Stop Letting Traffic Spikes Kill Your Business

It’s 3:00 AM. Your monitoring system is screaming. The marketing team’s email blast just went out, or maybe you finally hit the front page of a major news site. Your single Apache server is thrashing, the load average is sitting at 25.0, and your swap usage is climbing. By the time you SSH in, it’s too late. The connection times out.

If this sounds familiar, you are doing it wrong.

In the old days, the solution was throwing money at the problem: buying a massive F5 Big-IP hardware load balancer or upgrading to a monster dedicated server. But in 2011, that’s just burning cash. The smart money is on software load balancing. Specifically, HAProxy.

Why HAProxy? (And Why Nginx Isn't Enough Yet)

While Nginx is fantastic for serving static assets, HAProxy 1.4 is the undisputed king of TCP/HTTP load balancing. It is purely an event-driven, non-blocking engine. I have seen a single CoolVDS instance with 512MB RAM handle thousands of concurrent connections without breaking a sweat. It strips the SSL overhead (if configured correctly) and balances traffic based on actual server health, not just round-robin guessing.

Pro Tip: Don't rely on DNS round-robin. It ignores server health. If one web node dies, DNS will still send 50% of your traffic into a black hole until the TTL expires. HAProxy detects the failure in milliseconds.

The Setup: CentOS 6 & HAProxy 1.4

Let’s assume you have two backend web servers (Web01, Web02) and you want to put a Load Balancer (LB01) in front. At CoolVDS, we recommend CentOS 6 for its stability, though Debian Squeeze is fine if you prefer apt.

First, install the package:

[root@lb01 ~]# yum install haproxy
[root@lb01 ~]# chkconfig haproxy on

Now, let’s look at a battle-tested /etc/haproxy/haproxy.cfg configuration. This isn't the default garbage config; this is tuned for high-throughput environments.

global
    log 127.0.0.1   local0
    maxconn 4096
    user haproxy
    group haproxy
    daemon

defaults
    log     global
    mode    http
    option  httplog
    option  dontlognull
    retries 3
    option  redispatch
    maxconn 2000
    contimeout 5000
    clitimeout 50000
    srvtimeout 50000

frontend http_front
    bind *:80
    default_backend web_servers

backend web_servers
    mode http
    balance roundrobin
    option httpchk HEAD / HTTP/1.1\r\nHost:localhost
    server web01 10.0.0.2:80 check
    server web02 10.0.0.3:80 check

The Critical Details

• option httpchk: This is vital. HAProxy won't just check if port 80 is open; it will actually request the root page. If your Apache service is up but the database connection is dead (returning a 500 error), HAProxy can be configured to pull that node out of rotation.
• balance roundrobin: Good for stateless apps. If you need session persistence (e.g., for PHP sessions), stick with source balancing or inject a cookie.

Infrastructure & Compliance: The Norwegian Context

Here is where the physical layer matters. You can have the best config in the world, but if your underlying VPS has "noisy neighbors" stealing your CPU cycles, your load balancer will introduce latency.

At CoolVDS, we use KVM virtualization. Unlike OpenVZ (common among budget hosts), KVM ensures that the RAM and CPU resources allocated to your load balancer are actually yours. This stability is crucial when you are pushing packets at wire speed.

Data Sovereignty Matters

Hosting outside of Norway is becoming a headache. Between the Patriot Act in the US and the strict enforcement of the Personal Data Act (Personopplysningsloven) by Datatilsynet here, you need to know where your bits live. Running your load balancer and backends in our Oslo datacenter ensures your customer data never crosses borders unnecessarily. Plus, the latency to the NIX (Norwegian Internet Exchange) is practically non-existent.

High I/O for Logs and Caching

Load balancers generate massive logs. If you are logging every HTTP request for analytics, standard mechanical hard drives will choke. While we are seeing the dawn of SSDs entering the enterprise space, CoolVDS provides high-performance enterprise storage arrays (RAID 10 SAS/SSD hybrids) that ensure your disk I/O never becomes the bottleneck.

Final Thoughts

Redundancy isn't a luxury; it's a requirement. A single VPS costs the price of a few coffees. Losing a customer because your site wouldn't load costs much more.

Don't wait for the crash. Spin up a fresh CentOS instance on CoolVDS today, install HAProxy, and sleep better tonight knowing your infrastructure can take a punch.