Q: What's the typical setup for managed load balancing setup for two servers? Can I put one server in Los Angeles, and the other in Phoenix, and load balance between both?


A typical application is load balancing a website across two or more servers. In the simplest scenario, the website will be assigned a "public" IP address on the F5 BigIP load balancers. The servers will be connected to a dedicated back-end network on the load balancers and will have their own private IP addresses. This is the "server pool".

Website traffic will be distributed across the server pool according to a selected method, including directing traffic to the least-loaded servers or by using persistence. Persistence keeps a user's web traffic going to the same server throughout a web session. This is particularly useful for e-commerce transactions. If a server goes offline (e.g. for maintenance), the load balancers detect the failure and immediately redirect traffic to other servers in the pool.

Multi-site load balancing (e.g. Los Angeles / Phoenix) is used as a high-availability and/or business continuity strategy. For this application, we use F5's Global Traffic Manager (GTM) technology. The GTMs use a combination of DNS and deep integration with the F5 BigIP load balancers to provide a number of service delivery options, including geographic optimization of content delivery and disaster recovery. A typical deployment would be two load balanced server pools in geographically diverse data centers, controlled by two or more GTMs.


Q. Why does some of my internal traffic show up in my data center's internet bandwidth usage? Shouldn't internal communications be separate from my total usage?


Great question, Jim. American Internet Services (AIS) bandwidth measurements are based upon Simple Network Management Protocol (SNMP), the industry standard for measuring bandwidth among Data Center and Internet Service Providers. Simply put, SNMP counts the number of packets received and transmitted on a network interface. Bandwidth usage is calculated by recording the SNMP measurements; over time. - in our case, once every five minutes on a client's network interface. There is no way for SNMP to distinguish between internal traffic and external traffic, so any received traffic counts towards total bandwidth usage. If a data center client directs traffic to the network interface, the router has to look at it -even if it's not supposed to do anything with it. In the vast majority of cases this is not an issue, but if a client has a misconfigured network they may send a significant amount of unintended traffic to the network interface. This can result in bandwidth overage charges, fees that can be easily avoided by fixing the source of misdirected or needless traffic.
Here are two solutions for the most common network misconfigurations:

A) Configure internal servers to use non-routable network addresses for server-to-server traffic (e.g. backups), preventing any local traffic from being directed to the upstream (public) network interface

B) Place a router or firewall in front of the client network to isolate internal traffic from the upstream network interface

While a data center provider has no direct control over its clients' internal network configuration, AIS offers several engineering support options to help clients set up and troubleshoot their networks: AIS network engineers are available on a consultation basis for most network configurations. Alternatively, we can provide contact information for third party vendors who can set-up and manage our clients' networks.

Thanks for the question Jim, hope this helps!