Chapter 10. Network Operations


By definition, servers provide network services to other computers. All serversfrom small, general-purpose servers such as Windows Small Business Server (SBS) servers, to domain controller (DC) servers, to enterprisewide application servers, such as an Exchange, Oracle, or SQL Server serversrely heavily on their network I/O subsystem to maintain acceptable performance. Indeed, in the current architecture of the modern computer, network I/O is seen to be a primary performance bottleneck. That's why the network protocols used, how the network interfaces are configured, the hardware employed, and the monitoring software used are of critical concern to any server administrator and to an enterprise IT department's bottom line. We'll touch on these topics in this chapter.

High-speed server networking must support areas as diverse as the following:

  • Multimedia streaming

  • Server-to-server clustering

  • Server-to-storage communication

  • Wireless LAN (WLAN) networking

Each of these applications makes different demands on a server, and each has given rise to different networking standards. These areas of networking are cutting-edge technologies, and thus many of the newer hardware standards that are meant to support them are works in progress. In addition, many new networking standards are cross-fertilized with other standards. Thus, in this chapter you will encounter areas of technology replete with acronyms that you need to know.

Vendors take three approaches to the problem of improving and optimizing network throughput:

  • Make network components faster.

  • Separate different types of network traffic onto separate network infrastructures.

  • Manage the network infrastructure to eliminate bottlenecks and route traffic appropriately.

The first approach is exemplified by the introduction of Gigabit Ethernet (GigE) into the network over the past couple years. GigE is a faster standard that is overlaid on earlier, slower Ethernet technologies. GigE is backward compatible with Fast Ethernet (100BaseTX) and even slow Ethernet (10BaseT); thus it preserves your investment in infrastructure, such as CAT5 cabling and network management software.

The second approach is often referred to as "out-of-band" networking, with an example being Fibre Channel storage area networks (SANs) and storage interconnect technologies such as Giganet's cLAN, Myrinet, InfiniBand, Virtual Interface Architecture (VIA), Tandem ServerNet, and others. An out-of-band network, as differentiated from an in-band network such as your Ethernet infrastructure, is meant to direct specialized network traffic off your communications network and to offload processing from your server's CPU(s).

Backup traffic, for example, can saturate an Ethernet network, making all other traffic painfully slow. So out-of-band technologies are meant to be self-contained. That is, network adapters that support VIA, for example, expect to talk to other network adapters that support VIA; these networking standards are not generally interoperable with other standards, but sometimes they are.

This chapter includes a discussion of the Fibre Channel networking standard. Fibre Channel is really a server-storage interconnect technology that utilizes the SCSI interface. (The reason that SCSI doesn't qualify for inclusion in this chapter is that it is limited in terms of both cable length and the number of attachment points available.) Fibre Channel fiber-optic cables, on the other hand, can be used over several kilometers, if needed, with dozens of attachments on a Fibre Channel loop and tens of thousands of possible connections in a Fibre Channel fabric.

You will also find emerging standards such as iSCSI, which spans two different mainstream I/O technologies. iSCSI takes SCSI data traffic that would go between the server and the separated storage and sends the data over specialized parallel SCSI cables, packaged or wrapped inside an Ethernet frame. The Ethernet traffic is then sent over a packet-switching Ethernet network between two dedicated iSCSI network interface cards (NICs), using a CAT5 cable infrastructure. Because the traffic is now TCP/IP traffic, the network runs can cover long distances and are both managed and fault tolerant. That is, with iSCSI, the data is encapsulated into a TCP/IP wrapper and is a standard TCP/IP frame. You can use your standard Ethernet network management tools to manage iSCSI traffic, and because TCP/IP is a routed fabric architecture, it is redundant in the sense that it can survive path failures and be retransmitted and rerouted as necessary.

iSCSI offers a cost-effective way for small and midsize servers to send storage data over an enterprise's Ethernet network, and it has the added bonus of letting you leverage your staff's experience in TCP/IP networking.

iSCSI gives you the block-level access and storage I/O intelligence of SCSI, but it transforms your SCSI storage devices from server attached to network attached, which means they can be managed, partitioned, and allocated to servers and applications much more efficiently. iSCSI also provides much better reliability and availability than does regular SCSI. As with Fibre Channel SANs, there's no need to take down a server to add storage to an iSCSI SAN, and after it is added, that storage can be made available to any SAN-attached server.

When you consider the fact that higher-speed networking taskssuch as server-to-server connections, server-to-storage server communications, backbone transmissions, or any kind of networking that involves data transmissionfan out from a high-speed computer, you realize that you need to you learn about the newer interface technologies that are introduced in this chapter.




Upgrading and Repairing Servers
Upgrading and Repairing Servers
ISBN: 078972815X
EAN: 2147483647
Year: 2006
Pages: 240

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net