| This silo based methodology practically always results | | | | Then contemplate the complexity of integrating |
| in the use of several different availability / DR | | | | differing availability technologies from multiple vendors. |
| technologies from various vendors, with considerably | | | | Are they guaranteed to interoperate with one |
| different designs, capabilities and limited/no integration | | | | another? Is such interoperability built in (unlikely) or will |
| points. For example, an online web ordering system | | | | some degree of customization and manual scripting |
| may make use of network load balancing for front | | | | (very possibly) be needed, so that every tier can |
| end web servers, some form of data mirroring or | | | | communicate with the other tiers? If custom scripting is |
| clustering for backend database servers, and a 3rd | | | | required, what happens when even a single part of the |
| party availabilityalternative for middleware. Point of sale | | | | availability architecture changes? Will extra, custom |
| solutions, CRM tools, and even BlackBerry messaging | | | | consulting work be necessary to develop and re test |
| environments implement a similar prescription, utilizing | | | | existing scripts? Last but not least, if and when |
| completely different technologies for each layer in the | | | | something breaks down, whose responsibility is it to |
| application stack. | | | | identify the root cause? With different solutions from |
| Having such an approach to running a business | | | | different vendors, one must be wary of the inevitable |
| continuity solution for your enterprise application | | | | finger pointing that may result when things go wrong. |
| ecosystem has several drawbacks. First and | | | | Naturally one alternative is to simply not integrate the |
| foremost, one must be aware of the cost implications | | | | solutions, after all, so long as each portion is doing its |
| of utilizing different technologies within a continuous | | | | job, isnt it safe to assume that the entire system is |
| availability or DR architecture. The most obvious cost | | | | operational? Not necessarily. Consider for example the |
| is the equity outlay for the hardware/software itself. | | | | deployment of a multi tier, distributed architecture |
| By choosing (or being forced) to apply different | | | | across physical sites for DR purposes. If the entire, |
| solutions from different vendors, there is no possibility | | | | primary production site fails, will the servers start up in |
| to leverage economies of scale. Most hardware and | | | | the correct order and fashion at the remote site, or will |
| software companies offer volume based pricing | | | | some degree of interaction be needed from an |
| enticements for larger order sizes, but this possibility is | | | | administrator? |
| obviously squandered when various alternatives from | | | | Now examine the more probabletype of failure, when |
| different vendors are employed. | | | | just one component instead of an entire site fails. |
| Additionally if each solution leverages different | | | | Unless youve deployed a combined High Availability |
| underlying hardware, disk, or OS technologies, an even | | | | Disaster Recovery solution, chances are that the single |
| larger total cost of ownership will be noticed. Of | | | | failed component will resume operations at the DR site. |
| course cost extends beyond just hardware and | | | | But in most cases, the latency between sites will be |
| software to include implementation, coaching, | | | | too high for any multi tier application to function |
| andcontinuing management costs. Think aboutdeploying | | | | properly. In this situation, its best to actually fail all of the |
| even a pretty basic, three tier application architecture. | | | | components across to the remote site as a single, |
| In the online web ordering example discussed | | | | cohesive unit. But once again, how does this |
| previously, one would need to take on the somewhat | | | | coordination take place? Either we are back to |
| unnerving task of learning about not only the intricacies | | | | scripting the failover in some manner, or else some |
| of SQL clustering, but also deployment and | | | | hands on administrator involvement is needed. When |
| management of network load balancing and any | | | | that takes place,recovery times inevitably increase; |
| middleware components needed. Every time a new | | | | when recovery times increase, so does the bottom |
| variation of any of these solutions is made, theres the | | | | line cost of the outage to the business. |
| additional cost of relearning a new technology. | | | | |