By providing a solid base of Internet services, especially in the realm of e-commerce and Web applications, Windows 2000 provides an advanced and more developer friendly solution than Solaris 8.
Solaris 8 has strong Java features, including support for Java servlets, but its lack of message queue services, transaction processing, and all-round integration between the OS and Internet service means that it falls short of the requirements of a modern Internet-aware operating system.
One of the greatest strengths in Windows 2000 is the integration of the OS and the IIS 5.0 Internet Service component. The OS also provides the best reliability and scalability across an Internet server farm because Windows 2000 can distribute requests based on the resource load of individual servers. Using Windows 2000, it’s possible to maximize performance both in terms of servicing a client’s request and internally through the use of distributed development tools such as DCOM, MTS, and MSMQ.
Reliability
In order to ensure the maximum reliability of a system, features must be incorporated into every component of the operating system. The OS should be able to tolerate faults and monitor the health of a server so that it can predict when faults may occur and take action to rectify the problem.
At a core level, the operating system must be able to protect itself from the effects of the applications it supports. When an application hangs or crashes it should not cause the entire system to fail. This requires the OS to employ techniques to ensure that individual applications cannot affect each other or the OS itself.
For superior reliability, the operating system must be able to support fault tolerant devices to allow automatic fail-over when a piece of hardware fails. This includes support for RAID systems to allow for recovery from disk failures, and clustering to provide resilience in the event of a complete hardware failure. Other fault tolerant services include UPS support, which must be monitored so that a system can be safely shutdown in the event of a power failure to prevent data corruption.
To minimize downtime, various mechanisms help improve the recovery time once a fault has been identified or after faults have been rectified. These include tools for file system recovery and the backup and recovery systems. For diagnostic purposes, the system also needs to be able to log its current status and provide the tools to examine the logs so that the origin of the fault can be traced.
Although unplanned outages present the most serious problems to providing high reliability, planned downtime can also be a factor. The ability to dynamically reconfigure the OS without needing to reboot the system will help to improve the availability and therefore the perceived reliability of the system.
Core System Reliability
Both Solaris 8 and Windows 2000 provide extended facilities for ensuring that individual applications do not affect the operation of the operating system. Both operating systems support protected kernels and protected memory areas for individual applications. If an application fails due to a memory access failure, the crash will only affect the allotted area and not the entire operating system.
S
More Information
|
See the Hardware Support section earlier in this document for more details on the fault-tolerant support offered by Solaris 8 and Windows 2000.
|
olaris 8 incorporates a monitoring capability tailored for the SPARC hardware platform, but not for Intel hardware. In supported systems, this allows for components to be swapped while the machine is running.
For greater resilience, the Sun Cluster 2.2 allows up to four machines to be configured as a single cluster. If one machine within the cluster fails, the other machines within the cluster automatically take over the processes and services previously supported by the failed device. The clustering facility is provided by Solaris only on Sun SPARC hardware and requires specialized hardware to link individual nodes.
Solaris 8 includes Solaris DiskSuite—a software-based solution for disk management that provides RAID features in software without the need for special RAID hardware. Configuration of the disks is handled in real time and for most operations does not require a reboot for changes to take affect.
Windows 2000 Advanced Server and Datacenter Server support clustering of two and four nodes, respectively, to provide system-wide redundancy for an individual cluster. Unlike Solaris, which requires using special hardware to enable the clustering technology, Windows 2000 supports clustering on standard SCSI-equipped server hardware, lowering the cost for providing fault tolerant clusters. Advanced Server and Datacenter Server systems supporting clustering are tested under the HCT as detailed earlier in the hardware support section.
Windows 2000 includes built-in support for software-based RAID solutions and extensive support for many RAID controllers and RAID systems including those built into many data center servers. All disks and RAID systems are configured using the same Disk Administrator tool, providing a consistent interface to managing data storage.
In addition, Windows 2000 incorporates support for monitoring the system environment on suitable hardware and integrates with many UPS systems to support advanced reliability in the event of power failure.
