How many servers should a system administrator manage?
You could answer that question with the hoary consultants’ response: How long is a piece of string? But it’s a real question many IT mangers and CIOs have to contemplate when working out best practice for the staff to server ratio, especially when evaluating systems and processes in lieu of an infrastructure change.
So it deserves more than the flippancy sometimes inherent in the string analogy.
Computerworld Australia was recently asked by one of our Advisory Panel members to canvass the views of other IT managers and analysts on this question and the responses were reflective of a dearth of documented guidance and consensus.
In most cases the initial reaction to the question was “it depends” or “it's impossible to put a number to this without a lot more detail”.
And that’s fair enough – it is very difficult to fashion guidelines that apply to every IT department when there are so many diverse factors to take into consideration; kind of server, application type(s), ability and role of staff, business requirements, budgets, etc, etc, etc.
But that said, if you dig a bit more there are some indicative examples – or perhaps anecdotal evidence – of possible scenarios that may help IT managers formulate a ratio range to use when planning server set up changes or pitching strategies. Things will always need to be tailored to individual environments but having a starting point to work from can be highly beneficial.
“We have observed that it can be, for example with a physical server, as low as 10 per admin, and for virtual servers as many as 500,” Gartner analyst, Errol Rasit, said. “But it really depends on the type of application. We have seen as an example from a particular customer – from some of our larger customers – that they had their admins managing 15 physical servers and when that moves to virtualisation it moves to something like 75 virtual servers.
“To give you a different order of magnitude in another example one admin was looking at 50 physical servers and then moving to 250 virtual servers. I will say that we have seen maybe 500 or 600 virtual servers being managed by a single admin.”
IDC meanwhile notes that in Australia the ratio for an SMB would vary greatly from a hoster and again to a cloud provider like Amazon or Microsoft. The analyst house’s statistics suggest anywhere from 10,000:1 at a dominant vendor like Google down to the SMB average of 30:1 for physical boxes and 80:1 for virtual machines.
One enterprise IT manager told us the ratio for physical servers was roughly 50:1, another working for a government organisation said 15-20:1, and an IT director at a research and development outfit noted that in a mid-size organisation a system administrator could maintain 10-14 servers per week or if their role was merely maintenance (i.e. no projects, no debugging, etc) then they could look after 25-35 servers per week. The IT director added a bigger organisation with larger economies of scale could potentially increase the ration to 10-14 servers to each admin per day with staff dedicated to just maintenance.
One of the key factors in increasing the ratio, however, is how much automation can be rolled into the maintenance / management of the server farm.
“A lot of what changes the ratio in the physical world is the types of tools being used to automate a lot of the processes; so run book automation and these sorts of things,” Gartner’s Rasit said. “That tends to be the main differentiator. The problem with virtualisation and virtualisation tools is there are a lot of them. It is very, very easy for a lot of customers to try and automate everything and that doesn’t necessarily always bear fruit for the organisation because they are spending too much time doing that.
“They don’t know what is actually happening because everything is automated. We have some survey data on what administrators spend most of their time and around half of an admins’ time is spent on deployment and incident management, or resolution. If that is the case then it makes sense to automate that as much as possible.
“In a virtual world, if a VM goes down you kick off the VM again – automating that would save a lot of time. And then it is up to the admin to really select strategically what technologies help them automate their day to day tasks and actually save them time rather than having everything being automated and not having a great idea of what is happening.”
So is there any one answer to our question on how many servers should a system administrator manage? Looking at the above range of responses the answer is clearly no. But by at least looking at what others are doing and clearly noting the scarcity of best practice in this regards, IT managers and CIOs should be able to better formulate a server / system administrator strategy.