How to mitigate Hardware load balancer's risk

Hi, If we use hardware load balancer in front of clustered servers to take care of fail overs and load balancing in round robin manner, what happens if the hardware goes bad? How should we take care of this? Should we put one more load balancer in stand by mode to mitigate the problem?

Hello Rajan,

Personnally, i'm planing to do like the cade guide exemple, in the deployment diagram, just a box for the load balancer and another box for the backup load balancer.
If you can take a look at the litle image here, you can see that they use "Active/Hot StandBy".


If you permit it, since is the same subject, in the case of physically separate tiers, web servers and app servers, having a hardware load balaner for the web servers and another for the app server(plus two backup load balancer for the both), didn't this seem exaggerated ?

I hope this help.

True, it seems to be exaggerated with 2 balancers at each tier - that's why I posted this question. May be others can share their experience too. The only good part about the assignment is that we don't need to worry about the cost factor of the hardware, do we?

Rajan Choudhary wrote:True, it seems to be exaggerated with 2 balancers at each tier - that's why I posted this question. May be others can share their experience too. The only good part about the assignment is that we don't need to worry about the cost factor of the hardware, do we?

I reckon between web server and application servers you can use software load balancer. For example IBM HTTP server has a load balancer plugin that can be installed in HTTP server to do load balancing on App server cluster.

true, s/w load balancers can definitely be used in web servers but performance wise there is no comparison between s/w and h/w load balancers. H/W ones are much faster and reliable. That's what we pay for.

Kumar Amit:
I reckon between web server and application servers you can use software load balancer. For example IBM HTTP server has a load balancer plugin that can be installed in HTTP server to do load balancing on App server cluster.

I feel a confusion when we spoke about Web Server, somtimes we mean by it a servlet container, i'm not sure if IBM HTTP Server is a jsp/servlet container :s !

About the overkill of 4 hardware load balancers(2 active and 2 backup, at the servlet containers and at the EJB container), i'm feeling the same. I was searching for a h/w balancer with a low price, i found one at 2000$ approximatly, (it's funny to note that the 1 August mandatory course is more expensive), do you think even with that, it's still unreasonable ?

Rajan Choudhary wrote:Hi, If we use hardware load balancer in front of clustered servers to take care of fail overs and load balancing in round robin manner, what happens if the hardware goes bad? How should we take care of this? Should we put one more load balancer in stand by mode to mitigate the problem?

Rajan, did you end up showing a standby server to the hardware load balancer too ?
Since suD requires uptime of 99.99% for few hours, a failure to hardware load balancer will result in failure for this NFR.

no, not the stand by server but an additional h\w load balancer.