Prior to our public launch of www.knowledgeplaza.net, we had been working with the enterprise edition of KP. We have one on-premise client using KPE (client in-house infrastructure), rest of cases are SaaS (software accessed remotely). But moving from a on-demand model to an automatic online sign-up open for everyone requires an intensive workload from the development point of view. I have briefly collected some insightful opinions from some of our developers so that they share with us their experiences on this massive swift of going fully ‘cloud’.

According to Raphael Slinckx, lead developer at Whatever, “the main challenge was moving from a manually operated instance creation and administration to a fully automated process that allows people to write their name in a box and have a running instance few seconds later”.

As SaaS service, the various elements of each instance are dispatched across several machines both for load
spreading and fault-resilience. They are registered in a central repository so that we can easily manage all those processes running on the various servers. We have also deployed a virtual machine creation automation to facilitate the process of adding servers.

The integration of a billing-payment gateway is also a crucial aspect here. And we don’t mean just the paying process. Now, users can get more or less service provision with a click of a mouse and that needs to be implemented in the architecture in a robust manner.

We also want to emphasize the importance of having a private beta period, not only to fine tune the service by fixing bugs but also to test the server load. We monitored the server load to see if it grew according to our model (and adapted the model as needed).

SaaS might be still in its infancy (we’ll leave that discussion for now), but net users aren’t. As technology evolves fast, we are tuning into advanced users with a -logically- lower tolerance to slowness or irresponsiveness. Server load testing processes are necessary to get ready for ‘massive consumption and usage’, to provide a good responsive service to that end user.

Likewise, as Patrick Rácz -developer with UI focus- points out, “adoption is more volatile in the cloud“, meaning end-users are also getting used to user-friendliness on cloud services. Thus, enterprise 2.0 apps in the cloud have some quite challenge here in terms of usability.

One of the main concerns of companies considering a move onto SaaS enterprise 2.0 is data migration. Currently, there are no standards (just like there is not yet cloud interoperability or ‘cross-cloud compatibility’). If you have your data on a system and you’d like to start using a different one, exporting those data won’t be an easy task, at all. Given that the underlying concepts of apps in the cloud may differ from each other, you can’t have a “standard” to represent them without information loss. Any migration would have to be done the way we handled migration from data into KP for other customers: 1) extract the old data in “some” readable format, 2) make a manual analysis on the best way to map the old data model to the new data model and 3) create a custom written conversion script.

So, due to that lack of standards, we offer assistance service with data migration. Upon request, as we’ve done with companies using Knowledge Plaza Enterprise, we can either import or export a database dump of all the data. So far, it’s a manual process trying to ensure and validate data integrity.

Finally, apart from a challenging move, they seem to agree that this process of moving onto an automated online sign-up has been time-consuming. Indeed. And now we have new milestones ahead.

No related posts.