Look, let’s make this very clear, we love F5, we think they are awesome, the best ADC in the market. But please don’t be surprised when we charge you a fairly large amount for migration consultancy…
f5-load-balancer

Each F5 migration presents its own challenges, and you shouldn’t do this kind of thing without a lot of planning. This is why we offer a paid-for pre-migration health check, and depending on the size and complexity of the existing deployment it can take several days to come up with a plan.

Unlike other vendors, our recommendation might not be what you expected:

  • We might suggest that you reconfigure the existing F5 deployment, simplify it, renew the support contract and not replace it with Loadbalancer.org.
  • We might suggest that save money by just cancelling the security module contract and replacing it with our WAF.
  • We might recommend that you use the F5 for 20 external services, and re-deploy your 60 internal applications across 6 * Loadbalancer.org appliances.
  • We might suggest that you use a cloud based WAF or CDN in front of our load balancers instead of our built-in WAF.
  • Or we might recommend that you just move the current F5 deployment straight on to a pair of our ADCs. i.e. migrate it exactly as it is.

So what does any of this have to do with Jenga?

Well the other day I was helping one of our favorite partners, who in turn were trying to help a customer; they were planning to replace a pair of hardware F5 units with our Loadbalancer.org virtual appliances (for the obvious cost benefits). The customer had also requested that the infrastructure was tidied up, the security tightened, everything documented and training supplied.

Now that all sounded fine to me, even when I was given the existing 3,000 line configuration, I thought that’s OK - I could see some immediate gains in reducing the number of clusters, and simplifying some of the iRules... But when I started asking the customer if they agreed with my initial diagnosis, I got some blank looks. Slowly the story emerged that over the last decade the customer had used an assortment of external consultants to add or modify applications to the existing F5 configuration, and now it was so complex that no one actually knew what 60 of the 135 clusters were actually doing. The entire application stack had become a giant game of Jenga, where one false move could bring the whole thing crashing down!

So how did the story end?

Well, It took me 5 days longer than expected, but the customer is now very happy having ditched 60 phantom clusters, and splitting the applications across 2 separate pairs of Loadbalancer.org appliances, with approximately 40 services on each one.

Why did they split them across two pairs of load balancers?

Because they actually had two different application teams (from a prior merger), so it made a lot of sense to logically & physically separate the applications that they were each responsible for.

I’m glad to report that now when either team makes a change, they can do so with confidence that the whole thing won’t come crashing down around them. Long term it will save them a great deal of pain…

And no more Jenga consultancy fee!