Tài liệu Installing Windows 2000 Server pptx

Application server installation recipe Microsoft now uses the term application server to refer to a server running Terminal Services, but a server can still host applications without ter

Installing Windows 2000 Server

This chapter reviews the installation of Windows 2000

Server It discusses a number of hardware configurationsand setup options, and reviews potential obstacles

Installation and Configuration Strategy

If you have read Chapter 4 and have done your homework, youwill now be ready to begin installing Windows 2000 in your lab

You may be tempted, or you may have an urgent need, to godirectly to a working or production system, in a productionenvironment Perhaps your DHCP server died, or a new DNSserver is needed urgently, and so on Resist, or stick with whatyou know If you have an NT network and need to raise a newservice to fill an urgent need, stick with Windows NT

On the other hand, if you are a seasoned administrator andyou know what you’re doing, you will probably have items like

a hardware checklist, remote or unattended installation, hotstandby, and so on, well taken care of So only go directly to aproduction system if you know what you are doing and theproduction system is part of a conversion and rollout project

For the most part, you should always raise servers in a lab

Then you should burn them in (run them continually) for about

a week; hit them with work for at least another week After that,and if all test items check off, then ship or go live But no twoenvironments are the same Let’s look at the various installa-tion and configuration situations and then go from there

ConfigurationInstalling from Boot DisksInstalling from theCommand Line Installing over the NetworkThe Boot File

A lot of people ask how you burn in a server that is standing idle and has no usersconnected to it One simple way is to set up NTBackup to run continually Runningbackup is great physical therapy for a server It works the hard disks, memory, sys-tem buses, access control, permissions and the NTFS, remote and removable stor-age functions, and more You can also configure NTBackup (or any other backuputility for that matter) to perform both pre- and post-backup routines, such as send-ing alerts, moving files around the house, and so on Depending on your stress test,you may need to write a backup script to automatically overwrite media, and so on.See Chapter 17 for more information And if you want to test disk I/O and otherroutines, you might have to write some custom software for the job.

Getting Psyched Up about Installing

This chapter takes you through the basic install routines, and then to rollout andsophisticated deployment strategy We are going to help you cook up a variety ofserver meals Microsoft has spent many millions on the installation and configura-tion process So, for the most part, Windows 2000 rises well for the power that itwields It is certainly a lot smoother and friendlier to install than any other serveroperating system in existence (other than the machine you receive pre-installedfrom the factory)

We have installed the operating system more times than you care to know and onabout ten different platforms with a variety of hardware from scrap piles to brandnames We have also deliberately sabotaged our systems (like taking away drives,drivers, memory, and certain system files) and tried a variety of recovery tech-niques What we have to report to you is as follows: If you experience any difficultyinstalling Windows 2000 Server, you must be using very unconventional methods,thrift store hardware, or not paying attention to details and recommended strategy.Now sit back, close your eyes, and imagine you are in a class going through installa-tion training You’ll feel good thinking that you spent no more than the cost of a nicedinner on this book, and did not have to mortgage your house for a five-day course

Server Recipes

In evaluating the various needs in the enterprise, we classified our installation intovarious recipes of server installation, which are discussed in the following sections

Low-road or bare-bones system recipe

This option consists of using minimum hardware requirements to raise the server.You can use a no-frills motherboard with one CPU, the minimum RAM (64MB), a sin-gle IDE hard disk drive, a CD-ROM, a 1.4MB floppy disk drive, a standard networkcard, and a mouse, keyboard, and monitor (MKM)

Note

Microsoft now ships the Server OS and the upgrade version with a note that128MB is recommended However, the installation will still get the green light ifyou only have 64MB.

