environment that involves clustering, multiple IP address per website. How do we do it? We've combined the best hardware we could find with the best software we could find with some innovative architectures to achieve and build our current infrastructure. Our goal... simple... %100.00 availaiblity. To date, we've achieved it.