HP Bc1500 - BladeSystem - Blade PC Manual

Calculating the Maximum Concurrent Users in the HP

Consolidated Client Infrastructure (CCI) v1.3 Environment

Abstract.............................................................................................................................................. 2

Introduction......................................................................................................................................... 2

Calculating the MCU............................................................................................................................ 2

MCU Calculation Examples .................................................................................................................. 5

Example 1: Small Enterprise Company ............................................................................................... 5

Example 2: Medium Enterprise Company ........................................................................................... 5

Example 3: Large Enterprise Company ............................................................................................... 6

Static Allocation and Concurrency ..................................................................................................... 6

Summary ............................................................................................................................................ 6

For more information............................................................................................................................ 8

Abstract

A cost-efficient Consolidated Client Infrastructure (CCI) implementation is one that can satisfy end user

demands without an excessive amount of hardware. Since the maximum number of concurrent users

defines the amount of hardware to be purchased, calculating maximum concurrency becomes an

important aspect of designing a CCI solution. This document provides guidelines for calculating the

amount of hardware required to meet maximum concurrency.

Introduction

The most cost-efficient model of CCI uses Dynamic Allocation methodology to allocate one blade PC

to every end user who logs in. Blade allocation is essentially a random assignment. There must be at

least as many blades available as there are employees trying to log-in or already logged on at any

one moment in time.

Obviously, it is extremely unlikely that in a sample size of 1,000 employees, all employees will be

trying to use a PC at the same time. As such, with CCI’s Dynamic Allocation model, there is no need

to purchase and support a blade PC for every possible user. Instead, a CCI implementation needs

only enough hardware to support the number of Maximum Concurrent Users (MCU).

HP has concluded that a maximum concurrency rating of 70% is the recommended baseline for large

CCI implementations.

Calculating the MCU

In establishing 70% as the baseline MCU rate, the following assumptions are made:

•There are 250 work days in each year ((52 weeks x 5 work days) – 10 national holidays).

•Each employee is actively on their computer an average of 225 days per year, based on the

following number of days each employee takes off from work: 10 vacation days, 5 sick days, 5

paid time off days, 2.5 off-site training days, and 2.5 travel days.

•On average, each employee is not working an additional 10 days for the following reasons:

maternity and paternity leave, emergency leave, military leave, academic and other leaves of

absence,

•The above assumptions result in an average employee using their PC 215 days a year.

If a typical user was able to claim a blade for the entire workday, and absences were evenly

distributed across a year for all workers, there would only be a need to purchase and support 86%

(215/250) as many blades as there are employees.

Furthermore, if blade PCs can be returned to a common pool if idle for a specified period of time (as

can be done with CCI), then the MCU can be much less than 86%. The assumption is that at least

25% of a typical large enterprise’s personal computers are idle at least 2 hours a day even for those

people who are working. This is due to employees coming in late, leaving early, long lunches,

meetings, briefings, seminars, etc. This implies that of all employees, only 64% are active on their

computers at any one time. However, a computer that is idle but still logged on cannot be part of the

pool available for assignment to another user. In addition, since peak demand (not just the average)

must be met, we round up by 10% and conclude that an MCU assumption of about 70% is

reasonable as a starting point.

Although the HP baseline MCU of 70% is a reasonable assumption for most companies, the old

saying “Your mileage will vary” applies. There are many factors that can influence what a

corporation’s MCU will be. The amount of idle time before a user is automatically logged off has the

biggest impact on MCU. Figure 1 shows a conceptual model of how these factors interact to indicate

MCU percentage. The graph illustrates the results of the forced log off setting. Note that as the

“Duration of inactivity before forced log off occurs” setting increases in hourly increments, the number

of users remaining logged on is likely to increase rapidly.

Figure 1. Maximum Concurrent Users (MCU) Estimation Summary

The following factors can raise the MCU (requiring more blades):

m concurrency will be higher.

•ees will

The following factors can lower the MCU (requiring fewer blades):

oncurrency will be lower.

zones,

•is to log on only rarely – for instance, employees on a shop

•If the ‘log off if idle’ time is set to be more than 2 hours, maximu

•If the company has seasonal periods during which a high percentage of employees are working

(such as a tax preparation office in the spring), maximum concurrency will be higher.

If wireless devices such as PDAs and portables are used as CCI access devices, employ

remain productive by staying logged on and active even while at lunch, in meetings, etc., so

maximum concurrency will be higher.

•If the ‘log off if idle’ time is set to be less than 2 hours, maximum c

•If the pool of users are shift workers in one location, or are separated and span multiple time

the maximum concurrency could be lower. If they are so widely separated that their work hours do

not overlap, the maximum concurrency could be much lower, even less than half of the total number

of employees supported by the pool.

If the work force’s natural work model

floor that only work on a computer a few hours a day – the MCU could be much lower.

