HP Bc1500 - BladeSystem - Blade PC Manuel
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.
2
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.
24
0
100% 100%
0%
Duration of inactivity before forced log off occurs
(in Hours)
% of blades-
to-users
needed to
meet peak
demand
1%
6 12
50% MCU = 70%
(with forced log off set
to 2 hours
)
% of CCI
users on
the
network
18
3
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.
4
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
5
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.
6
7
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
© 2005 Hewlett-Packard Development Company, L.P. The information contained
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
8
Autres manuels pour Bc1500 - BladeSystem - Blade PC
9
Ce manuel convient aux modèles suivants
1
Table des matières
Autres manuels HP Matériel réseau
HP
HP 205H-MNT2 Manuel utilisateur
HP
HP 201723-B21 - HP StorageWorks Modular SAN Array 1000 Hard... Manuel de programmation
HP
HP XP 9000 Manuel utilisateur
HP
HP BladeSystem C-Class Interconnect Component Instructions d'utilisation
HP
HP Ethernet Network Interface Card hn230e Manuel utilisateur
HP
HP Aruba 640 Instructions de montage
HP
HP ProCurve Networking J8451A Manuel utilisateur
HP
HP HSR6602-G Instructions d'installation
HP
HP E-MSM466-R Instructions de montage
HP
HP In/Out P2P 10/13dBi MIMO 3 Instructions de montage
HP
HP StorageWorks x9300 Manuel de service
HP
HP Enterprise Aruba 7210 Instructions de montage
HP
HP E8491B IEEE 1394 Manuel utilisateur
HP
HP 30020B Manuel
HP
HP X1600 - StorageWorks Network Storage System 5.4TB SAS Model NAS... Comment utiliser
HP
HP X1600 - StorageWorks Network Storage System 5.4TB SAS Model NAS... Manuel utilisateur
HP
HP A5920AF-24XG Bk(pwr)-Frt(prt) Instructions de montage
HP
HP ProCurve Radio Port 210 Instructions de montage
HP
HP procurve radio port 220 Instructions de montage
HP
HP AdvanceNet 30244M Manuel
Manuels Matériel réseau populaires d'autres marques
Matrix Switch Corporation
Matrix Switch Corporation MSC-HD161DEL Manuel utilisateur
B&B Electronics
B&B Electronics ZXT9-IO-222R2 Manuel utilisateur
Yudor
Yudor YDS-16 Manuel utilisateur
D-Link
D-Link ShareCenter DNS-320L Manuel utilisateur
Samsung
Samsung ES1642dc Instructions d'utilisation
Honeywell Home
Honeywell Home LTEM-PV Instructions de montage
