Pure Storage presented at Tech Field Day Extra at VMworld 2020 on how you can connect an external FlashArray to VMware’s Cloud Foundation (VCF) architecture to get a more flexible storage layer for your infrastructure.
This highlighted, once again, the major challenge that enterprises have when trying to operate their infrastructure in a cloud-like way: you have to invest in making it easy to operate.
It seems obvious to me, but a lot of the industry still decries the idea that a lot of the cost of IT is due to operations. Well, yeah, because the value of IT is from using it, not building it, and using it is called… operations.
And operating infrastructure is still pretty tricky.
In my consulting for PivotNine I often talk to clients about preserving optionality which is a fancy way of saying keep your options open.
Changing your mind costs time and money. If you want to be able to react quickly to change, you need to reduce the amount of time and money it costs to change your mind. This is the main reason cloud is so attractive: changing your mind is usually pretty easy.
Keeping your options open costs money (it’s part of why AWS is so profitable) but it can be worth it if it keeps your switching costs low and you have a reasonable expectation you’ll need to change your mind.
Think about your own experience of IT in the past few years. How often have you needed to adapt to new circumstances? How often did you need to make those changes in a hurry?
Still keen to lock yourself into a decision you can’t easily change?
Make Change Easy
By adding a Pure Storage FlashArray into the VMware Cloud Foundation mix, it becomes much easier to move data around, both within a cluster and also across clusters.
Instead of needing to copy data within the cluster or across workload domains, the FlashArray can take a snapshot of a VM and then move the connection to the data, rather than moving the data itself. This is much faster, particularly for large amounts of data.
It’s not a new idea, and is the same reason snapshots are faster than streaming backup data to a physical tape drive or into a remote device over a 1Gb/s network pipe. Well-defined abstractions and loosely-coupled modules are really useful things.
Keeping everything inside VCF means the storage abstraction is within the VMware infrastructure, so it’s more closely coupled to the failure domain of the cluster. This might be fine, if you’re confident that you won’t need to move things between clusters very often. But if you think you might need things to change, people are always happy to discover that changes are fast and easy.
Data migrations are some of the most tedious and soul-crushing projects I’ve ever been on, and people are rarely happy to learn that they’ll need to take an outage and that it’ll take hours (or days!) to complete the migration. Oh, and that work needs to happen out-of-hours, so enjoy doing your tedious data tasks at 2am instead of sleeping.
Invest in Flexibility
If we’ve learned nothing else over the past year, hopefully we’ve absorbed that being able to react quickly to change is very valuable, and often the true value of that agility isn’t easy to predict. Flexibility requires investing in things that are able to change, and that can require a bit more time and money up-front.
So as you’re designing your clusters, think about how things will be operated day-to-day, because that’s what your clusters will (hopefully) spend at least 80% of their time doing before they’re replaced.
If you do this right, the job of building new clusters to replace old ones will become operations. Changes and improvements should simply be routine, and ideally no one but you will even notice it happening.