1. Capacity Planning Plans Capacity Planning Operational Laws

2. Building an Efficient Capacity Plan

3. Steps to Create a Capacity Plan
Define the Business Problem
Define the Service Level
Define the Current Capacity: Workload Characterization
Plan the Future

4. Define the business problem
Service Provider
Vice and Data
Drop Calls Rate =5%
Service Level
Acceptable Drop calls < 1%

5. More Business Data 55,000,000 subscribers Lagos issue with drop calls
1 Erlang= 60 minutes of talk
BHT Busy Hour Traffic
5 pm
4 calls in average
5 minutes
1000 subscribers
Erlang for the BTS = 1000 x 4 x 5 /60=

6. Business problem Congestions in the core network
Congested networks lead to packet loss, jitter and packet delays
SLA to customer is 70% of guaranteed rate
Issue: at peak time some customers experience less than 70% of guaranteed rate

7. More data
Peak hour: 10 pm, 1000 customers connected
5 packages

8. Capacity Planning Input and Output Variables
Processors, disks, network
Max no of connections
RT <= 2 sec
Throughput >10 tps
Basic
Service
Parameters
Workload
Evolution
Desired Service
Levels
Capacity Planning
Saturation Points
Cost-effective
Alternatives
When are Service Level violated?

9. Business Problem Response time of CRM is too slow ~5 s
SLA: 0.5 s – 1 s is acceptable

10. Application Server(s)
End-Users
Web Server(s)
Application Server(s)
Storage
SAN
Database Server(s)

11. Workload Characterization
CRM: create customers, look up for customers, delete customers, search customers, analytical reports (data mining, big data, Hadoop)
Classify workloads:

12. MTN Nigeria’s Capacity Plan
Increasing the Performance and Efficiency of Your Capacity Planning

13. Capacity Planning Framework for Application X
Workload
From Application X
System X
Parameters
Desired
service
Levels for Application X
Saturation
Points for system X
Cost-effective
Alternatives for System X

14. Business Models
Resource Model

15. Application Y

16. Application Z

17. Business Problem X

18. Business Problem Y

19. Business Problem Z

20. Performance and capacity mgt in a Software Defined Datacentre (SDDC)
Most important, the SDDC enables automated, policy-driven provisioning and management of data center resources.
Program interfaces make it possible for applications to request resources based on clearly defined rules and policies. The result:
A more responsive, agile, secure and high-performing data center that takes full advantage of the underlying hardware

21. Three keys to having an SDDC
1) Capacity management
2) Multi-virtualization and multi-cloud management platforms
3) Configuration management

22. Capacity Management
A SDDC is about rapidly provisioning hardware to users. But, a critical element to that is to ensure there is enough capacity to provision.
One of the first steps in a SDDC migration is to ensure that your data center/IT shop has enough capacity for the needs of the organization, applications and services.
You can’t automate the provisioning of resources unless you have enough resources to serve the business.
Level setting the needs of the business and ensuring you have the capacity is an essential first step.
There are a variety of tools that can help with this, including ones from BMC (PractiveNet), CA Performance Management and even smaller providers like VMTurbo.  

23. Multi-virtualization and multi-cloud management platforms
Data centers will have complicated architectures. It’s rare today to find a data center that’s all in with one vendor; it’s usually a mix of technologies from multiple different providers.
Maybe a business used to be a VMware shop for its virtualization but it’s recently begun using Microsoft Hyper-V. Maybe it has used Amazon Web Services, but it wants to start using a private cloud from a more niche service provider.
A key to managing this complex, heterogeneous environment is to have a multi-virtualization and multi-cloud management platform
Increasingly, cloud management vendors are embracing a strategy of supporting multiple platforms.

24. Configuration management
Another key to a true SDDC approach is to move from a manual to automatic provisioning of resources. This would be exemplified by an operations professional getting the specifications of an application or service, then setting up the hardware on a case by case basis. The more efficient alternative is to instead automatically provision based on the applications need.
This is basically the idea behind a “devops” mentality, meaning that developers and operations folks work much closer together. Tools like Puppet and Chef help companies achieve automatic provisioning. Whereas capacity management will ensure there are enough resources to provision, configuration management will automatically allocate those resources without the need to have manual scripts.  
Overall, the idea of the SDDC is that it provides an additional layer of abstraction above the hardware components, public and private cloud, which “empowers applications to define their own environments, based on performance, security, availability and further policy requirements."

25. Capacity Management and Software-Defined Data Center

26. Capacity Planning Operational Laws
Utilization Law
The utilization (Ui ) of resource i is the fraction of time that the resource is busy.
Ui = Xi * Si = i * Si
Or Xi = i
UTILIZATION LAW Ui = Xi So
LITTLE'S LAW N = X R
RESPONSE TIME LAW R = N/X – Z
THE FORCED FLOW LAW: Xi = Vi Xo
The Forced Flow Law
Xi = Vi * Xo
XO = Xo/ Vi
B is the length of time that the resource was observed to be BUSY.
C is the number of request DEPARTURES observed.
D Service Demand is the sum of all service times for a request at resource i N
Q Queuing Time is the sum of all waiting times for a request at resource i
R a system response Time or a Device Residence time
S service time; period of time a request is receiving service from resource i, such as CPU or disk
T is the length of TIME we observed the system.
Vk the visit ratio, is the number of times device k is visited per transaction.
W is the ACCUMULATED TIME for all requests within the system – time spent both waiting for and using resources.
X throughput of a device or a system
Z is the think time of a terminal user
Service Demand Law
Di = Vi * Si = (Xi/Xo)(Ui/Xi) = Ui / Xo
Dcpu = Vcpu * Scpu = Ucpu / Xserver
Little’s Law X N
Residence Time (R’i) at resource i is the sum of service demand plus queuing time. R’i = Qi + Di
Response time (Rr) of a request r is the sum of that request’s residence time at all resources. Rserver = R’cpu + R’disk R
N = R * X