| Managing the Raging Pro
Manageability starts the instant you power-on the server. For example, servers from Compaq can boot from a configuration CD-ROM. Similarly, on a CD-ROM or via the EISA configuration disks, servers from HP and Compaq let you create a hidden EISA configuration partition. This lets administrators add and remove hardware from the server without having to
maintain customized EISA floppy disks. Once the hardware is configured, you can run the installation programs for various operating systems. The NOS software can be purchased from server vendors such as IBM, HP and Compaq. The server installation programs automatically let you choose the OS you want to install.
Once your server is up and running, server management software really comes in handy. The vendor's own and perhaps proprietary server management software can give you detailed information on all the vendor components inside the server and can perform predictive failure analysis for you. It is wise to take a proactive approach to heading off problems. For example, the Compaq Insight Manager (CIM) utility runs on a workstation and collects data from your servers. The data indicates potential drive and memory failures and lets you
browse through the correctable error logs maintained in the BIOS. If an error occurs, you can receive it via a Simple Network Management Protocol (SNMP) agent or by pager.
If your operating system freezes, the error is logged and the server is restarted--automatically. Additionally, the software can list the various ROM levels on your servers and indicate whether a flash ROM upgrade is necessary. Vendors release firmware upgrades every few months or so. The upgrades generally are necessary to support newer peripherals such as the ability to recognize a 4.3-GB drive in an older system and to fix firmware code bugs.
Beware of third-party management software. Many third-party tools, in their design to work with any server, often will not report details unique to a component. Since there is no standard where this aspect of manageability is concerned, if a server vendor does not include its own management software, you probably won't get the full benefits.
If your company has a central management console, you may want to check if the vendors' management software allows you to send SNMP-based alerts to that console. This will let you tie your server into the alerting processes
that have been established. Other options are flash firmware and remote management hardware. The former allows you to fix bugs or upgrade the code on components to support newer peripherals without having to replace chips; simply boot up with the firmware disk, and your components are upgraded. The latter lets you dial into a crashed server via a modem or the network on a self-powered card. Retailing for about $1,500, these cards help you to analyze the operating system startup logs, reboot the server and perform diagnostics. Intel makes a card that will work in any server.
A Server's Life Cycle
Scalability of a server's hardware design determines how long the server can be useful. If the processor unit is on a daughterboard, then check its upgrade path. Can you replace the 166-MHz CPU with a 200-MHz Pro CPU? This feature will get
you to the next level of processing power without having to replace the entire server.
If you anticipate rapid growth or want to make sure you can scale up years fr
om now, make sure there are enough drive bays for expansion. If the server comes equipped with a SCSI controller, check if it has more than one channel and that it can handle all RAID levels. If you run out of internal slots, you can add an external drive-array cabinet. For instance, the dual channel Smart-2 controller from Compaq (which only works in Compaq servers) not only handles all RAID levels, but also allows for online expansion of a RAID set. Normally, once you have a server running, if you want to expand the capacity you must back up, add the new drive, format, reinstall the OS and then restore from tape. With online expansion, you add the drive to a running server and expand the RAID set capacity, all without any reinstall or restore, and your OS will see the additional space. This capability is offered by vendors such as Compaq.
Check the number of bus slots. After you insert your network card and the SCSI controller, how many PCI slots are left? How many EISA slots are left? If you're planning
for expansion, this is an important point. Make sure you have at least one PCI slot to spare. If your server has only two PCI slots and those are filled with a network card and a SCSI card, you won't have the ability to expand using another PCI card later on. In a dual-bridged PCI design, you shouldn't place heavily used EISA cards in a PCI slot, since this will cause a large degradation in performance of the PCI cards on the secondary PCI bus. Although legacy EISA cards can be phased out, you can use a low overhead EISA card, such as a remote management board, and then use the secondary PCI bus slots for network cards.
Sending an SOS
The initial cost of the server is actually only a fraction of the cost you'll incur in supporting it. Management software can help reduce operating costs. If you need phone support from the vendor, look fo
r seven days a week, 24 hours a day (7x24) availability. HP, for example, offers free support only during business hours. Compaq and IBM have free 7x24 support. I
f you're looking for international support, note that some vendors, Compaq, for example, outsource to Digital Equipment Corp. in some areas. Others, such as HP and IBM, have their own support mechanisms worldwide. If this is a concern, you may want to discuss it with the vendor early on. Finally, take stock of the server's warranty. Similar prefailure warranties are a useful way to achieve simple proactive manageability.
Kiran Movva is a systems analyst at a major energy corporation on the West Coast. He can be reached at kmovva@nwc.com.
|