Home
 > Research and Reports > TEC Blog > How Do You Categorize Servers?

How Do You Categorize Servers?

Written By: R. Krause
Published On: September 5 2000

How Do You Categorize Servers?
R. Krause - September 5, 2000

Overview

In the world of Wintel-based Servers (a/k/a PC Servers), the types of machines offered are as varied as the applications for which they are used. Although it is risky to apply terms too broadly, we can break the server market into six segments, and provide a guide to the features and characteristics of each. (There are other ways the market can be segmented - we have chosen this because it more "accessible" to the general user.)

In general, servers are segmented by the size of the user population they serve - from small "Mom and Pop" shops up to large companies with thousands of employees. However, there are exceptions, which will also be explored herein.

Segments

The size-related segments we have defined, from smallest population served to largest, are:

  1. Workgroup
  2. Department
  3. Mid-Range
  4. Enterprise
  5. Super-Enterprise

All of the above segments are for general-purpose servers - those which try to have a broad mix of features and functions, to appeal to the widest range of customers The sixth segment is relatively new, only a few years old. Rather than being defined by number of users, it is categorized in terms of what it does, not whom it serves. This segment is variously referred to as "Single-Function Server", "Server Appliance", "Information Appliance", "Thin Server", and "Task Server" - each designation having slightly different connotations. For simplicity, we will lump the sub-genres together and refer to the general product group as "Appliance Server". This segment (for both Wintel and Unix/Linux-based systems) is predicted to grow to $8 Billion by 2003. [Source: IDC].

Segment Definitions

For the purposes of this document, we will use the definitions below. (Over time, these definitions have changed, and will continue to change, and we expect that segments will consolidate.) Refer to Graph 1 for approximate user quantity vs. segment map.

  1. Workgroup:
    Users: Generally serves small groups, from as little as two or three, up to 20 or 30 clients/users.
    Pricing: Usually low-priced (under $2000)
    Typical Uses: E-mail, file, and print serving.
    Functionality: Medium functionality, positioned slightly above a high-end PC.
    Examples: Compaq ProLiant ML330, Dell PowerEdge 1300, HP NetServer E60, IBM Netfinity 3500 M20
     
  2. Department:
    Users: Handles moderate-sized groups, from 10 users up to 50+ (or even 100) users
    Pricing: Under $2000 base price for some models
    Typical Uses: E-mail, file, and print serving
    Functionality: Increased reliability, power, and functionality
    Additional Comments: The line between workgroup and department servers is becoming blurred as high-end computing power increases.
    Examples: PowerEdge 4400, NetServer LC2000, Netfinity 5100, ProLiant ML350
     
  3. Mid-Range:
    Users: Handles larger groups, starting around 50+ users, up to 100-200+ users.
    Pricing: Starting around $3000-$4000 for base systems
    Typical Uses: Messaging/e-mail, file/print serving, mid-range compute-intense tasks
    Functionality: Reliability and performance features become more significant/powerful.
    Additional Comments: These servers are sometimes referred to as "Small-Medium Business Servers". The line between department and mid-range servers is becoming blurred as high-end computing power increases.
    Examples: NetServer LH4r, Netfinity 5600, ProLiant ML370, PowerEdge 6450
     
  4. Enterprise:
    Users: Handles large groups, 200-500+ users/clients.
    Pricing: Starting above $5000, generally in the $7K-$8K range for a base system. Fully configured systems can easily exceed $500K
    Typical Uses: Data warehousing, large database management, and heavy-duty transaction processing.
    Functionality: Much larger emphasis on reliability, performance, and functionality.
    Additional Comments: At this level, servers are sometimes architected to have functionality moved "outside the box". For example, hard disk storage within the server may be minimal (just enough to hold the OS), and the server is designed to be rack-mounted with a large RAID array installed in the same rack.
    Examples: Netfinity 7600, ProLiant ML570, PowerEdge 6400, NetServer LH6000
     
  5. Super-Enterprise:
    Users: Generally exceeding 500 users/clients, sometimes more than 1000
    Pricing: Starts in the $20K range for basic systems, but a fully configured system, plus associated storage, memory, and options can exceed $1 Million.
    Typical Uses: Data warehousing, massive DB management, high intensity transaction processing (>40,000 tpmC).
    Functionality: Reliability is extremely important, more so than pricing. Computing performance and functionality are also key.
    Examples: ProLiant 8500, PowerEdge 8450, NetServer LXr 8500, Netfinity 8500
     
  6. Appliance:
    Users: ISPs, small groups,
    Pricing: Around $1000 for a base system
    Typical Uses: Web caching, web serving, intranet serving, Storage Area Networks
    Functionality: Easy to install/set up, scalability, CPU density (in rackmount systems)
    Examples: Cobalt Networks RaQ, Netfinity 4000R, NetServer LPr, Compaq TaskSmart, Dell PowerApp

Correlation Factors

Now that we have defined the segments, what can we expect to get for our money, and what is the correlation between market segment and feature/function set? In other words, on which areas (e.g., processing power, reliability, price) can we expect each of the segments to focus?

