Posted on Wed, Jun 16, 2010 @ 01:16 PM
WordPress offers blog services to about 10 million users. It is a true SaaS application designed with a multitenant architecture. In this architecture all users are "tenants" sharing the same database and application logic and are in virtual isolation from each other but physically in the same building. Recently WordPress had an outage which affected all 10 million users. But what was different in this scenario vis-à-vis network issues, is the outage was caused by a code change to the application.
In multitenant SaaS applications new changes are rolled out on a regular and frequent basis. Every customer is always on the latest release. However when something goes awry, everyone also gets the hit. Performance, security and service levels are all dependent of the design of the application as well as the datacenter infrastructure where the application resides.
What may be a potential solution to the ripple effect inherent in SaaS applications is a move to a "multi-apartment building" concept. In this model you have the tenants grouped into different buildings as opposed to the current model where everyone is in the same building. Using this model changes could be tested in buildings or groups of customers before being rolled out to the general population.
Having such a design would also enable testing for security vulnerabilities, specific performance enhancements and for evaluating the impact of new code changes. Therefore, when in the market for SaaS solutions, it may be wise to ask potential vendors if their application supports multiple apartment buildings or is just a single building design. You may avoid a lot of anguish in the future by making the right choice.
Posted on Wed, Jun 10, 2009 @ 12:43 PM
Recently I attended a webinar that discussed the need for a business continuity plan (BCP) in case a pandemic virus was to affect hundreds of thousands, if not millions, of people around the world, possibly in your area. Obviously playing off of the fears from the recent H1N1 (a new flu virus of swine origin according to the CDC; http://www.cdc.gov/H1N1FLU/), the presentation seemed to focus on how your business would continue to operate if such an event were to happen.
To be sure the possibility exists of regions/localities being closed off from access if the outbreak was to happen, but I think the presenter was missing the bigger picture.
I don't think you need a business continuity plan for a pandemic. You need a business continuity plan for any event that can cause business operations to slow down or potentially stop for multiple days. Let's not specify a BCP for a pandemic, rather let's create a BCP that includes measures for situations that would not allow workers into the building, street, or city for a certain period of time. Certainly a pandemic would qualify for this, but so would many natural disasters including hazardous chemical leaks/explosions massive civil disturbances, etc.
Trying to plan for a specific event is tricky at best - it is almost impossible to guess every potential situation that may cause your workers to not be allowed into their workplace. Do you have a separate pandemic plan for a one-day outage? Two-day outage? Five days? You get my point.
I would suggest that rather than planning for specific events, instead plan for estimated days of non-access regardless of the reason. Your BCP should have plans for multiple day events, multiple week events, and at least a framework for multiple month events. While multiple-month events may seem far-fetched one only has to remember September 11, 2001 or Hurricane Katrina to at least give some consideration for the possibility.