HP Device Manager 4.5 - Deployment Guide

Deployment factors
• Total number of managed thin clients
– Use more powerful, server-level hardware configuration; for example, RAID.
– 20,000 is the recommended maximum number in one HPDM system. Lab testing shows that HPDM performs most
efficiently up to this number.
– Current customers successfully use HPDM to manage up to 50,000 devices with a single server with these
considerations:
• Use of networks faster than 1000 MB.
• Efficient placement of gateways and repositories to maximize the use of LAN-based communications.
• Use of scheduled tasks to execute tasks during non-peak hours of operation.
Note: HPDM is a very flexible system and supports the use of any number of HPDM Servers, HPDM Gateways, and
repositories to match the customer’s existing network architecture.
• Volume of transferred files
– Add more Child Repositories.
• PXE imaging
– Deploy HPDM Gateway in the same subnet as the thin clients.
– If the subnet is a NAT subnet, configure port mapping on NAT to make sure that the HPDM Server can talk to the HPDM
Gateway directly.
Advantages of additional HPDM Gateways
• Required for PXE imaging tasks.
• Consolidated communication between branch offices and DMZ.
• Accelerated task delivery speed in case of NAT Gateway in branch offices.
Advantages of additional repositories
• Reduced traffic between branch offices and DMZ.
• Faster software updates and image deployment.
Determining number of repositories
• Number of repositories = (transferred data ÷ bandwidth) ÷ expected time spent
For example, you have 20,000 units to be re-imaged. Each image is 1 GB. Therefore you have 20,000 GB (20 TB) of data to
transfer.
Your connection from one repository to thin clients is 100 Mbps. It will take 444.4 hours to transfer all data.
20,000 GB ÷ (100 Mbps ÷ 8 bits per byte ÷ 1000 x 3600 seconds per hour) ≈ 444.4 hours
To reduce it to 48 hours, you need 10 repositories (444.4 ÷ 48). There is some overhead to synchronize from the Master
Repository to the Child Repositories. Take that into consideration.
Optimizing large image deployments using Batch Control
Batch Control is an optional task parameter that can be used when deploying images or other tasks with payload to a large
number of thin clients. The batch settings control how many thin clients are sent the task at a time, which gives you some
control over the amount of network traffic HPDM generates.
The continued write speed of current thin client models is around 3.5 MBps (megabytes per second). Use that number to
decide how many imaging tasks you can send in one batch for each repository and how long the interval between batches
should be. For details about setting batch control, see the Batch tasks
section later in this document.
For example, you are deploying a 4 GB image to thin clients. The network is a high quality 1000 Mbps Ethernet, which means
it can provide continuous theoretical bandwidth. Many of the latest 7200 rpm hard drives have higher continuous read
speeds than that, and those using RAID even higher. Assuming that the network is the bottleneck, we can have the
following:
1000 Mbps ÷ 8 bits per byte ÷ 3.5 MBps ≈ 35 concurrent imaging tasks in a batch
To calculate the time it will take:
4