I'm bumping into this sort of madness at an alarming rate. Any illusion that I once held that business is sane is rapidly diminishing. I'm going to use an example, not including names and slightly modified to include other examples of practice just to ram the point home. This covers a company working in the financial sector.
The company has its own data centres and it is rapidly growing which is creating a problem because of the time it takes to order and install new servers. The basic process is fairly straightforward, they order new servers, the servers arrive at goods in, they are then modified and mounted. The company analysed the flow (a rough diagram is provided in figure 1) and found that the bottleneck was caused by modification of the servers due to the number of changes which need to be made before mounting.
Figure 1 - The flow.
The solution? To automate the process further including investigating the possible use of robotics. It turns out that the process of modification is not only time consuming but can cause errors and the cost of automating this process (given the rise of general purpose robotic systems) can be made given the future expectation of growth. The figures seems encouraging. Getting rid of bottlenecks in a flow sounds like a good thing. It all sounds reasonable.
But it’s all barking mad. The first thing to do is map out their environment including the flow. The map is provided in figure 2.
Figure 2 - The map
Now, as we can see from the map then goods-in and ordering are fairly commodity (i.e. well defined, commonplace) and that makes sense. Computing on the other hand is treated as a “product” which seems suspect but far worse is they are making some fairly custom modifications to this “product” before mounting. Also, what is racks doing in the custom built section? Racks are pretty standard things aren’t they?
Well, not in this case. Here the company is using custom built racks which are slightly smaller than the 19 inch standard. Which is why they need to modify the servers. “Eh, What?” I hear you scream. Yes, the company needs to take the cases off the servers, drill new holes and mount rails to make them fit. “Eh, What?” I hear you scream. And so, yes, the proposed automation is about using robots to improve that process of drilling and modifying servers to fit into custom built racks. Wait, can’t we just use standard racks?
If we take our map, and mark on how things should be treated (see figure 3) then surely it makes sense to use standard racks and get rid of the modification. “But, that’s how we’ve always done it!” comes the cry. Oh, heaven help us. It may have somehow made sense in the long distance past but that's the long distant past.
Figure 3 - The modified map.
In fact if we look at the map, we should also mark compute as a utility (which is what it is now). I’ve done this in figure 4. Which now begs the question that we’re a finance house, what the hell are we messing around with servers in the first place for, let alone drilling holes in servers to make them fit custom built racks. We shouldn't be "improving" this flow, we should be getting rid of it entirely.
Figure 4 - An alternative flow
I'm writing this whilst looking at another example which has nothing to do with finance or racks but it's as equally daft in all senses except scale. This one is a lot bigger, a lot more expensive and they've even splashed out a good few million on consultants to help define their "reasonable approach". This includes lots of nice charts, flow diagrams and even a value stream map (a mechanism for improving flow). I'm starting to think that large chunks of business can only operate because competitors are doing equally daft things. Making ineffective mechanisms more efficient isn't a good thing. You really need to understand context before embarking on this type of stuff.
I need a large cup of tea, a nice biscuit and a good sit down. I'm trying desperately hard to be polite and hold back the expletives.