100% 100%

Duration of inactivity before forced log off occurs

(in Hours)

% of blades-

to-users

needed to

meet peak

demand

6 12

50% MCU = 70%

(with forced log off set

to 2 hours

)

% of CCI

users on

the

network

Though every effort should be made to correctly gauge the MCU as close as possible, CCI scales

easily and blade PCs can be added quickly should a company’s MCU be underestimated. Though

acquisition cost does matter, the cost delta between a typical 70% MCU and 95% MCU on a per

seat basis is estimated to be only several hundred dollars more per seat, and as such, because 80%

of traditional PC total cost of ownership (TCO) is support costs and not acquisition costs, the impact to

the TCO should be negligible over the long-term. The best practice is to error on the high side when

calculating the MCU.

Note that CCI blade PCs are sold in 10-packs and deployed in enclosures that hold 20 blades. To

take maximum advantage of the solution, all calculations should be rounded up to the nearest 20

blade PCs. For instance, if a corporation concludes they need 815 blades to support 1200 users, the

actual number of blade PCs purchased should be 820.

The calculations above assume large pools of employees (more than 200). With smaller pools of

users, the averages and standard deviations of employees using their computer can vary widely from

day-to-day in response to small changes in the number of people out sick, etc. For instance, to use an

extreme example to illustrate the point, with a workgroup of 10 employees, an MCU of 70% (7

blades) will not be enough if 8 or more employees are at work on a given day.

Consequently, the smaller the pool of users, the higher the MCU should be. As a rule of thumb – and

in conjunction with other factors listed in this paper, HP suggests an initial MCU value for the

following user pool sizes:

•Less than 80 users, MCU of 100%

•80 – 140 users, MCU of 90%

•140 – 200 users, MCU of 80%

Note:

Returning an idle blade PC to a pool of blades effectively ‘kicks off’ a user, and files that have not been

saved can be lost. This scenario makes it essential that all users invoke ‘auto save’ features and that the

timing for auto save is less than the maximum idle time before forced log off. For instance, if the forced

log off time is 2 hours, then auto save for all applications hours should be set to less than 2 hours.

MCU Calculation Examples

Example 1: Small Enterprise Company

The business in this example is a 4,000 employee enterprise in the central US for which the

management has concluded the value proposition of CCI is appropriate for 1000 users. Though

many are work-at-home employees and often access their blades from a number of satellite offices as

well as their home, they are all within 200 miles of each other in the same time zone and work

generally the same hours.

Table 1. Blade calculation for small enterprise company

Parameter Value

Number of employees using CCI 1000 users

86% of employees at work on any given day 860 active users

25% lower concurrency due to forced logoff time set to 2

hours

645 active users

10% delta added for peak planning 710 active users

Calculated MCU percentage rate 71%

Number of blades needed 710 blades

Number of blades ordered 720 blades

Example 2: Medium Enterprise Company

The business in this example is a 10,000 employee global enterprise with 8,000 employees in

America, and 2,000 in a call center in India that is only open 9 hours a day. The management has

concluded the value proposition of CCI is appropriate for all Call Center staff, and 2,000 employees

in the corporate HQs, for a total of 4,000 employees. All blades are located in America, and

accessible and usable from anywhere in the world over the corporate WAN.

Table 2. Blade calculation for medium enterprise company

Parameter Value

Number of employees using CCI 4,000 users

Maximum number of employees using CCI at either site

during normal business hours

2,000 active users

86% of employees at work on any given day at either site 1,720 active users

Assumption that maximum number of users still logged

during business hours at other site is 5% of 2000

1820 active users

10% delta added for peak planning 2002 active users

Lower concurrency due to forced logoff time set to 2 hours

does not apply as call center staff are discouraged from

leaving their desk for more than 60 minutes

n/a

Calculated MCU percentage rate 50%

Number of blades needed 2002 blades

Number of blades ordered 2020 blades

Example 3: Large Enterprise Company

The business in this example has 15,000 full time employees. It is a US-based tax preparation firm

with thousands of retail offices throughout the United States. This firm adds an extra 5,000 temporary

employees from January thru April to assist with preparing customer’s tax documents. All employees

are asked to avoid taking vacation or any other type of leave in the critical months of January thru

April. This corporation feels that 79% of their total employees can migrate to CCI due especially to its

very low break/fix field costs and because no customer’s personal files are stored on local devices.

All blades are located in NA, and accessible and usable from anywhere in the world over the

corporate WAN.

