In the last period, many times I was involved in discussions with customers regarding the adoption of the out-of-the-box Windows Server NLB (Network Load Balancer) or dedicated hardware solution (Cisco, F5, Coyote, etc.)
I’ve tried to summarize my mental pattern about this argument, in the table below (…well, I censored Duff & Donuts from my thoughts):
| | PROS | CONS |
| NLB | - It’s cheaper (already available as part of Windows Server stack).
- Rapid deployment and adoption:
- SharePoint team doesn’t need to rely on the infrastructure team for the configuration;
- No real technical expertise needed;
| - NLB works at socket level (TCP/UDP) and doesn’t provide any specific feature or optimization for http/https.
- A dedicated NIC (network card) is strongly recommended.
- Governance and Operations of NLB cluster could be tricky:
- more people must be made aware of NLB configurations
- Configuration could be tricky in presence of multicast traffic.
- No caching capability is provided:
- for http/https is expected to rely on Microsoft ISA or IAG;
- No certificate management:
- Certificate must be individually managed in IIS;
- Some Governance is needed;
- No compression capability:
- for http/https is expected to rely on IIS 7.x
- Technology is …antique, well not really an issue but NLB was created to balance COM+ application with NT4/OptionPack
|
| Hardware Load Balancer | - Improved Performance, https traffic is managed at hardware level;
- Low latency during the switching in case of High-Availability configuration.
- SSL and HTTPS configuration is managed internally, making it transparent to IIS/SharePoint configuration.
- Caching capabilities if needed (don’t abuse of this).
- High-Availability generally supports dependency rules on how to route packets in case of unavailability of specific servers/application tiers.
- Support of protocol specific rules (http, https, etc.)
- Support of Security Rules;
- Technical Agnosticism, the tools can be used to balance Windows, Linux, Web Server, sockets, email servers.
- Governance in the sense that there is a centralized point of management for all the needs regarding balancing, high-availability, security etc.
| - Expensive, for sure it something to be acquired and identifying the best solution won’t be easy ‘cause the huge amount of options in the market.
- Learning Curve
|
Rules of Thumb
- Use NLB if hardware load balancer is not available and there are no plans on that (polite way to say no budget);
- In the Intranet, if reverse proxy isn’t available (sometimes the hardware load balancer is available only for Internet traffic);
- As Tactical Solution (as example for running stress test on your new project in a stage environment if the hardware solution is not available or cannot be used);
- We can definitely state that a strategic solution must rely on hardware load balancer, a tactical solution could rely on a software NLB.
