• Finding an authorized dealer is not the only step in a selection process, the important issue is the total operational cost
• The industry points out that the server-hardware can be classified into three broad categories—entry-level server, mid-range servers and enterprise-level servers
• Two basic things while selecting a server are the server environment and subsystems. Space and accessories are important to deliver the level of reliability and performance

A quick run down of Gordon Moore’s observation points out that a new chip gets released every eighteen months, while at the same time, the capacity increases two-fold. The statement was made in the mid-Sixties. And it needs no great mind to work out on the capacity, as of today. So, what a server does today and will be doing in future, is only a limitation of imagination. Nonetheless, the basics have not changed. It lets network users share files and printers, makes management of data easier, makes data more secure when held centrally and protected by a single centralized backup regime, runs shared applications, such as web, e-mail servers, fax servers and databases. Clearly, the server signifies plenty of processing power and enough memory to cope with the demands of the applications. And in today’s age of technological resourcefulness, purchasing a network server is extremely important for upgrading the status and efficiency of any business.

Here, we take a look at what is happening in the server-hardware space, and glance at some of the trends. Further, we also understand that for businesses, either entry-level or the established ones, the questions at present, are—"How do I go about selecting the right network server for my business? What details should I be concerned with?" We have spoken to the leading vendors on this, and arrived at some basics.

What’s Important?

The simplest way to select a network server would perhaps be to refer to a Yellow Pages directory or a local newspaper advertising or the Internet, and order it. But the question is, from whom? Authorized dealers claim several advantages that they can offer over the local guys. This is definitely not an inflated claim, as they undergo sufficient training and are also abreast with the current and upcoming technological advances in the field and have access to technical support from the principals. Also they could be committed. But finding an authorized dealer is not the only step in a selection process, feel almost all the people we spoke to, in the business. The important issue is the total operational cost. Explains Amar Babu, general manager, channels, South Asia, Intel Corp., "With the current business environment being extremely tough, it is all the more important that IT managers focus on optimizing their costs on IT purchases. At the same time, there should not be a compromise on performance and features, which may impact their businesses negatively. Intel recommends the customer to opt for the server, which meets his requirement and also offers the best price-performance within his available budget".

The same point is reiterated by every one in the industry. Look at lowering your future total cost, is what they say.

The other important consideration in managing the total cost of operation is to go for a well-known, brand network servers. The advantages that pop out to the customer include, on-site warantees to protect them from paying extra for minor glitches in the system. Further, servers also offer a variety of options, such as tape drives, additional processors and hard drives that integrate with the system. So, the process for purchasing a network server and having it installed is straightforward—find a local authorized dealership and select a branded system.

Nonetheless, while buying a network server, one is really lost due to a wide range of suppliers and a wide spectrum of kits. Though this is good when it comes to prices, it is dilemmatic when it comes to working out what exactly to buy. What the leading vendors say is that by an understanding what servers are all about, what features to look for, choosing a server is really no harder than buying a desktop PC at the entry level. "A network server is a computer that manages network resources. Any user on the network can store files on the server, the network server also manages network traffic. It primarily lets network users share files and printers. It makes management of data easier, and can be made a lot more secure when held centrally and protected by a single centralized backup regime. The main hardware requirement is plenty of fast hard disks and is much more important than the speed of the processor, memory, etc. Additionally, servers can run shared applications, such as web, e-mail and fax servers, and databases. The server then needs plenty of processing power and enough memory to cope with the demands of the applications".

But since applications drive the basics for one’s choice, what is it that needs to be understood? The industry points out that the server-hardware can be classified into three broad categories—entry-level server, mid-range server, and enterprise level server. The high-end servers are the most powerful servers of the lot, with cost of Rs 4.5 crore, and are capable of serving massive organizations; the mid-range servers cost between Rs 40 lakh and Rs 4.5 crore, mainly comprising of RISC/UNIX-based machines. And the low-end typically, that can be split up as the Standard Intel Architecture Servers (SIAS) and low-end RISC/UNIX servers, cost less than Rs 4 lakh a piece.

