[Episode 07] Migrating the instant messaging service to Google

The history of our messaging server began in 1996 with Oléane which included 10 accounts. In order to optimize costs, we internalized this service in 1998 and managed our server on our own with a sendmail (up to 80 accounts). In 2002, in order to improve services and facilitate administration, we deployed Extremail, an OpenSource solution in linux (up to 250 accounts). In 2007, we decided to reinforce our messaging service and have a multiuser calendar program by deploying Zimbra, initially with its community version and then with the commercial version in order to have the push mode for mobile equipment. We now have more than 600 users and we are trying to further improve the service in order to make sure that we reach the 1,000 or 2,000-account milestone in the best conditions.The aspects that we needed to improve were:* Increasing messaging availability: In 2010, we had 600 minutes of unavailability due to different reasons and never at the right moment.* Reducing mail delivery time: With the arrival of significant mail flows, the antivirus and antispam server delayed the delivery of emails by a few minutes (sometimes by up to 30 minutes).* Reinforcing security: Having a better performing antispam and antivirus engine.* Having a storing space: This means that we need to offer everyone the possibility to leave their emails on the server in order to ensure better preservation.And all this starting from what we already have: push mode for mobiles, intensive use of distribution lists, user-friendly administration. We also needed to reduce the budget allocated to platform administration and maintenance. Even though we have the adequate infrastructure, we didn’t need to maintain the platform in our infrastructure.We have compared the capabilities of Zimbra 7 and Google Apps for Business. On paper, Google has a high rate of availability (30 minutes of unavailability in 2009, 0 minutes in 2010). Delivery times are less than 5s between a Pentalog domain and Google (with an attachment of 100k). The storing space per user is 25GB. The annual net cost is 50 euros/user. The cost for zimbra with the same features (maintenance, storing space etc.) is 108 euros / year / person. These costs are calculated for 600 people including migration, maintenance and actions taken over three years. If we leave storing aside in this cost calculation, we obtain almost the same cost.The question is more related to security. Can we trust Google’s security systems? We can also call into question the security of an outsourced service in which a great amount of information is sent. Although we are constantly investing in security, we cannot rise to Google’s level. Google has certificates on infrastructure security which allow to accept outsourcing. Despite having consulted local economic intelligence services, they don’t have a definite answer to the question. We need to pay particular attention to transfers (encrypting exchanges via https) and to the security level of passwords.In the end, why have we chosen Google? Although messaging will remain an essential tool, it will decrease in importance as our social network develops. The messaging service and calendar will be the first functions to be used, but the scope will extend to Google Docs, Google Chat and we are already looking into ways to integrate Google Apps into our information system.In order to prepare this migration, we have paid a great deal of attention to change management:* A team consisting of 25 pilot users has been created (different levels of responsibility, different locations, different needs etc.).* The satisfaction survey (carried out using GoogleDocs) sent at the end of the test period yielded positive results and enabled us to identify subjects on which we needed to provide more information. I would like to thank them for their participation. They also offered arguments for choosing this solution.* A presentation will be made in all offices to explain how this migration is going to be carried out.* We are already posting best practices into our internal knowledge base (text and videos).The migration in itself will not actually be risky. Messages have already been delivered for the two platforms (Zimbra in production and Google) in order that pilot users continue using Google Apps. The longest phase will consist in transferring the entire Zimbra content over to Google. We will therefore have no service availability breakdowns, except for the history feature during the transfer phase.Before the end of February we will therefore have changed our messaging solution. It will open new possibilities for us. Everyone will benefit from it in no time. If some of you have fears, don’t hesitate to contact me or the local infrastructure team so that we may dispel them.