System Performance : Performance and Scalability



Connect Xf has been designed to deliver more performance for the same resource utilization, when compared to traditional systems Our engineers worked hard to design a system, which delivers more performance and scale for the same resource utilization, when compared to any other collaboration system. Our product incorporates our vision of being able to do more with less.


Adaptive Mail routing

Adaptive Mail routing

Speed of mail delivery is crucial for effective collaboration. To ensure that there are no bottlenecks within the system, the Connect Xf engineers designed multiple ways to route mail such that each mail has the shortest path to the destination, being extremely stingy with hops.

In addition it is easily possible to configure parallel paths for mail flow by adding queues to ensure that the queues never pile up due to a problem with one of the recipient system. With the introduction of such innovate features over the OSS MTAs, we were able to get more than a 300% jump in mail delivery times with a high level of consistency.


SISA (Single Instance Storage of Attachments)

SISA (Single Instance Storage of Attachments)

Besides the server capacity, one of the biggest enemies of high performance is choking within the network leading upto the server or in the last mile to the client. The network capacities can be easily consumed when large mail are transported between servers or downloaded by clients. We realized that one big problem causing the network bottle-necking was mail with large attachments sent to multiple people (groups or distribution lists) within the network, and then each client downloading those large mail to their clients. By stripping mail of large attachments and storing them at a central place, we observed a reduction in 60 % on the network utilization. Not to mention the savings in storage due to retaining only one copy of the original large mail (Application level de-duplication).

To enhance performance for large sets of users having a large mail flow (10,000+ users) the attachments are stored in sub folders which saves time in saving or reading attachment from the folders as the size of each folder is small.


Aggressive Caching

Aggressive Caching

The Connect Xf engineers identified another point, where even excessive use of high performance components like LDAP and Database (which have inbuilt caches) can degrade performance substantially. While letting the application developers have the flexibility to write code as they want, they built in a strong caching layer in between, which holds frequently accessed data from different parts of the system in the memory which reduces the calls to the database, LDAP and mailstore.


Enhanced end user usability leading to higher system performance

Enhanced end user usability leading to higher system performance

Right from bringing out frequently used applications and options in easily usable interface options, to intelligently searching only the minimum data to get the required results, to painstakingly designing the system to reduce clicks or eliminate them during any user operation, Baya, the collaboration web client within Connect Xf, has been specifically designed to reduce the operating path for the users, to allow them to access and view information with minimal clicks. This in turn has an impact on server performance since the server now has so much less to do.


Scale only on demand and do it with ease

Scale only on demand and do it with ease

The Simple High Availability architecture and the ability to setup distributed site servers, makes it very easy to scale the system on demand. As you grow vertically (more users at one site), you can simply add or consolidate servers within the server farm to achieve more capacity and as you grow horizontally (more sites), you can easily add servers the enterprise tree to allow users to be located remotely.


Minimize data movement across servers

Minimize data movement across servers

Connect Xf has been designed to optimize the movement of data across enterprise servers to reduce delays in data replication and also to optimize server and network performance. The simple philosophy is that if you need it frequently replicate else access from where it is. e.g. your mailbox is never replicated across distributed site servers or even on the server farm at a single site, thus if you temporarily access your mailbox from any site (other than your own), the system will connect remotely and show you the mailbox.


Monitoring out of Tune components

Monitoring out of Tune components

Typically during deployment the system is tuned to the specified work load. However as the system is used over time, the work load may vary (up or down) and it may be necessary re-tune some of the components and sub-systems to work optimally with the new loading patterns. You don’t have to worry about this since the system automatically monitors the usage and sends alerts when a re-tuning is required.


Nip DOS attacks and system overuse in the bud

Nip DOS attacks and system overuse in the bud

Connect Xf has been designed to optimize the movement of data across enterprise servers to reduce delays in data replication and also to optimize server and network performance. The simple philosophy is that if you need it frequently replicate else access from where it is. e.g. your mailbox is never replicated across distributed site servers or even on the server farm at a single site, thus if you temporarily access your mailbox from any site (other than your own), the system will connect remotely and show you the mailbox.