How many HP-UX servers can ONE person handle in remote support?

[Univer_1]

How many HP-UX servers can ONE person handle on remote support in general? 50, 100, 200 or more?

The job about remote support include some Telephone support and some trouble shooting.

[Ivan Krastev]

It depends.
If you have some automated monitoring - OVO for example you can handle up to 100 servers.

regards,
ivan

[Anshumali]

it actually depends the nature of the servers and the response time, SLA agreed. I can handle all the servers with whatever size DC if there is no SLA. :)

Basically, considering the layered nature of support, i think the number of servers directly handled reduces with the level of support.

Thanks,
Anshu

[Robert-Jan Goossens]

Hi Univer,

my 2ct's

add on to the above answers, this will also depend on what you need to give support to, and what level of support eg fist line, second line or even thirth line.

First and maybe partial second line

OS + HardWare only 75-100+
OS + HardWare + middleware software support 50-75
OS + HardWare + full software support (even home build apps) 25-50 MAX

You will also need to know if the cu would like to have support on storage products (san, back-up and recovery software and (v)tape solutions).

Best regards,
Robert-Jan

[VK2COT]

Hello,

I agree with the previous replies.

To add more to discussion...

I worked for several major outsourcing
companies. Typical figure for an outsourcer
to "make money" on support teams was between
20-40 average-size servers per Unix admin.

I know it because various outsourcing
companies were retrenching staff once the
number of servers dropped below around
20 per person :(

Of course, complex servers deserve more
attention so the figures should
change...

Personally, when I worked for one
large service provider between 1994
and 2000, I alone managed around 100 Unix
servers (Corporate Oracle payroll and
procurement, firewalls, Web proxy servers,
email gateways, Web servers, even some
Cisco routers).

When I resigned on 4th of January 2000
(overworked and unhappy about not having pay
increase for five years, which was quite common in Australia), I was replaced by
seven Unix admins. That is one of
magics of life :)

In short, how many staff you need depends on
your SLAs and knowledge of support staff.

Cheers,

VK2COT

[Bill Hassell]

The answer is 1 to 100. Some companies throw everything at you (SAN disk, tape, networking, web management, database, backup, hardware upgrades and fixes), still others have hundreds of home-grown applications that require constant maintenance. This reduces the maximum machine count to just a few. Remote support is often grossly oversimplified and not until a major problem occurs will management realize that telnet to a single LAN just won't work.

Given enough time (about 6-12 months), you can develop tools to automate managment (log scan and trim, disk space monitors, performance monitors, backup scripts, security checks, spooler monitors, network monitors, etc) to make 20-50 servers work fairly well. Or given enough money, you can buy OpenView packages that will do the job quite well for hundreds of servers.

HP-UX servers tend to be very reliable and if nothing changes (no new RAM, disk, peripherals, I/O cards), the hardware just hums along. Most maintenance tasks will be failed disks (a very difficult taks to handle remotely with cheap JBOD disks) and finding/fixing obscure problems caused by changes to the system, especially when multiple users have the root password.

The system administrator's job is very poorly understood by management and when you do your job perfectly, people will often ask: "What is it that you do around here?" I would estimate the number of machines after first defining exactly what is expected. And if remote access does *NOT* also include a protected connection to GSP/MP ports, reduce the quantity by half. Trying to talk some remote PC user through the steps to boot into single user mode when you can't see anything will be a very slow process.

[Geoff Wild]

I say - 32.

That would be the limit to give great support (not just get by). Anything more would mean not doing a thourough enough job. Depending on outage windows - one could spend up to 2 months a year just in the normal patch cycle. Sure - some people could do more servers - a lot of it depends on what you have to support. If you are talking about the tasks of an Unix Admin - then I'd still go with 32. A lot depends on how complicated the environment is...etc...etc...

32 just seems like a nice number :)

Rgds...Geoff

[John Payne_2]

I agree with Bill and Geoff. Your number is not likely to be high.

One more thing about Bill's comment about custom apps. Bill talked about stable HW, but how stable is the software? If you have someone else constantly tinkering with the applications everywhere, or absolutely positively worse, someone who has root password and things they know what they are doing, there is no way you can even approach 32 servers. You will spend all your time cleaning up messes.

You will begin to hate nothing more than the phrase "We didn't do anything..."

Now, If they are systems you control - Installation, patching, application modification, root password, etc. You can do a lot more. If the systems and apps are fairly stable, there's no reason why the number of servers couldn't increase...

John

[Sp4admin]

I currenly support 36 serves remotely. This is for OS and some application support. This keeps me very busy.

sp,

[Michael Steele_2]

Telco companies like ATT, SBC, Telcordia, etc., have used a production support policy 'called' the mushroom policy for several years now. They build their prod. boxes, put them in a dark closet and forget about them. (* They don't feed them s###, like your supposed too, this they reserve for their admins. *)

Anyway, under this policy a department of 22 SA's might be responsible for ~750 servers. But as this is a 7x24x365 mushroom closet and all shifts are covered including the late night grave yard shift, you'll only have one SA on-call for all ~750 servers in the mushroom closet.

What you end up with is a tier one comprised entirely of OpenView monitoring, numerous false alarms, and a tier 2 group comprised of the 22 SA's. There are no SA's at the tier one support level. Tier 3 will then be the developers who built the box.

However, as time goes on, these tier 3 people move along to other projects. Leaving orphaned mushrooms behind without a parent.

Orphans really suck.

[Emil Velez]

one additional consideration is the hardware platforms and patch management.

How similar is the hardware of all of the HPUX systems. Are they all the same model with the same OS load ? If so very preductable as far as software maintenance.

How often does the CU want you do do patch analysis for security and OS patches. if you do it once a quarter and all of the systems are at the same release that simplifies things further. IF you have dissimilar hardware and different OS versions or the same OS versions loaded at different times on dissimilar hardware that adds another level of complexity.

[Steven E. Protter]

Shalom,

I currently work on a team of 4 that handles support on 85 HP-UX and 100 or so Linux servers.

WE don't do much on networking or the SAN side, well only a few.

We provide on demand support to several hundred other servers around the globe.

We work pretty hard but are not at capacity.

SEP

[Bill Hassell]

My real world example is what I am doing now. We have a DBA and 5 SA's to run 5 rp8420's, 4 rp2470's and about 25 Linux/Windows boxes. One SA is dedicated to disk management of 4 EMC Symmetrix and about 150 TB of EMC Centera. Another is dedicated to Data Protector plus custom tape applications that write, duplicate and read hundreds of GB each day. A 3rd SA is managing aging data (7-10 years) requiring endless scripting to track millions of data records. The 4th is primarily focused on Linux+Windows boxes that run custom apps for banks and I work mostly on general sysadmin tasks for production, development and a DR site.

There literally thousands of custom scripts handling all the tasks for this banking application and dozens of gigabytes of logs required by auditors. Although full root access is limited to selected SA's, sudo is used extensively for privilege management. Change management is quite detailed but there are 5-20 changes each week due to the nature of the business. 50-60 hours/week are fairly standard and major architectural changes are planned for this year (more disk, major application changes, more customers).