Several years back, when I was a munchkin – okay not really, there was new trend called Boot from SAN. This trend continues today. I hopped on that bandwagon quickly when UCS released and the ability to use Service Profiles with Boot from SAN was a brilliant way of being able to recover server environments in a hurry, and even automatically, when a server failed. This was High Availability at its finest.
Then, I started thinking about it. In doing so, I realized that the base reasoning for Boot from SAN, in most environments these days, is not necessary anymore. The ability abstract the server OS install from the underlying hardware for HA purposes, in my opinion, is the primary reason for Boot from SAN. Now, if you know me, you will understand that I have stated for a while now that unless you are doing something weird – like using a dongle in your environment – there are very cases where you cannot virtualize all your systems. If you can virtualize (the right way), then you have high availability built in – so why do you need that twice? You don’t.
This leads to my biggest pain point with Boot from SAN, migrations from SAN to SAN. It is just plain painful and time consuming. Imagine pulling your toenails out of one foot, and having them moved to the other foot. It just does not work. This is the same as moving a Boot LUN from SAN to SAN unless you own expensive migration tools, and are willing to go through the extra effort and hassle. Without the tools, you need to:
Whereas, if you are local disk boot, you just attach the new SAN and VMotion the virtual machines. Quick, fast, easy-ish.
Now, I will admit that there are a few cases where Boot from SAN may be useful – where you are not running a hypervisor or a good one anyway –but if that is the case, perhaps you should rethink that as well. If you don’t believe me that you you can virtualize almost any, if not all, servers in your environment – reach out. I would be happy to be proven wrong – but I won’t be.