We have raised servers (both Server and Advanced Server) on CPUs ranging fromold Pentium 133s, 166s, 200s, Pro 200s, and 266s, to Pentium II and III 300s, 450s,500s, 700s, duals (two CPUs), quads, and so on You can raise a Windows 2000server on 133MHz and 166Mhz Pentiums, but we don’t recommend it for anythingmore than the smallest role server, discussed later On the other hand, an oldPentium Pro with a lot of RAM will serve many of your needs You can usually pickthese servers up on the Internet for a song, and if they are good brands, they’ll dowell for many years

Small file and print server recipe

A small file and print server caters to small business’ file and print services Youshould still use the bare-bones components, add a second large IDE hard disk drive

for file and print services, the usual peripherals, and so on The amount of RAM

will depend on the number of connections and users Printing services will require

a lot more RAM than file services If you are supporting a lot of printers, then go to128MB of RAM and stick with no less than a Pentium Pro (a PII or PIII of 300Mhz andhigher is better)

Don’t be a chump and install 32MB RAM modules (unless they are lying in your topdrawer gathering dust) In fact, it is almost impossible to buy anything less than64MB nowadays Check out the chapters on File Systems and Printer Services for fur-ther advice on hardware requirements (Chapters 21 and Chapter 23, respectively)

Your hard disk demands will be higher, and you should now consider adding a ond drive You can stick to a cheap IDE disk (even the cheap IDE or EIDE drives aregood disks) or begin thinking about SCSI But let’s hold the thought about harddisks for later on in this chapter

sec-You may have read information elsewhere calling for more firepower in aWindows 2000 Server We would preempt that question with the answer that ourassessment is based on various experiments, projects, pilot systems, and deploy-ment Every situation is different, and the only way to really know what you need

to throw at a situation is to test

Application server installation recipe

Microsoft now uses the term application server to refer to a server running Terminal

Services, but a server can still host applications without terminal users You maywant to install applications on servers for users who will load them into local mem-ory at their workstations The application is thus loaded across the network, butthe “footprint” and ensuing resource consumption is local to the user’s hardware

Note Tip

You may also have applications that are server-based or server-oriented Thesemay include database front-ends, communications software, processing-orientedsoftware, and network management applications There may be hundreds of appli-cations that are suited to server-side execution and that need no user interaction,such as process control applications and data processing.

You could use the recipe for file and print servers; it will take some testing to raisethe ideal configuration for your purpose Depending on the availability require-ments, you might need to add RAID, hot-insert drive-bays, and so on, which are discussed later in this chapter

Terminal Services installation recipe

A terminal service application server is a whole new ball game The WinFramelicensing arrangement between Citrix Systems, Inc., and Microsoft (see Chapter 25)was the origin of Terminal Services Terminal server, under the Hydra project name,first made its debut in Windows NT 4.0 in late 1997 It was then launched as a sepa-rate NT 4.0 operating system called Windows NT 4.0 Terminal Server Edition (TSE).Terminal server is no longer a separate product under Windows 2000 Server All servers come with Terminal Services built into the OS (see Chapter 1), and,

as described in Chapter 25, you either configure Terminal Services as a remoteAdministration tool (which does not require licensing) or as an Application Serverservice (which does require licensing) This computing model is known as thin-client/server computing

With Windows 2000 acting as a Terminal Service application server, all your usersrun all their applications on the server There is no such thing as a local TerminalService client The client can be a browser, a fat client running a Terminal Serviceterminal application (like a TN3270 character-based terminal running on Windowsand accessing a DB2 database on the mainframe), a dumb terminal (known asWindows-based terminals), or terminals running on the Windows CE or Pocket PCplatforms Your users’ terminals can also be installed on any non-Windows platform,such as Macintosh, DOS, and UNIX, but these require the MetaFrame product suitefrom Citrix, which uses the ICA protocol

Terminal servers can be raised with any of the recipes discussed so far However, it

is not what you start up with, but what the terminal users do when they are attached

to the server that matters We have tested these services and deployed them in orous real-life situations since 1997, and the following configuration pointers, whichapply to a different configuration recipe we will shortly discuss, are key:

vig-✦ Restrict your users from having more than four applications open at a time.For example, make sure they can comfortably open and run a database appli-cation, a word processing application, e-mail, and a Web browser

✦ Configure the applications to run without fancy splash screens, animations,

or any resource-intensive software This may not always be possible (seeChapter 25 for further information)

