Scaling Zimbra Collaboration Suite

One of the main tasks for business is growth and development. In today's realities, an increase in the number of production facilities, as well as the emergence of new employees and contractors, implies a constant increase in the load on the enterprise's IT infrastructure. That is why, when implementing any solution, the IT manager of an enterprise must take into account such a characteristic as scalability. The ability to handle a large workload while adding large compute loads is especially critical for ISPs. Let's take a look at the scaling options Zimbra Collaboration Suite offers as a product used by various SaaS providers around the world.

Scaling Zimbra Collaboration Suite

There are two types of scalability: vertical and horizontal. In the first case, the performance gain of the solution is achieved by adding computing and other capacities to the existing IT infrastructure nodes, and in the second case, the performance gain is achieved by adding new computing nodes, which take on part of the load. Zimbra Collabration Suite supports both horizontal and vertical scaling.

Vertical scaling in the event that you decide to add computing power to your server will not be much different from migrating to a new, more powerful server with Zimbra. However, if you decide to add secondary storage for email to the server, you will definitely run into a limitation inherent in the Zimbra Open-Source Edition. The fact is that in the free version of Zimbra you cannot connect secondary volumes for storing email. To solve this issue for users of the free edition of Zimbra, the Zextras PowerStore extension is designed, which allows you to connect both physical and cloud S3 secondary storage to the server. In addition, PowerStore includes compression and deduplication algorithms in Zimbra that can improve the efficiency of storing data on existing media.

Particularly in demand is the creation of secondary volumes among ISPs, which make fast but expensive SSDs primary storage, and place secondary storage on slower but cheaper HDDs. Through the use of transparent links, which are stored on the SSD, the system continues to work quite quickly, and through compression and deduplication, each server can store much more emails. As a result, the cost efficiency of servers with secondary storage and Zextras PowerStore is significantly higher than when using the standard functionality of Zimbra OSE.

Scaling Zimbra Collaboration Suite

Horizontal scaling, by definition, can only be used in a multi-server infrastructure. Since in a multi-server installation all Zimbra modules are distributed over different machines, the administrator has the opportunity to add more and more LDAP Replica, MTA and Proxy servers, as well as mail storages almost indefinitely.

The process of adding new nodes repeats the process described in one of our previous articles about a multi-server Zimbra installation. It is enough just to install the necessary Zimbra modules on the server and specify the Master LDAP address, as well as enter the data for authentication. After that, the new nodes will become part of the Zimbra infrastructure, and the Zimbra Proxy will provide load balancing between the servers. At the same time, all previously created mailboxes and their contents remain on the storages where they were before.

Typically, new mail stores are added to the Zimbra infrastructure, at the rate of one server per 2500 active users of the Zimbra web client and up to 5-6 thousand users of desktop and mobile email clients. This number of users allows you to achieve the most responsive server experience and avoid problems with availability and long load times.

In addition, multi-server infrastructure administrators can also connect secondary stores, as well as compression and deduplication on each email store using Zextras PowerStore. Using this zimlet allows you to save up to 50% of disk space, and, coupled with an increase in the economic efficiency of the entire infrastructure. In the case of large ISPs, the economic effect of such infrastructure optimization can be really large.

Source: habr.com

Add a comment