Traditionally, server load balancing has been viewed as a problem for large enterprises and a costly and unnecessary expense for SMEs – and it was certainly not an issue for the comms team to worry about. But that was when most SMEs ran their businesses using just a couple of servers. Today, the number of servers used by SMEs can often be 10 or more, while the need for load balancing is also being further fuelled by developments at Microsoft.
Many SMEs moving to MS Exchange 2010 to handle increasing volumes of email are realising the need for load balancing to ensure performance and high availability. Microsoft has supported this by choosing to adopt a distributed, role based architecture. The same is true for those deploying Microsoft SharePoint and Lync, which presents some unique challenges in the area of load balancing and managing available bandwidth.
What is load balancing?
The basic idea with load balancing is to share incoming connections across multiple hardware devices. For example, with back-end applications such as order processing, billing and customer management being integrated into complete supply-chain, web-enabled applications, website reliability, scalability and performance is essential.
To deal with this, server load balancers or Application Delivery Controllers (ADCs), provide the ability to direct traffic to the best performing, most accessible servers based on factors such as concurrent connections and CPU/memory utilisation. If a server or application fails, the user is automatically re-routed to another functioning server.
In the case of Exchange 2010, changes Microsoft has made to its core server architecture including the use of Exchange Client Access Server (CAS) to handle client connections, make load balancing necessary to automatically re-route and reconnect users to optimised servers to avoid poor performance and deliver high availability. Server load balancing also helps to optimise the performance and resilience of Lync; and unlike Exchange, Lync has to support real-time VOIP traffic flows that are jitter- and latency-sensitive.
Microsoft Lync Server 2010 supports two load balancing solutions. DNS (Domain Name System) load balancing or so-called ‘round robin’ balancing is a basic software solution that works by responding to DNS requests by simply alternating these requests between servers, without consideration for matching the user IP address and its geographical location, server load or network congestion, for example. This is used to balance specific Lync network traffic such as SIP and media.
However, hardware load balancing is also recommended for enterprise scale, high-available deployments where Lync Server requires multiple servers in Front End pools, Director pools and /or Edge server pools to withstand a server outage. It is used to balance the web services/http traffic such as address book, meet/dial in sites and meeting content across servers to ensure performance and provide scalability and availability.
By adding hardware load balancing, traffic can be shared across Lync servers and if a server becomes inaccessible, the balancer will take it off-line and automatically re-route and reconnect users to other functioning servers to guarantee application uptime.
Another function performed by the load balancer is SSL Offloading. The encrypted SSL session can be terminated at the load balancer so that the headers and content can be read in order to direct it to the correct servers. The real benefit for this is when traffic is coming from the same IP address such as a reverse proxy and persistence – also known as stickiness – will ensure even load distribution across servers. Re-encryption is required in Lync as it does not support SSL offloading.
So, hardware load balancing has a role to play in Lync deployments that require high availability. But in fact the term hardware load balancing is slightly misleading as it doesn’t actually have to be hardware at all. A number of virtual/software load balancers for Hyper V or VMware environments have also completed Microsoft qualification testing with Lync Server 2010.
While server load balancing has traditionally been a big ticket item from the likes of vendors such as F5, there is now a growing demand for affordable and scaleable load balancers. Virtual load balancing solutions start at just over £1,000.
For the comms channel, load balancing offers another value add opportunity that goes hand-in-hand with end user migrations to Lync, MS Exchange or SharePoint, or to a as part of a new server deployment. The noise around Lync is certainly growing and while there may still be more hype than real deployments, there is no doubt that it is only a matter of time. And when it does, having a well-balanced solution will help to ensure Lync delivers on its promises.