Management model for large amount of MS Windows Servers

Dear all,

I am working in a big company with around 30 applications, 100 server
machines and over 1000 employees. We manage our servers and run the
system for several years and now it becomes much complicated model,
the amount of applications and server hardwares tend to increase time
after time. We think about restructuring the server system to match
our requirement below.

1) Decrease amount of physical server hardware to decrease the load
for our data center accordingly.
2) Expansion capability for future application and hardware
3) High Availability for 99.99% system up.

Our approach to satisfy above requirements is:
1) and 2) Use virtual machines and SAN disk
3) Load balancing for web app servers, mirroring on database servers.

I wonder if anyone may share hands-on experience and best practice of
system administrator for similar caliber of system. Very appreciated.

Thank you.

Regards,
Khanh

 

Khanh,
Those are fairly generic expressions of "good things" when running
infrastructure. Starting from there you will end up with generic solutions:
Virtualization; High-availability techniques (cluster, RAID, load balance,
mirroring); SAN storage etc. You need more specific requirements for your
organization. The real problem is what technology to use where, and for
what.
Some analysis first is a good thing:
- There are some great tools to find out what data you are storing. You will
probably find that 80% is rarely if ever used, or is even business related.
- Tools to measure server utilization over an extended period. You can then
identify servers that are under-used. You can also see what services could
be amalgamated e.g. SQL services, web services
- Power consumption: how much power are you using in the data center and for
what?
- Service requirements: not just availability but recovery time, DR,
regulatory requirements. Probably you will have some services that are more
important than others.

Then you can start to put a solution together. You will have virtualization
and SAN in the solution somewhere, but as I say the problem is where and for
what.
Hope that helps,
Anthony





"Khanh" <khanhtruong@yahoo.com> wrote in message
news:f4a54921-ee4f-4dfe-90a2-42b9b697535c@l5g2000pra.googlegroups.com...<!--coloro:blue--><span style="color:blue <!--/coloro-->
> Dear all,
>
> I am working in a big company with around 30 applications, 100 server
> machines and over 1000 employees. We manage our servers and run the
> system for several years and now it becomes much complicated model,
> the amount of applications and server hardwares tend to increase time
> after time. We think about restructuring the server system to match
> our requirement below.
>
> 1) Decrease amount of physical server hardware to decrease the load
> for our data center accordingly.
> 2) Expansion capability for future application and hardware
> 3) High Availability for 99.99% system up.
>
> Our approach to satisfy above requirements is:
> 1) and 2) Use virtual machines and SAN disk
> 3) Load balancing for web app servers, mirroring on database servers.
>
> I wonder if anyone may share hands-on experience and best practice of
> system administrator for similar caliber of system. Very appreciated.
>
> Thank you.
>
> Regards,
> Khanh <!--colorc--><!--/colorc-->

 

Hi Anthony,

Actually we have a lot of tools to monitor our system including
monitor, server, network, I just like to look into a standard model in
infrastructure management for such our size. Our draft plan is as
follows:

1) For Web Application:
- First layer: utilize an ADC such as Crescendo or F5
- Second layer: Websphere(Java), IIS(.NET) and PHP webserver
- Third layer: DB server such as SQL Server, Oracle...

All webservers and DB servers are restructured to stay on Blade and be
virtualized if needed to have HighAvailability

2) For other applications:
All servers are also to be virtualized for availability if needed

3) Existing other specialized servers such as DataTransform server,
SysLog server for utilities.

Just wonder if you have any advice on above plan:
1) Does it exist any kind of future system which may be outside of our
structured plan and break our planned system?
2) Is there any problem on ADC device in reality?
3) Any other hidden issue that I may not see at this time?

Thanks in advance.