✦ Assign and enforce hard disk quotas This is important to do for all users, butespecially useful when you are dealing with terminal users

A server hosting no more than five terminal users should be running on a CPU of

no less than 300MHz Each user (depending on the applications and the type of cessing) should be assigned no less than 32MB of RAM You should also install fastSCSI drives and support them in hardware RAID configurations on fast controllercards In short, there is no bare-bones situation when it comes to Terminal Servicesand application hosting After all, if you were deploying to standard clients, theywould likely each be above 266MHz with 32MB or more of RAM

pro-At 32MB each, the recipe thus calls for the following total server RAM:

✦ Operating system = 128MB

✦ Five users @ 32MB each = 160MB

✦ Total RAM needed = 288MBYou will probably have a hard time adding 32MB modules into a modern mother-board Your configuration would thus be two 128MB modules and one 64MB mod-ule, or a single 320MB or larger RAM module

We have actually succeeded with less RAM, and you could count on a 300MHz system with 128MB RAM and a couple of fast IDE drives to service three to fiveusers But you should know that this will work only if you can guarantee that theusers keep no more than two apps open (say their e-mail and one work app, like adatabase front-end) This latter “easier” configuration should be your bare-bonesrecipe for Terminal Services (to give users a reasonable work environment)

We talk about which applications work well on Terminal Services in Chapter 25

Role server installation recipe

Role servers are servers running services like DHCP, WINS, DNS, and ActiveDirectory Your application and needs may vary widely, depending on the serviceand how many subscribers it has A small company might get away with a light-weight configuration, like the small file and print server recipe offered earlier Inother cases, you may require much more firepower, especially on medium to largeintranets For the record, we’ve been running DHCP, WINS, and DNS on Windows NTProfessional on Pentium 200s with 128MB of RAM in each, servicing several thou-sand users over a nationwide WAN for several years But there is a lot more replica-tion and dynamic configuration overhead with Windows 2000, and you might have

to shell out for a Pentium II or III machine

Cross-Reference

BackOffice, high-road, or mission-critical recipe

Mission-critical servers should have no less than 300MHz in CPU capability For themost part, and if you have more than a handful of users, your CPU should be morethan 400MHz You might consider equipment running two-CPU configurations, orpossibly deploy quad systems

Hard disk needs may vary, but you’ll need to configure a second drive letter ning at RAID 5 under hardware control (in case you’re wondering, these are SCSIdevices, which we discuss later)

run-Redundant or stand-by system recipe

Any of the server recipes mentioned so far could be cloned to provide an offline

or hot spare These are obviously not cluster or automatic fail-over machines If the primary server goes down, you could pull dynamic volumes out of the primaryarrays and install them into the hot spares But a better solution, if you can afford itand have the budget, is to install Advanced Server and run cluster services and net-work load balancing

Large systems, clusters, and Datacenter server installations

Advanced clustering (high availability) and Datacenter Server solutions are beyondthe scope of this book, although most of the configuration information in this bookwill apply to the high-end operating systems As discussed in Chapter 21, there isnot that much difference between the operating systems, at least not from the aver-age user’s point of view Any large system will call for an external SCSI-based stor-age silo under hardware RAID-5

The various recipes we’ve discussed so far are summarized in Table 5-1

Table 5-1

Hardware Guide for Server Recipes

Recipe CPU/MHz RAM/MB HDD

Bare-Bones 200 64+ IDE Small F&P 200-300 64+ IDE App Server 300+ 64+ IDE/SCSI Terminal Service 300+ 300+ SCSI-RAID Role Server 266+ 96+ SCSI-RAID BackOffice 300+ 128+ SCSI-RAID Standby 300+ 128+ SCSI/IDE Large 450+ 300+ SCSI-RAID

Choosing hardware is not a difficult exercise at all for Windows 2000 Server There

is really not a lot that you will put into your system The list of hardware we will cuss is as follows:

✦ Network interface cards (NICs)

The Hardware Compatibility List (HCL)

