Using Console Server Technology In Server Farm Environments.
Server farms offer a considerable number of advantages--scalability and flexibility being two of the more obvious ones. The permutations available to businesses can be endless. Server farms can range anywhere from a handful of processors (less than 10) to hundreds or thousands, and can be used to handle everything from general computing and database management tasks to more specialized engineering, design, simulation, and animation functions that require infinitely more processing power than any individual server can ever hope to deliver.
While distributed computing delivers some undeniable advantages, it also brings with it some inherent challenges. For one, centralizing the management of these devices becomes increasingly complex, and depending on the method used, extremely costly. For another, there are simply fewer IT resources to go around, especially in a world that expects round-the-clock functionality and systems support. Simply put, skilled help is hard to find and having individuals on-site at various locations--next to impossible. So, the more efficiently an operation can structure its service requirements and the better it can manage its systems remotely, the less it requires in terms of human resources.
Another management challenge lies in the fact that the computers used in server farm environments are more numerous and smaller in size to allow for more efficient racking of devices. As a result, much of the "superfluous" hardware, such as KVM (keyboard, video, and monitor) cards have been eliminated in the newer systems, in the interest of saving space. This is also referred to as running "headless". As a result, administrators who have become familiar with these more traditional means of access must rely on alternative methods to gain the visibility into the multiple devices at any given time without having to interrupt service.
There are numerous ways and means to manage these new environments from a hardware standpoint. What has aided the server management cause to a great extent is the prevalence of the Internet and the role it can play in supporting computing systems. Through the Internet, a systems manager can gain access to any machine to monitor and diagnose hardware devices such as hard disks, fans, and so on--in other words, all the devices that an operating system does not support.
Yet while the Internet is an all-pervasive force that is impacting how server farms are managed, there are still limitations from a purely practical standpoint when it comes to accessing the actual hardware in the devices. In-band management via SNMP, telnet connections, or proprietary management software, for example, allows one to manage devices over a LAN or WAN, but at the same time, its reliability on this network leaves one open to problems in the event of a unit or network failure. In addition, the failure of a single device on the network could potentially prevent management connectivity throughout the network itself. Although some management software can provide some additional functions and features, it can be costly in terms of implementation, maintenance, and training.
As mentioned earlier, many of the newer devices in a rack have no KVM ports, which also means that any of the above mentioned methods can only be as good as the connections they make. If a critical device on a network goes down and the only alternative access is via a KVM port, there is little benefit to be gained from an Internet-based solution. The only consistent access to a given device is the tried and true console port: a serial interface that allows communications to various devices, including mainframes, servers, and peripherals.
Console Ports To Console Servers
System administrators have been using the console port for a considerable length of time. But its original use was very much dependent on being in the right place at the right time. In the early mainframe days, the console port allowed managers to access hardware in a geographically contained environment, usually within a single building. This required, above all else, proximity to the device, which was not an issue in the days preceding distributed computing. But in today's distributed world, geographical disparity can be counterproductive to efficiency in systems management, at least from a hands-on perspective. For many organizations, there are simply not enough systems support people available to provide round-the-clock service and support at multiple locations.
The logical step therefore is to look to managing devices over a network. In a networked environment, it is easy enough to access the console port using a dial-up connection, but then cost and complexity enter the equation, since each device would require the installation of a modem and phone line. Not only would this be extremely expensive and cumbersome, it would be next to impossible to manage. The cabling requirements alone would be beyond the capabilities of most managers to handle. It is evident that this mode of access is becoming less and less flexible as uptime demands--and the number and complexity of the devices--increase. Some administrators have tried to avoid downtimes for service through makeshift measures, such as putting servers on a UPS or continuously adding software patches, but this can be a daunting and time-consuming task in that it requires administrators to configure every box on the network. Then there is the basic challenge of actual physical access to ertain devices. For one, it ma y be physically impossible to gain access to the serial port while the device is in the rack, or there may be a need to connect to multiple devices at one time. The climate in the computer room may not be conducive to working for extended periods of time. In many cases, the administrator simply may not be on-site, causing unnecessary and costly delays.
All things considered, console management has become a highly complex function fraught with limitations and roadblocks. Despite the limitations that have arisen to date, however, there is one very effective solution that is quickly gaining ground in the distributed computing world: the console server. A console server is a flexible and highly efficient management device that allows system administrators single, permanent serial management port access to multiple servers and other hardware devices on a network (starting at 16 ports and up), while reducing manpower requirements, cabling complexities, and overall management costs (see Figure).
Through the console server's single point of access via dial-up or LAN/WAN connectivity, system administrators can easily obtain status, manage servers, and diagnose problems. More importantly, this can be achieved without shutting down devices on the network or requiring an on-site visit. The console server can also be connected to other network devices that provide serial console management ports, such as mission-critical routers, switches, remote access servers, etc. By adding all these devices to one system (whether using multiple console server units or not), administrators can create a common, simple method of access to devices to help save time and control costs.
The ability to service devices without bringing the system offline (a feature that is not available on all console servers) is a particular boon to Sun Solaris environments, since these processors incorporate a unique open boot prompt feature that will allow a system or device to be shut down when given a break signal. While a number of more "traditional" console servers send this break signal every time they are powered off, newer product offerings will not. This allows system administrators in Solaris environments then to quickly and easily perform the required functions safe in the knowledge the systems will not go down unexpectedly. When one considers that Sun is one of the leading vendors in the server farm arena, this is proving to be an invaluable--and critical--product benefit.
Of course, training is always a consideration in any IT environment. Using a centralized management tool not only reduces overall support requirements, but can also simplify and reduce the costs of training. The learning curve is substantially reduced, which dramatically improves proficiency gains in device management.
Depending on the product selection, console servers can provide a number of benefits, including In-Band and Out-of-Band management. With this feature, when a device failure prevents connectivity through the network, management connectivity to the router/switch console can alternatively be made through a dial-up connection. Since a console server with a modem can support a number of console devices through a single dial-up connection, it also eliminates the need for individual modems and phone lines to connect every device or system on the network.
Because connection to a console device via telnet eliminates the need for vendor-specific software, the console server can also prove to be a highly efficient use of resources (both technology and human). System managers can identify device consoles by name and run multiple, simultaneous management sessions on multiple devices from a single server connection. Also, selected servers can be rebooted from remote locations on the LAN or through remote dial-up, thus eliminating the need for an onsite visit. As market demand increases, and technology evolves, we can soon expect to see vendors adding additional features and benefits to console servers to increase their functionality and ease of use, including Web browser access and added security.
There is no question that in today's processing-intensive environments, the demand for server farms will continue to grow substantially, which means administration requirements will become increasingly challenging. Centralized, easy-to-use technology tools such as console servers will play an integral part in managing these environments. Not only can they improve efficiency, they can ultimately ease the burden in terms of labour, time and costs, while ensuring uninterrupted service. As their capabilities evolve, we can expect console servers to become a mainstay in the world of distributed computing.
David Krull is product manager of serial products for Perle Systems Inc. (Oak Brook, IL).
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Industry Trend or Event|
|Publication:||Computer Technology Review|
|Date:||Jul 1, 2001|
|Previous Article:||Getting [Virtual] Religion.|
|Next Article:||Enabling Technology For The Data Center OF The Future: DAFS And The InfiniBand Architecture.|