Continuous Web Availability

By Keith Tilley Managing Director, Europe Comdisco Availability Services

Nearly one-half of organisations need at least 99% availability for their critical applications. Whether these applications support e-business, enterprise resource planning, customer resource management or other strategic business applications, they need to be up and running all the time

Almost every front-end or customer facing application is today being architected to be Web-facing. For example, in a traditional client/server model, corporate users could have a CRM application running on each of their desktops and in turn connected to the server via a LAN. If one of the users had a failure of an application, only that user would be affected

Today however, with the use of thin client software, all users have to go to a Web server that supports the front-end. This obviously has many benefits, but it also introduces a new component that has to be included in your recovery model. This means that if you use high-availability techniques for your back-end environment, but you haven't developed your front-end to protect against downtime, you don't have a highly available infrastructure

Defining web availability
How do you get to the stage where your Web infrastructure can support your e-business needs? The first step is to understand what Web availability is in the context of the Web. High availability means transparent recovery from system failures and site failures to ensure a continuous Web presence. This boils down to three key components:

Transparency - which is needed to ensure that users never experience a loss of Web presence

Recovery from system failure - that keeps your site going in the event of a failure within an Internet Data Centre, e.g., hardware or software failure

Recovery from site failure - which keeps your site going in the event of a failure that takes down the entire Internet Data Centre, eg, natural disasters or utility failures

To determine whether or not you need this level of availability comes down to establishing the cost of downtime on your e-business applications, i.e., revenue lost and understanding where these disruptions can come from. You also need to consider other risks such as bad publicity

Understanding the sources of downtime and the vulnerability they place on your site is also essential. These sources typically fit into three categories:

Unplanned disruptions - these can include equipment failure, human error or natural disasters
Planned disruptions - this time you take intentionally for hard/software maintenance, data backup or restoration
Peak demand - this can be caused by inadequate infrastructure to cope with high user demand on a site, meaning transactions can not be completed

Creating High Availability Architecture
Your availability solution should include the following:

1. Redundant components - all components within your Web architecture should be redundant. This includes using multiple ISPs and routers, redundant switches and firewalls and having duel power and backup in place

2. Off-site data protection - as with any availability solution, moving your database images off-site is absolutely essential. Many of the high availability techniques used in traditional distributed environments apply to the Web as well. For example, disk mirroring and database shadowing which is dependent upon how quickly you need recovery and whether any loss of data is tolerable

3. Geographic load balancing - with this you have duel protection sites in operation, helping to ensure a transparent recovery from a site failure. One of the critical issues in moving to a geographically load-balanced solution is determining how to handle replicating your databases between the production sites so they maintain synchronization. Options include, using multi-mastered databases that replicate the data in real-time or using a single database master that both sites can tap into

4. A consistent backup and restoration strategy - make it a priority to practice the restoration of your servers from your daily, weekly and monthly backups. It is alarming how many users perform the backups but are not sure if they can restore the systems

While 100% redundant components at two or more sites may seem costly, there are ways to reduce the requirement. For example, if you are using a managed hosting service, the provider may be able to arrange a standby server for you at the time of failure or planned downtime at your primary server. This can help reduce the cost of the redundant equipment, but always remember to regularly test this facility for its appropriateness and effectiveness to your business needs

Another option for companies using dual production sites is to eliminate some of the site redundancy, for example, a standby local database server, where the two sites back each other up. However, this is more convenient than having the redundant equipment on-site and requires modifying your applications to be connection aware. Should the primary server fail, the applications should automatically connect to the standby server

Another strategy is to use your staging servers to perform the task of staging and testing and to be used as a backup for the primary production site

Getting started
Moving to a high availability Web infrastructure is not something you need to do in a single step. There are steps that you should consider taking to make transition easier:

Consider a hosted environment - using a managed hosting provider allows you to economically move toward a high availability infrastructure. The hoster can provide you with an environment that includes multiple ISPs, redundant switches, firewalls and gateway signal routers. Also look at a hoster who can offer burstable bandwidth options that allow you to meet peak bandwidth requirements without having to oversubscribe beyond routine capacity requirements
Move to a load-balanced configuration for your Web servers - this is a fairly painless step, just needing you to purchase a local load balancer. This at least makes your Web servers more maintainable
Protect your database in real-time off site - this is essential just to ensure the currency of your data. But it also provides you with a migration path to eventually Move a secondary, geographically load-balanced production site

Companies that don't think about the importance of a highly available Web infrastructure until after they have a high-volume critical site, often have trouble making the transition easily. However, if you engineer you site early on for availability, as you scale up, the initial pieces are in place to help you move smoothly into a high availability environment

For over 20 years, Comdisco has provided global technology management services designed to maximise all benefits of technology. Comdisco's complete suite of information technology Availability Services includes Business Continuity, Managed Network Services, Web Availability and Managed Storage Solutions. IT Control and Predictability Solutions SM