Before you go buying parts, review the Hardware Compatibility List (HCL) in the

\supportfolder on the Server CD, for your own peace of mind If your part is notlisted, check out the HCL on the Microsoft Web site at www.microsoft.com/hcl.You will probably find the HCL a little amusing because Windows 2000 has alreadybeen available to the leading manufacturers for more than a year as of this writing,and they have a ways to go before their products will be logo-compliant Some pop-ular brands are not even listed as compatible, and some have even stated that theywill not go for the logo that in-house tests are good enough

Also, according to Microsoft policy, Microsoft will not support you if the item is not on the HCL, but not many items are on the HCL yet And if you spend $195 withMicrosoft to figure out if hardware is the reason a server will not start, will theyrefuse to take your money? They never have to date Microsoft’s paid support team

is very responsive and will help you determine if hardware is your problem At least

if they tell you that you have a hardware compatibility problem, that’s probably allthe advice you’ll need

The HCL aside, you should heed the following advice: Most large companies willbuy brands from the likes of IBM, Compaq, Dell, or HP, and so on And if the budget

is there, a small company looking for one business server should go this route aswell The servers will be burned in and tested, and the manufacturer will standbehind the compliance of its product running Windows 2000 Server, logo-compliant

or not The servers will also come with warranties and various levels of support,and you can get as much as a 30 percent discount on the operating system, whichcomes preinstalled

If, however, you plan to build your own server, or if you need to upgrade a machinedown the road, then by all means buy your own parts and knock together your ownserver However, for best motherboard results, try to stick to made-in-America com-ponents, or well-known and popular foreign imports For RAM, there is only a handful

of factories left, but you’ll be okay buying products from the likes of NEC, Compaq,IBM, TI, and others For hard disks, IBM, Quantum, and Seagate are the leaders nowand may soon be the only players For CPUs, there’s Intel, AMD, and Cyrix If you arethinking PowerPC and other marginal CPUs, you’ll need to talk to the likes of IBM orMotorola The other peripherals will not interfere with your server

Motherboards

Motherboards come in a variety of sizes and shapes The essential components of aserver motherboard are as follows:

✦ Mother form factor: Motherboards come in several sizes: AT (the form factor

of most server boards), ATX, BabyATX, and MicroATX BabyATX and MicroATXare aimed at the home and have fewer slots than you would want for a servermotherboard Go with AT or ATX

✦ Slots: Slots come in three standards: ISA (the older and slower slots), PCI

(which caters to faster data transfer rates), and AGP (Advanced GraphicsPort), which is more suited to graphics components (not becoming of aserver) Most motherboards include all three of the slot types Since AGP isfor a graphics interface card, there is usually only one AGP slot Choose amotherboard that gives you about two ISA slots and four to five PCI slots

✦ RAM slots: The RAM slots include SIMMs and DIMMs SIMM slots are the older

72-pin slots, and the modules have to be mounted in pairs DIMM memory ismuch faster DIMM modules come in 168-pin slots, and the memory can bemounted as single modules You can put more DIMM RAM in a server thanSIMM, which is important for future expansion

✦ CPU Sockets: The CPU sockets include Socket 7, Slot 1, and 370/PPGA Socket 7

is the older Pentium Pro-type socket, which is inserted like a pancake into themotherboard sockets Slot 1 CPUs are for the new Pentium II and III CPUs, whichare inserted into a single slot and protrude away from the motherboard Slot

370 CPUs are cheaper than Slot 1 CPUs and are for the Intel Celeron PPGA CPUs.One of the top motherboards in the United States is SuperMicro, which suppliesmany leading brands You can buy SuperMicro boards at Motherboards.com, whichsells several other leading brands, including Soyo Soyo is gaining market share and

is becoming very popular with places like Best Buy and CompUSA Another board maker that has become popular is Tekram (www.tekram.com)

mother-Central Processing Units (CPUs)

