There is a growing trend of open source-based applications running mainstream business-critical applications. Part of this is due to the availability of new solutions software that, when added to the open source application stack, greatly expands the overall reliability and scalability of the resulting applications, allowing them to be "enterprise class."

TCO. That simple three-letter acronym is widely used and also widely misused. It's not the initial key driver for critical business purchases even if some vendors would like you to think that it is.

Let's consider the issue in more detail. Linux is often cited as a lower-cost alternative to proprietary systems. The lower TCO is a result of lower software licensing costs combined with commoditized and inexpensive hardware. But there's more to TCO than that.

You are a manager who needs to evaluate and purchase the needed software components to build and run a credit card billing system. This system is the heart of your business. If it's down, your business is down. If there are transaction bottlenecks - or worse, lost transactions - your customers are not happy, to say the least, and your manager won't be either. Imagine explaining to an irate customer that his or her problem is because you opted for the lowest TCO solution. I wouldn't want to be handling that phone call.

To be fair, lowest TCO can be the decision driver for applications that do not have high service-level requirements: ones that don't need to run 24x7 and can fail without impacting your business. We find many open source applications running such applications. This is often the stealthy entrance of open source products into Global 2000 companies. It's clear that there are not many budget dollars for applications that are nice to have but have low-business impact.

Widely held as the fastest growing database, a great deal has been written about the success of MySQL and, as an open source application, it gives users a low TCO among its many other benefits. However, some of this growth is due to its use in very simple, single user applications. For example, Gartner did not include MySQL in its new database study because the researcher didn't see the company's database as "enterprise class." An AMR Research study surveyed 140 information technology managers about open source databases. Overall there was satisfaction with price, performance, and ease of use of the open source database systems, but satisfaction lagged when it came to "scalability," the ability to handle large workloads, AMR noted. Further, it has been reported that some companies simply outgrow MySQL and determine that they need the "scalability and server redundancy of Oracle's database clustering technology."

Then, TCO by itself is not enough to drive key business-critical application decisions. A key driver seems to be "enterprise class." But what does "enterprise class" mean? The answer can be seen above: scalable or the ability to handle large work loads, redundancy, and close to zero downtime. In other words, meeting the needs of real-world, mission-critical applications, not tracking my wine collection.

How can a full, open source application stack be "enterprise class"? By incorporating an additional layer in the solutions that's built to deliver these missing capabilities: Oracle-like scalability and redundancy on Linux, MySQL, JBoss, and Apache. The resultant application will be "enterprise class" and deliver low TCO. Figure 1 illustrates such an architecture.

Emic offers transparent application clustering solutions for open source Web platforms and applications built on SQL databases, application servers, and Web servers. Emic Application Cluster (EAC) software provides an efficient and cost-effective solution for creating scalable, high-performance, highly available database applications and Web services with dynamic load balancing and seamless fail-over and recovery. Emic provides the single-point-of-manageability and application-level clustering for MySQL database server, Apache HTTP server, Tomcat container, and JBoss application server for J2EE deployments.

While there are hardware replication options, such as shared disks or file systems, as well as replication options in some open source projects, such as bind (DNS) and OpenLDAP (directory services), these focus on solving a different problem. Emic's application clustering solves the ordering problem. Solving this problem guarantees that the correct answer is generated by the application all the time no matter what physical server processed the user's query. This cannot be done with lazy or asynchronous replication.

Figure 2 shows how the management of a cluster can be accomplished. It's one thing to build a cluster and yet another to manage it. Cluster management requires servers to be taken offline for maintenance and brought back online. Therefore, you must be able to reliably manage it and nominate a donor node to be the node that will update the formerly offline node so that it will be brought back in sync with the current state of the database. Manageability is key.

Figure 2 shows how the cluster is managed with the Emic Application Cluster management console. The upper-left frame shows a simple three-node cluster with one node in standby mode. This type of configuration is very easy to create and will withstand a high transaction rate as well as be quite cost effective.

For example, CanadaDrugs.com operates its entire workflow application on open source, commencing from an initial pharmacist review of every new prescription, which involves a minimum of three pharmacists throughout the dispensing process therapeutically screening and assessing prescription orders for appropriateness, accuracy, potential allergic reactions and interactions, duplication of therapy, even direct contact with the patient's U.S. prescribing physician for verification concerns and all possible drug-related problems. Following the pharmacists final checks, approval, and release the prescription order is billed, then mailed. CanadaDrugs.com keeps the system up and running with Emic's Application Cluster.

Another case is Securities America, which is part of the American Express family of companies. It has 1,500 independent brokers and almost 400,000 individual account holders.

"Nobody has a database cluster like what Emic is able to offer. In almost every instance of a commercial database cluster, they use a single disk copy of a database. With Emic, we have an individual copy of the database on each node. That gives me a much higher level of availability and allows me to run it on much cheaper hardware," stated Joe Dennick, IS director at Securities America.

Open source is here and open source application stacks can move into business-critical applications when there are additional software solutions that add the "enterprise class" functionality to the resulting application. You can meet the enterprise class requirements of availability, scalability, and reliability and still deliver low TCO. Penguins can wear more than tuxedos - they look good in business suits, too.