Regards,
Khanh

 

Hi Khanh,
There isn't a standard model based on size. It depends what your
requirements and priorities are.
What business are you in?
Anthony,




"Khanh" <khanhtruong@yahoo.com> wrote in message
news:d7d4f257-cd0e-42d1-bcbe-383770b73d93@i18g2000pro.googlegroups.com...<!--coloro:blue--><span style="color:blue <!--/coloro-->
> Hi Anthony,
>
> Actually we have a lot of tools to monitor our system including
> monitor, server, network, I just like to look into a standard model in
> infrastructure management for such our size. Our draft plan is as
> follows:
>
> 1) For Web Application:
> - First layer: utilize an ADC such as Crescendo or F5
> - Second layer: Websphere(Java), IIS(.NET) and PHP webserver
> - Third layer: DB server such as SQL Server, Oracle...
>
> All webservers and DB servers are restructured to stay on Blade and be
> virtualized if needed to have HighAvailability
>
> 2) For other applications:
> All servers are also to be virtualized for availability if needed
>
> 3) Existing other specialized servers such as DataTransform server,
> SysLog server for utilities.
>
> Just wonder if you have any advice on above plan:
> 1) Does it exist any kind of future system which may be outside of our
> structured plan and break our planned system?
> 2) Is there any problem on ADC device in reality?
> 3) Any other hidden issue that I may not see at this time?
>
> Thanks in advance.
>
> Regards,
> Khanh <!--colorc--><!--/colorc-->

 

Hi Anthony,

I am in insurance industry.

Thanks,
Khanh

 

"Khanh" <khanhtruong@yahoo.com> wrote in message
news:98e9ed7d-93d4-40fd-b99d-4c6caf6df849@x6g2000prc.googlegroups.com...<!--coloro:blue--><span style="color:blue <!--/coloro-->
> Hi Anthony,
>
> I am in insurance industry.
>
> Thanks,
> Khanh<!--colorc--><!--/colorc-->


Your plans sound find. I would also look into using blade technology along
with virtualization to reduce physical number of servers, as well as reduce
space, cooling and electrical requirements, increased wiring managment, etc,
striving for simplification, centralization and enhanced recoverability
(that virtualization provides). Many of the blade systems have built in
management tools that will also be beneficial.


--
Ace

This posting is provided "AS-IS" with no warranties or guarantees and
confers no rights.

Please reply back to the newsgroup or forum to benefit from collaboration
among responding engineers, and to help others benefit from your resolution.

Ace Fekay, MCT, MCTS Exchange, MCSE, MCSA 2003 & 2000, MCSA Messaging
Microsoft Certified Trainer

For urgent issues, please contact Microsoft PSS directly. Please check
for regional support phone numbers.

 

Hi Khanh,
1) General points
You will generally have requirements in several categories: Performance,
Availability, Recovery, Security, Compliance etc.
There is a huge range of technologies and practices to meet these
requirements. Just as ONE example, here is a reference to SQL design for
availability and performance:
.

2) Specific points
- The separation of roles is a good thing
- But I don't see virtualization as a method of achieving high availability.
Virtualization is a way of running more discrete workloads on a physical
machine and therefore increasing your effective computing capacity. If you
increase your effective capacity, then you can put some of it towards
achieving availability. VM's use different techniques to achieve high
availability.
Anthony





"Khanh" <khanhtruong@yahoo.com> wrote in message
news:98e9ed7d-93d4-40fd-b99d-4c6caf6df849@x6g2000prc.googlegroups.com...<!--coloro:blue--><span style="color:blue <!--/coloro-->
> Hi Anthony,
>
> I am in insurance industry.
>
> Thanks,
> Khanh <!--colorc--><!--/colorc-->

 

"Anthony [MVP]" <anthony@no-reply.com> wrote in message
news:00C900DF-C5FD-4E0A-8CB8-B7BF13D22A8F@microsoft.com...<!--coloro:blue--><span style="color:blue <!--/coloro-->
> Hi Khanh,
> 1) General points
> You will generally have requirements in several categories: Performance,
> Availability, Recovery, Security, Compliance etc.
> There is a huge range of technologies and practices to meet these
> requirements. Just as ONE example, here is a reference to SQL design for
> availability and performance:
> .
>
> 2) Specific points
> - The separation of roles is a good thing
> - But I don't see virtualization as a method of achieving high
> availability. Virtualization is a way of running more discrete workloads
> on a physical machine and therefore increasing your effective computing
> capacity. If you increase your effective capacity, then you can put some
> of it towards achieving availability. VM's use different techniques to
> achieve high availability.
> Anthony
>
><!--colorc--><!--/colorc-->


That's a good article on SQL design. It should be a good guideline to use.



Ace