The leading CPU maker is still Intel However, AMD and Cyrix are making their own

noises in the x86 market Contact all three manufacturers for Windows 2000 Server

compatibility, or check out the HCL under CPUs

Memory

As mentioned earlier, stick to DIMM slots, which are 168-pin and are much fasterthan SIMMs; besides, you will not find a newly manufactured motherboard for aserver that comes with SIMM slots DIMM RAM sizes start at 16MB, but it is gettingharder to find modules under 64MB Because Windows 2000 Server requires at least64MB, there is no need to look around for smaller modules

Hard Disk Drives

The biggest names in Hard Disk Drives (HDDs) are Seagate, IBM, Fujitsu, Quantum,Western Digital, Maxtor, and Hewlett Packard, in no particular order For small orbare-bone servers, you can escape with IDE or even Enhanced IDE (EIDE) drives

Anything more — supporting multiple users or high-end applications, TerminalServices, and BackOffice or high-end role servers — demands SCSI hard disk drives

SCSI comes in several flavors: SCSI, SCSI-2, SCSI-3, and Ultra SCSI are the base cols Several years ago, it was thought that SCSI would go the way of the dodo, but

proto-it continues to surge, and new SCSI standards are emerging all the time

The advantage of SCSI over IDE or EIDE lies in several factors:

✦ Speed: SCSI drives are much faster in access time and transfer rate than IDE

drives; however, new IDE drives are being introduced every month that perform recent additions to the SCSI lineup

out-✦ Capacity: SCSI drives are currently available from 9.5GB to 72GB (although

this will be out of date by the time you reach the end of the chapter, and IDEtechnology continues to surprise us)

* Addressing: Many drives or devices can be chained on a single cable You

can currently address up to 15 SCSI devices with the Ultra SCSI standard

✦ Support: There is more supporting technology available for SCSI that is

targeted to server solutions These include high-end RAID controllers, hotinsert hardware, storage silos, and drive array enclosures

In addition to SCSI, a new standard is beginning to take hold and is showing ing promise It is called Fibre Channel Arbitrated Loop or FC-AL FC-AL drives areconnected by coax cable The drives are incredibly fast, and you can currentlyaddress up to 126 of them, as opposed to 15 under SCSI Table 5-2 lists the differ-ences between FC-AL and SCSI

amaz-Table 5-2

SCSI versus FC-AL

Specification SCSI FC-AL

Cable distances 1.5-3 meters 30 meters between devices Data rates 5MB per second 200MB per second Addressing 15 devices 126 devices Array support Parity Hot insert support with no special controller,

dual porting, CRC for integrity

In addition to the differences listed in Table 5-2, FC-AL is extremely tidy in son to SCSI; the devices have a single port for a coaxial connection, and no otherconfiguration is required SCSI, on the other hand, requires ribbon cable, jumpers,and terminators

compari-The downside of using FC-AL is that they are a little more expensive than SCSIdrives and have not been extensively tested with Windows 2000 (nor have the largeSCSI drivers, for that matter) Both standards in large capacity are worth experi-menting with; after all, hard disk prices are dropping everyday At the current rate,

a 96GB HDD is expected to cost about $300 by 2002

Hot insert or hot swap is also an important consideration for server class machines.Your servers should be configured with hot swap drive hardware that will allow you

to remove a dead drive while a system is hot and online This allows you to replacedrives without bringing mission-critical servers down

Hot swap standards come in four levels, as illustrated in Table 5-3

Table 5-3

Hot Swap Levels

Level Capability

Level 1 Cold swap System is offline; no power is applied to the drive.

Level 2 Hot swap reset System is held in reset state; power is applied.

Level 3 Hot swap on an idle bus Power is applied.

Level 4 Hot swap on an active bus Power is applied to the drive.

Mission critical or maximum availability servers should be configured with Level 4hot swap capability.