Table 3. Blade calculation for large enterprise company

Parameter Value

Number of employees using CCI 15,000 active users

Percentage of employees using CCI at work at any given

day (80% between May and December, 95% between

January through April, so use 95%)

14,250 active users

Lower concurrency due to forced logoff time set to 1 hour,

but rare that computers are not in use for more than 30

minutes, (“50” only 1%)

14,108 active users

1% delta added for peak planning (low percentage due

to already near maximum utilization rate)

14,249 active users

Calculated MCU percentage rate 95%

Number of blades needed 14,249 blades

Number of blades ordered 14,260 blades

Static Allocation and Concurrency

With a static allocation model of CCI, individual employees have a dedicated blade that they return

to every time they log on. This is comparable to having a reserved parking place in a parking lot – no

matter how many or few of the other parking spots are taken, one spot is designated for a particular

person with their name on it. This means that there will be a parking spot for every car, even if not

always filled. With the static version of CCI, there must be a blade for every end user whether they

are logged on or not. If the corporation plans to migrate 2,000 users to CCI in a static model, then

2,000 blade PCs are needed.

In the static model, HP recommends that customers add 1% to be designated as standby hardware in

case a blade PC fails, resulting in a calculated MCU of 101%. So in this example of a static

allocation, the total number of blades ordered and put into production should be 2,020.

Summary

The intent of this white paper was to provide information on how HP has concluded 70% maximum

concurrency as a baseline for CCI implementations, and include some general guidance and

information on how MCUs could be estimated for CCI implementations. Though some examples and

basic guidelines were presented here, every enterprise is different. CCI architects should work with

the customer’s HP account team and HP Service to calculate the MCU, and consequently their

hardware needs for a CCI solution.

For more information

For more information on implementing a CCI solution refer to the following URL:

http://h20223.www2.hp.com/NonStopComputing/cache/10793-0-0-225-121.aspx

HP Implementation Service for CCI

herein is subject to change without notice. The only warranties for HP products and

services are set forth in the express warranty statements accompanying such

products and services. Nothing herein should be construed as constituting an

additional warranty. HP shall not be liable for technical or editorial errors or

omissions contained herein.

Itanium is a trademark or registered trademark of Intel Corporation in the U.S. and

other countries and is used under license.

385672-001, 01/2005

Otros manuales para Bc1500 - BladeSystem - Blade PC

Este manual sirve para los siguientes modelos

Tabla de contenidos

Otros manuales de Hardware de red de HP

HP Enterprise Aruba 7210 Instrucciones de montaje

HP StorageWorks AB466A Manual de usuario

HP 5500-24G-SFP EI TAA Instrucciones de montaje

HP MSA 2040 Manual de usuario

HP StorageWorks HA-Fabric Manager Manual de usuario

HP HP 16500 Instrucciones de instalación y mantenimiento

HP A5920AF-24XG Bk(pwr)-Frt(prt) Instrucciones de montaje

HP AJ878A - StorageWorks 320 GB Hard Drive Manual de usuario

HP HPE Synergy 12000 Frame Instrucciones de funcionamiento y seguridad

HP StorageWorks X9320 Manual de servicio

HP X1600 - StorageWorks Network Storage System 5.4TB SAS Model NAS... Manual de usuario

HP 12500 Series Manual de usuario

HP Mv2120 - Media Vault Network Drive Manual de usuario

HP Presario X1000 - Notebook PC Instrucciones de uso

HP ARUBA 8320 48 10/6 40 X472 5 2 Bdl Instrucciones de montaje

HP StorageWorks 2324fc Manual de usuario

HP Edgeline EL1000 Manual de usuario

HP FF 12910 Instrucciones de montaje

HP StorageWorks 1/8 Tape Autoloader Manual de usuario

HP 800 Series Manual de usuario

HP Aruba 640 Instrucciones de montaje

HP Wireless LAN 802.11b/g Device WN4201B Manual de usuario

HP c-Class Guía de instalación

HP X1600 - StorageWorks Network Storage System 5.4TB SAS Model NAS... Manual de usuario

Manuales populares de Hardware de red de otras marcas

Matrix Switch Corporation

Matrix Switch Corporation MSC-HD161DEL Manual de usuario

B&B Electronics

B&B Electronics ZXT9-IO-222R2 Manual de usuario

Yudor

Yudor YDS-16 Manual de usuario

D-Link

D-Link ShareCenter DNS-320L Manual de usuario

Samsung

Samsung ES1642dc Instrucciones de uso

Honeywell Home

Honeywell Home LTEM-PV Instrucciones de montaje

evertz

evertz 7700 Series Manual de usuario

D-Link

D-Link ShareCenter DNS-343 Manual de usuario

Asante

Asante AsanteBridge 1012 Manual de usuario

FireBrick

FireBrick FB2700 Manual de usuario

Audiovox Electronics

Audiovox Electronics ARWH2T Manual de usuario

ADC

ADC DSX-1 Manual

NetComm

NetComm NF2 Manual de usuario

Alcatel-Lucent

Alcatel-Lucent 5620 Manual de usuario

Dell EMC

Dell EMC XC6420 Series Manual de usuario

Lanner

Lanner NCA-1020A Manual de usuario

Allied Telesis

Allied Telesis AT-TN121-A Manual de usuario

Arista

Arista CCA-ETM-Q12 Manual de usuario