Failure Recovery
The most critical task to be carried out after a failure is to verify the integrity of information on permanent storage. Solaris 8 includes the UNIX fsck utility, which enables the OS to repair a file system during a system reboot. However, recovery is not always perfect and it is possible to lose files and data. For fault-tolerant file systems it’s possible to use the third party Veritas file system, which supports journaling and writes file system changes to a log that can then be recovered during a reboot. This is similar functionality to that available natively with the Windows 2000 NTFS file system. Unlike NTFS, the Veritas system is only supported for data storage volumes; it cannot be used with the OS boot volume.
The Windows 2000 NTFS file system is highly tolerant of disk failures. It employs a transaction logging technique to ensure that all requests are recorded on the disk. In the event of a system failure, the file system can process the pending changes recorded in the log to return the file system to the operational state prior to the machine failure.
If the operating system fails, Solaris 8 automatically writes a log of the failure to disk for later study. The machine then automatically reboots to its normal operating state. On Sun’s Enterprise Server systems, Solaris 8 also supports Automatic System Recovery (ASR). ASR monitors the components within a Sun server and when a component failure occurs, it automatically reboots the system. During the reboot process, the failed component is then ignored or disabled by the operating system, which then continues to boot up as normal. This allows for completely unattended operation—the machine will restart even if a critical component within the machine is faulty.
For more serious problems, Solaris also supports an interactive environment for repairing the operating system. However, effective repair requires extensive knowledge of the underlying file system and the tools available for repair. For non-technical users this is a daunting interface and even with UNIX experience competent repair requires a high level of expertise.
Windows 2000 also supports an automatic restart option with the ability to save the memory contents to disk so that the problem can be traced. To aid in the recovery process, Windows 2000 includes the Repair Command Console (RCC), which allows you to manage NTFS volumes from floppy disks, or from a bootable CD-ROM. The RCC allows recovery of the system even in the event of file system failure on the boot disk using the Windows Scandisk and management tools. In addition, you can boot a server running Windows 2000 into “Safe Mode” to allow for system reconfiguration without loading additional drivers and services.
To ensure the active status of a system, Windows 2000 allows you to store the current configuration for the entire server in a separate location. The configuration can then be recovered to reset the system back to its original state. This is useful not only in the event of system failure, but also when installing new device drivers and hardware that may upset the system. Solaris 8 does not offer this capability. And it would be impossible to implement such a system because the configuration information is spread across a number of files and locations.
Solaris 8 includes a basic tape backup system, but is neither reliable nor practical for most server installations. The ufsdump utility requires experience of the command line to be able to use it effectively and is not guaranteed to restore your system perfectly. To recover a system after failure you must boot from CD and re-install the base operating system before restoring the files stored on tape. This increases the recovery time significantly.
There are other solutions available for Solaris 8 backups, such as Legato NetWorker, but this is yet another application that sits on top of the operating system. Legato provides certain levels of integration with other application software (Oracle, Sybase, Veritas), and facilities for backing up entire networks and many different operating systems. However, it remains a separate component, with a separate management interface that serves to increase costs—both in software and staff time.
Windows 2000 includes a powerful backup application that can backup files to tape based on a number of different criteria. Windows Backup also supports powerful scheduling features. For recovery, you can create a recovery boot disk that can be used to boot the system and then recover files from tape to recover your system in the shortest possible time. The backup is also integrated into the rest of the operating system allowing for safe backup and recovery of Active Directory and Microsoft Exchange.
A summary of the standard system recovery features offered by each operating system is shown in Table 7 below.
Table 7: System Recovery Features
|
Feature
|
Solaris 8
|
Windows 2000
|
File System Recovery Tools
|
Y
|
Y
|
Fault-Tolerant File System
|
N
|
Y
|
Crash Logs
|
Y
|
Y
|
Kernel Dumps
|
Y
|
Y
|
Automatic System Recovery
|
Y
|
N
|
Interactive System Recovery
|
Y
|
Y
|
CD/Floppy Boot Recovery
|
Y
|
Y
|
Configuration Backup/Recovery
|
N
|
Y
|
Simple Tape Backup System
|
Y
|
Y
|
Enterprise Tape Backup System
|
N
|
Y
|
| 