HDD Controllers

The standard HDD controllers that are built onto most motherboards suffice formost server and data processing needs Small business systems configured withIDE or EIDE cards will work well with onboard controllers (that is, on the mother-board) or your vanilla controller that sells for under $50

SCSI drives need SCSI controllers These can range from your standard SCSI trollers to faster cards and cards that support mirroring, disk duplexing, disk arrays,and so on One of the best-known names is Adaptec Most branded computers thatare configured for SCSI usually make their own SCSI controllers or use Adaptec

con-A good SCSI controller is of paramount importance, because it makes no sense toinstall a fast SCSI drive or drive array and then go cheap on controller cards

Network Interface Cards

Any seasoned network administrator will tell you that you can buy a $10 networkinterface card (NIC) and spend hundreds of dollars trying to get it to work, or youcan spend $75 to $100 and have it installed and bound in less than five minutes

Stick with the brand name cards like 3Com and Madge Most of 3Com’s productshave been tested compatible with Windows 2000 There is zero setup effort withany of the latest 3Com NICs, and as long as you install any of 3Com’s 900 series(such as the 3C905), you won’t have any hassle

Plug and Play

Plug and Play (PnP) has arrived on Windows 2000 Server This technology makesinstalling devices far less painful than was the case on Windows NT, which did notsupport PnP For the most part, the operating system will be able to detect yournew components and automatically configure them for operation You may onlyhave to provide addressing or name configuration We will look at PnP and theDevice Manager in the Post Installation section later in this chapter

Getting Ready to Install

Before installation, you should prepare a checklist detailing what you are going toinstall and the items you need to have handy The following checklists cover sev-eral types of installation

Standalone Servers

Standalone servers do not connect to any domain, but rather to a workgroup Youcan create a workgroup from one standalone server or join the server to anotherworkgroup, Windows for Workgroups-style You can also join a domain post installa-tion For a standalone server, you need the following items:

✦ Workgroup name

✦ An administrator’s password

✦ Network protocols

✦ IP address

✦ DNS IP addresses and host names

✦ NetBIOS name of host

✦ DNS IP addresses and host names

✦ NetBIOS name of host

✦ DNS IP addresses and host names

✦ NetBIOS name of host

✦ Role service information

According to our definition, a member server can be a role server only if you installspecific role services onto the server during installation, such as WINS, DNS, DHCP,and so on.

Domain Controller

There are two approaches to installing a domain controller First, you can raise themachine as a member server and promote it post-installation, and even post-burn-in

Or you can promote it to domain controller status during an automated installation

We recommend against the latter option, unless you are really confident about yourmachines and their configuration If you are an Original Equipment Manufacturer(OEM), you would not need to be concerned about domain controllers and ActiveDirectory because the domain specifics, such as creating a new tree or forest, orjoining existing trees and forests, is something that gets done on the customer’snetwork On the other hand, if you, as a consultant or network engineer, have cre-ated an extensive unattended or remote installation regimen that automaticallyraises the machine as a domain controller, then you know what you are doing Wewill look into this a little later

For now, there are several reasons for not promoting during or just after initialinstallation First, promoting a domain controller is a time-intensive operation(Active Directory goes through extensive self-configuration before the installationcompletes) Second, if there is a problem with the machine, you will have to demotethe domain controller, which can be a complicated process And third, once youhave installed and raised a domain controller, you do not want to demote it due to

a hardware problem, or risk trashing your domain controller

The faster the CPU and the more RAM you have, the quicker you can install andconfigure Active Directory On a new server with a 500MHz CPU, 256MB RAM, asuper-fast motherboard, and lightning-fast SCSI drives, we were able to installActive Directory, start-to-finish, in under five minutes On lesser machines, it took

as long as 45 minutes

When Active Directory is demoted, it tears down everything it created and restoresthe machine to the control of the registry and the local SAM In fact, it is like watch-ing a movie going backwards Active Directory asks you for a new Administratoraccount name and password for the rollback All configuration changes made to themachine, such as desktop settings, are restored to the default, newly created set-tings When you reboot the machine, you are back to where you started You willnot even get earlier changes you made to the registry because the registry is essen-tially reinstalled when Active Directory comes down (because it is wiped out whenyou promote the server)

There is a good reason for this Everything configured on a domain controller

is stored in the directory databases, and after the registry is restored, you can re-promote it from scratch Promoting a domain controller is dealt with in Chapter 9

