How Many Servers Can One Admin Manage?
There’s a discussion today at Slashdot about how many servers or users an admin can manage. The question seeks to establish whether a department in which sysadmins each manage 900 user machines is understaffed. As in any discussion of IT or data centers, practices vary widely. But here are a couple of relevant factoids from our travels tracking the data center industry.
Jeff Rothschild, the vice president of technology at Facebook, said in a recent presentation that Facebook has 230 engineers supporting data for more than 300 million users. He says Facebook seeks to maintain a ratio of one engineer for 1 million or more users. Facebook is vague about exactly how many servers it has, saying it’s “more than 30,000.” But 30,000 servers and 230 engineers works out to a ratio of about 130 servers per admin.
Microsoft says it has automated its data center operations to the point where its admins can each manage between 1,000 and 2,000 servers. That matters, as the company may pack more than 300,000 servers into its new container data center in Chicago. It expects to support that facility with about 30 employees, including admins and facility maintenance staff.
At Tagged, we saw server count grow quickly from ~300 to ~1000. We needed more IT staff, but found this demand driven by the increasing number of applications, not the number of servers. Today we index IT staffing to the number of software engineers.
Anders GregersenPosted December 31st, 2009
I think the size of the IT staff depends on more than the number of servers, applications or software engineers (we all know that). Our IT department consist of both operations, 1st line support, change management, software engineering, project governance and knowledge management. Depending on who you ask we have a size of 2 to 15 people. Management sees us as 15 people, but if we talk server operation, we are 2. We are supporting every backend component (network, servers, storage, security, e.g.) including participating in all projects that impact our infrastructure.
As a companys size go beyond a certain threshold, standards emerge on how and what are required, reducing the required staffing and maturing the business use of IT. Most companies need the same functionality and in most only the scale is different (yes its simplified).
Daniel MatthisPosted December 31st, 2009
One problem with the “how many servers can one admin handle” really depends on how the server is managed, what type of server it is, and the it staff structure around those systems. In the SMB world many sysadmin’s double as the appadmin in which case the number depends on the number of apps and servers not just servers.
Other place like Microsoft probably have sysadmins who take care of just the OS and relegate app duties to an analyst some place. Thus they have a couple of standard OS to support a huge cluster of servers. Give an admin one app, one OS configuration to support and automation can allow them to handle hundreds if not thousands. The moment he takes on an app and multiple OS configurations that number drops exponentially.
at my former company 2 of use managed 160+ servers. 40 of which were standalone & 120+ were virtual. Plus we managed the SANs.
Depends how resource/fail intensive the job is. As we were solely handling internal operations for the company our load was much lower than if we were handling Internet operations.
For the most part 98% of our system was low maintenance, there were a few applications which constantly hiccup’d. We only were strained when a major system needed to be rebuilt over several days
ChunenPosted January 1st, 2010
Shouldn’t the question be:
“What is the ratio of Sysadmins-to-UniqueServerConfigurations?”
For example, you could have a front end web-server farm with 100 identical servers, a second level with 25 identical application servers, and a third level with 10 identical databases. This may require 3 admins for 3 unique server configs to a ratio of 1:1.
Compare this with 135 unique configurations managed by x admins.
The question could be further refined the operations organization was divided into separate backup, resource monitoring and integrity monitoring groups.
AustinPosted January 1st, 2010
Our ratio is 1 per 300 servers. Or about 1 per 1 million accounts. We use a lot of virtualization.
robertoPosted March 4th, 2010
“Our ratio is 1 per 300 servers”.
Not sure how that is possible. The person is either superman or your servers are a mess.
rmxzPosted March 16th, 2010
Totally the wrong metric.
Managing 700 servers or 7000 servers is about the same work so long as they’re identically configured.
Wouldn’t a better metric be “how many sysadmins per distinct server configuration”.
I’ve seen the # alot lower in environments where poor standards were implemented. More like 1:20 server ratio. The problem is nobody has the vision to automate tasks in the beginning where they could support a 1:1000 servers.
PhilPosted June 21st, 2010
At my company we manage around 800 dedicated servers and 200 virtual servers, since it’s a web host about 80% of those are unique/standalone/require you to remember somewhat what makes them different. We have a lot of standardization in place which makes life easier. But there is typically only one man working on each shift handling all technical aspects of the business (support/alert response/full management (i.e. no matter what happens to the site/server it’s our problem and you need to know how to fix it). During busier shifts we have 2 men working.
[...] how much of a sysadmin’s time will be spent managing these servers. Data from multiple sources shows that a sysadmin can maintain at least 100 servers, particularly if they are homogeneous as [...]
WSPosted September 6th, 2011
As others have hit upon, the number of servers is irrelevant – it is the configuration, the purpose, etc. Obviously, one server may do only one purpose, e.g. DNS, that is simple to deal with and which also might be automated. But if you work in a shop with complex and diverse configurations, and no budget for automation tools, then the amount of support per server will increase. Are the servers virtual or physical? That’s another variable. Internal users or external? etc. In other words, “number of servers per admin” is a meaningless number – can’t be applied across different environments. And that doesn’t even touch on who is a “server admin”.
BBPosted March 22nd, 2012
The question is straight forward but answering it looking from different views and aspects of IT would be complex. In my opinion and experience in IT, I can say that the ratio goes down as the IT organization maturity improves and has implemented automation across Data Center Infrastructure and Operations. As the number of servers in the Data Center grows – wither physically or virtually the demand for automation becomes higher. Budget availability then comes in and the ratio worsen if automation initiatives are ignored.
seshu kumar.gPosted August 24th, 2012
so finally all the admins are decide who may member are maintain , but one think is write each one person will maintain their stamina. if other wise they will face the problems in future, write ,…. so Facebook no has 30,000 servers will continuously use all admin’s .. is it possible…
As mentioned, it seems to do more with the software systems being managed, not the actual server count. We have server admins on our team that support over 200-1k servers(vps machines) with ease, even with modest automation and deployment solutions.
The more physical servers you manage within a network, the more maintenance there is going to be all-around. Unfortunately there’s not too much you can do to automate this level of things unless you’re putting a lot of capital into it. Either way though, physical or virtualized, very large networks are indeed easily manageable when approached correctly.
Plexxi and Boundary team up to deliver a model for the application-aware network — Tech News and AnalysisPosted February 4th, 2013
[...] to Amazon, Amazon must figure out how to economically scale its business — and having the old industry standard of one systems administrator managing 500 or maybe 1,000 servers isn’t going to let AWS keep [...]