Admins-per-server ratio - observations from own experience
I have been asking myself this question a long time ago - what “N” number of admins are needed to service M number of servers? Or how many support service employees are needed for a certain number of workstations. As it turned out, there is no unequivocal answer to this question. Let me cut the article title to the APS-rating. The indicator depends on:
Nevertheless, general approaches to the definition of this relationship are indicated in the following sources:
If we make some generalizations (for example, removing the data center values that are clearly non-standard for the average Russian business, with the ratio 1 / 300-1 / 1000), then the picture looks like this (for comparison, it also cites the values for its activities):
David Capucci (2) proposed the following division of users into groups:
To determine which group the users of my organization belong to, I used this particular model, with the prevalence of group B. In this case, the scope of the APS rating is shifted from 1/80 to 1/120. I have 1/99.
In addition, using the data of the application system, I was able to analyze the applications of users for 2010 and the 1st half of 2011.
Total applications received:
- for the 1st half of 2010 - 1118 (of which 781 change requests and 337 failures)
')
- for the 1st half of 2011 - 1282 (of them, 843 requests for change and 439 failures)
those. The total increase in the number of applications was 14.7% compared to the previous year. For the subsequent analysis, it is important to know that we operate with two temporary values characterizing each application — the time of actual closure (ie, from the moment when the user opened the requests to the moment when the IT officer closed the application) and the time spent by the IT officer to fulfill this application. . These time intervals may vary by times. This situation (the difference between the actual execution time and the labor costs) arises in the following cases:
It should also be noted that in the 1st half of 2011, the number of IT employees was 9 (of which 4 were directly involved in supporting) people, in the 1st half of 2010 there were 8 of us (3 support employees). Accordingly, with the old number (8), for the six months of 2010 we managed to reach the average actual closing time of 1 application in 79.4 minutes. In 2011, with the number already in 9 people, the actual average time for closing an application was already 52.6 minutes. In fact, we improved our performance by 33.7% compared to last year, having increased our staff by 12.5%. Interestingly, the IT worker’s labor costs per request decreased by about 1.3 minutes (ie, they improved by 2.5% compared to last year).
What conclusions can be drawn? For example, it seems to me unequivocal that 3 IT staff to support 395 users is clearly not enough (this is evidenced by at least a 33% improvement in the time spent on an application after increasing the number of support staff to 4).
About the company-consumer of services: wholesale and retail sale of electronics, system integrator with a developed branch network
- server homogeneity
- nature of tasks
- operating system
- application software
- and even “advanced” users
Nevertheless, general approaches to the definition of this relationship are indicated in the following sources:
- How Many Administrators are Enough?
- Know The Types: Sizing up Support Staffs
- Improve staffing ratios
- Is there a best practice for a server to system administrator ratio?
If we make some generalizations (for example, removing the data center values that are clearly non-standard for the average Russian business, with the ratio 1 / 300-1 / 1000), then the picture looks like this (for comparison, it also cites the values for its activities):
- the ratio of the number of administrators to the number of servers is from 1/10 to 1/20 (1/25)
David Capucci (2) proposed the following division of users into groups:
- Group A - users at the forefront of progress (traders, developers and programmers) require (or want) the most new and fast equipment. APS-rating - 1/30!
- Group B - users with a common set of software on computers, use the tools of group work and analytics. however, stopping the operation of these users is tantamount to stopping the business. APS rating from 1/60 to 1/100
- Group C - users with a standard set of software, without a critical impact on the business. APS-rating - 1/125
To determine which group the users of my organization belong to, I used this particular model, with the prevalence of group B. In this case, the scope of the APS rating is shifted from 1/80 to 1/120. I have 1/99.
In addition, using the data of the application system, I was able to analyze the applications of users for 2010 and the 1st half of 2011.
Total applications received:
- for the 1st half of 2010 - 1118 (of which 781 change requests and 337 failures)
')
- for the 1st half of 2011 - 1282 (of them, 843 requests for change and 439 failures)
those. The total increase in the number of applications was 14.7% compared to the previous year. For the subsequent analysis, it is important to know that we operate with two temporary values characterizing each application — the time of actual closure (ie, from the moment when the user opened the requests to the moment when the IT officer closed the application) and the time spent by the IT officer to fulfill this application. . These time intervals may vary by times. This situation (the difference between the actual execution time and the labor costs) arises in the following cases:
- when the processing of an application is related to the transportation of equipment. For example, in the case of a geographically distributed network (we, for example, have the northernmost subdivision in Khanty-Mansiysk ~ 900km from the head office, there are no IT specialists there), with regular delivery once a week;
- when connected with the order of equipment / software out of stock . For example, they order expensive and piece equipment, which only arrives from Moscow to the head office in 4 weeks.
It should also be noted that in the 1st half of 2011, the number of IT employees was 9 (of which 4 were directly involved in supporting) people, in the 1st half of 2010 there were 8 of us (3 support employees). Accordingly, with the old number (8), for the six months of 2010 we managed to reach the average actual closing time of 1 application in 79.4 minutes. In 2011, with the number already in 9 people, the actual average time for closing an application was already 52.6 minutes. In fact, we improved our performance by 33.7% compared to last year, having increased our staff by 12.5%. Interestingly, the IT worker’s labor costs per request decreased by about 1.3 minutes (ie, they improved by 2.5% compared to last year).
What conclusions can be drawn? For example, it seems to me unequivocal that 3 IT staff to support 395 users is clearly not enough (this is evidenced by at least a 33% improvement in the time spent on an application after increasing the number of support staff to 4).
About the company-consumer of services: wholesale and retail sale of electronics, system integrator with a developed branch network
Source: https://habr.com/ru/post/125871/
All Articles