The features and functions that matter in servers are separated into two general groups, which we can call "Goals" (which are more strategic in nature) and "Features" (which are more tactical in nature). In other words, "Goals" are the overarching categories, and "Features" are the implementation methods or physical constructs by which the goals are achieved.

The factors listed in Table 1 are a sampling of some of the more important ones; there are others that users will find relevant to their particular situation or need. Different users will have different requirements, and thus factor weightings will vary by user, so the relative "weight" of each factor is not part of Table 1.

Customers can use the Table, in concert with other sizing tools (usually based on the software or infrastructure, such as SAP R/3 or Microsoft Exchange), to help them determine which class of server will be best suited to their needs.

Table 1 shows the correlation of the various Goals and Features to the defined segments.

Table 1

App. Wkgrp. Dept Mid-Range Ent. Super-
Ent.
Goals
           
Performance/
Power
H
M
M
H
H
H
High
Reliability
L-M
M
M
M-H
H
H
Price
H
H
H
M
M
M
TCO
H
H
H
H
M
M
Flexibility
L
L
M
M
H
H
Service/
Support
L
M
M
M
H
H
 
Features
Multiple
CPUs (qty.)
L
L
L
M
H
H
Storage (internal)
L
M
H
H
H
M
redundancy
L
L
M
M
H
H
     Fans
L
L
L
M
H
H
     Power
M
L
M
H
H
H
     NICs
M
L
L
M
H
H
Hot Swap/
Maintainability
L
L-M
M
M
H
H
     Fans
L
L
 M
M
H
H
     Drives
M-H
M
M
H
H
H
     Power
L
L
M
M
H
H
 
How to Interpret
Table 1
Correlation Factor
 L
Low
L-M 
Low/Medium
M
Medium
M-H
Medium/High
H
High
For any given factor, a correlation rating (Low, Medium, High) is given for each of the segments. This rating is an indicator of how much emphasis is placed on that particular factor for that particular segment. For example, "Price" is very important in the selection of a Workgroup or an Appliance server, but not quite as important for a Super-Enterprise server. Similarly, "High Reliability" is very important in the Enterprise and Super-Enterprise market space, but less of a consideration for an Appliance server.

Factor Definitions :

Performance/Power: Computational (or similar) power, the ability to perform a large number of operations or to handle a large workload.

High Reliability: The ability to run without a non-recoverable failure for long periods of time. (Note that "reliability" is a goal for all classes of server. The term "high reliability" is used because different server classes require different levels of reliability.

Price: Base system price, or price with a modest complement of features added. Total Cost of Ownership (TCO): An amalgam of fully-configured system price, installation/setup/training costs, infrastructure costs, downtime costs, and other factors.

Flexibility: The ability of the server to be reconfigured for disparate tasks/functionality, as needed by the user. Service/support: Related to availability and reliability, maintainability, and ease-of-use considerations.

High CPU qty.: Whether a server can support four, eight, or more CPUs

Storage (internal): The storage capacity, both in terms of raw space (i.e. how many gigabytes) and number of disk drives, a server can support inside the chassis.

Redundancy: Whether a given component (e.g. fan, power supply, network interface card, disk drive) has a backup and a failover mechanism. Redundancy is a method for improving system reliability/availability.

Hot Swap/Maintainability: The ability to remove and replace a component (e.g. fan, disk drive, power supply) without having to shut down the system or cause processing functions to stop.

Outlook

As the server market consolidates, market/product segment distinctions - especially at the low end - will blur. We believe that within 12-24 months, there will be only three size-related segments (Entry/Workgroup, Department, and Enterprise) plus the general class of Appliance servers. As the number of segments decreases, the distinction between them will increase. However, certain patterns will be maintained, such as Enterprise Servers needing to be extremely reliable. Customers should keep these factors in mind as they move toward future server purchases.

 

 

 
comments powered by Disqus

Recent Searches
Others A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

©2014 Technology Evaluation Centers Inc. All rights reserved.