Tip Note

Burn in a domain controller machine for several weeks, if possible, before you mote and deploy it A domain controller running Active Directory for several weekshas already accumulated an extensive configuration, not a good idea for a machinethat has to be replaced.

pro-The checklist for a domain controller is as follows:

✦ Domain name If you are creating a new domain, you will need the name of theparent domain you are installing under, or the existing tree name (or the for-est name if you are installing a new domain tree) If you are adding a domaincontroller to an existing domain, you need to have that name handy as well

✦ An administrator’s password

✦ Network protocols

✦ IP address

✦ NetBIOS name of host

✦ DNS IP addresses and host names

In all of the previous checklists, be certain of the NetBIOS name you are going touse Once you have supplied this name, it cannot be changed short of reinstallingthe machine, if it is still a standalone, member, or role server If you need to changethe NetBIOS name of a domain controller, you will have to reinstall Active Directory

We talk about naming again in the next section

Installing Windows 2000 Server

We have found, after dozens of installations, that the best practice for installingWindows 2000 Server is to raise a bare-bones operating system and then configure

or promote the machine after it has been running and burning in as a standalone

or member server for a while Not only is the installation far quicker, but also thispractice allows you to get a minimal server going and then install services andapplications into a running system after it has met your specific quality levels.The only service that should be configured during installation is Terminal Services.Raising a server with this service running in remote administration mode will allowyou to connect to a remotely installed server for configuration and promotion Thenext few sections will take you through the bare-bones installation

Partitioning Hard Disk Drives

Give Windows 2000 a hand, and it will take an arm or at least another drive.Installation will assess all the hard drive resources in the system, and if there aretwo drives (or partitions), the OS will attempt to use both The first active partition

Tip

will get snagged for the system files the minimum required to raise the system

to a point where you can run recovery tools, or the Recovery Console Windows

2000 calls this volume — you guessed it — the system volume.

You can control how installation deals with hard disks and partitions by usingcommand-line switches and parameters Some of these are discussed later in thischapter and as part of unattended installation in the Windows 2000 Resource Kit

in Appendix B

Windows 2000 will then snag a second drive or partition and use it for the boot files the files needed to boot the rest of the operating system all the way to the

desktop where you can log in Windows 2000 calls this volume the Boot volume.

(This is a reversal of the old naming convention for boot and system partitions.)There are two reasons for the dual disk consumption First, Windows 2000 is opti-mized to use more than one hard disk drive Second, a minimum boot disk can beconfigured to hold just the boot files and be formatted as FAT or FAT 32, instead ofNTFS The theory is this: If you lose the base operating system — that is, if you areunable to boot to the desktop — you can at least boot to a DOS diskette and thenfrom DOS copy new base files over the corrupt ones (or replace a defective drive)

Many NT and NetWare systems have been configured this way But a well-designedand managed system need not retain a FAT boot disk, which, due to its poor secu-rity, is a risk to the entire system because it does not support file-level security

However, Windows 2000 Server will allow you to boot to the Boot Options console(when it detects a disaster) Here you have several options, such as Safe Mode withNetworking, and from there you can attempt to boot without certain services anddebug the problem once you have the OS up and running You can also boot theRecovery Mode Console, which takes you to a command line that you can use toaccess NTFS partitions (see Chapter 17 and Appendix B) and the Boot Disks So, the idea of leaving boot or system files on FAT volumes is old fashioned, the result

of bad memories from Windows NT days We recommend the following partitionarrangement options

Option 1: One HDD

The following arrangement uses one hard disk drive, which forces Windows 2000 toput both boot files and system files onto the same drive and partition

1 Configure the system with one hard disk drive of about 2GB in size

(Microsoft’s official recommendation is to supply at least a 1GB partition, butwith service packs and new features coming down the road, you need to leaveroom for expansion)

2 Format the partition during the install as NTFS (see Chapter 21).

3 Let Windows 2000 choose the partition name, the default.

Note