A look at the IDC report on the the total server market in the country, would give us an indication where the buying has been. The report says, the entire server market was Rs 1,359.8 crore in 2000. While the SIAS category accounted for 43 percent of the market share, the non-SIAS category entry level accounted for 30 percent of the market share. The mid-range accounted for 21 percent of the market share and the high-end for 6 percent. In real terms, the SIAS spend was Rs 587.5 crore and non-SIAS spend was Rs 772.3 crore. Compaq led the way with 36.5 percent of market share in the SIAS category and Sun, the leader in the non-SAIS category. While HP, Compaq, and IBM, are present in both the categories, Sun is purely non-SIAS. According to the same IDC report, in the SIAS category, Compaq is followed by IBM with 27.3 percent of the market share, HP with 10.3 percent of the market share, HCL with 5.8 percent of the market share, and DELL with 5 percent of the market share. And in the non-SIAS category, it has been Sun with 28.5 percent share followed by IBM with 25.1 percent share, HP with 21.7 percent share, and Compaq with 20.3 percent share. This also points out the co-relation between vendors and support on the operating system. In operating systems, at the entry level is NT operating system, while on the mid- and high-end, it is UNIX which is the preferred system. In fact, UNIX is the table leader with 49 percent share and NT close to 29 percent. IDC estimates that UNIX will continue to garner close to 57 percent of the market share by 2005, while NT would be having a 30 percent share by then. Meanwhile, LINUX too, is slowly catching up. So the link, Microsoft OS and SIAS on the entry level, and RISC and UNIX at mid- and high-end.

At the entry-level server, it has been a SIAS-based story. Mainly because these products are cheap, and in most cases, will have just one processor, and enough memory and disk space to share files and printers on a small network of up to twenty to twenty-five users, efficiently. Additionally, they can run a couple of applications like web and e-mail servers— the most common requirements. However, these file and print servers are not meant for department-level use or critical data storage. Typically, Pentium III is ideal for small workgroups. They come with a minimum of100 MHz bus, 128 MB RAM, and a 10/100 Ethernet controller, and are dual-processor capable. And can have two hard drives, a tape backup drive large enough to cover all the data on the system. Other features like built-in temperature, voltage and fan monitors, with the software to monitor and control the server (from either a custom console or a popular SNMP management platform), become integral. Also rack-mount forms are also becoming available in a standard 19 inch rack.

Then there’s a degree of overlap between such mid-range servers and the final category, enterprise servers. These are designed to run the most demanding of database, e-commerce and other applications, and also to share files and printers among hundreds of users. As a result, the biggest differentiator is the number of processors supported. Many quad-processor systems are available, plus eight-way servers for the really big installations. The type of processor also changes—Pentium III Xeon is most commonly specified. While this is happening on the hardware level, at the OS level, Microsoft products have been the popular choice for small to medium-sized businesses, and UNIX continues to be very popular at the high-end, and there are LINUX implementations too. In the non-SIAS category, it has been Sun, IBM, HP and Compaq, mainly because of the architectures supporting UNIX platform, which has been the sturdy base of critical applications and database management

And with server consolidation (bringing together applications), databases and services onto fewer, highly reliable servers, which is believed to cut high management costs associated with the runaway proliferation of small and mid-size distributed servers in their organizations, these companies are the only few to take measures and support the cause. The interesting trend is that all the four big guys have been carving niches and driving the same fundamental of more flexibility and reliability, at lower costs. Sun has continued to work with Solaris and Spark architecture, and has brought in products for both the workstation and the mid-range category, HP worked with Intel to come up with Itanium, IBM extended its championing to LINUX, Compaq worked with Intel for support on its Alpha structure, and SGI is working on its NUMA architecture to enable business intelligence applications of the organizations.

While these are all some of the technology considerations, two basic things that need to be kept in mind are the server environment and subsystems, explain pundits. Space and accessories are important to deliver the level of reliability and performance. Further, uninterrupted power supply and ventilation for high-performance, are also important. And as high-end servers will consume a lot of electricity, it needs to be ensured that overloading of the circuit is avoided. In fact, these day options like the rack-mounted servers are available to address the needs listed above and additionally then, several servers can be mounted in a single rack or in multiple racks, and this saves the need for multiple monitors, keyboards, and mouse devices that can be controlled through a single set of I/O devices.

Servers also display plenty of other characteristics that can prompt one to decide on a box that delivers speed, capacity, and fault tolerance. For example, processing, memory, storage, etc. Most server-specific motherboards support two CPUs or more. It has been observed that multiple processors offer the benefit of aggregating more computing power within a single machine, but the underlying operating system, and its applications and services, must be multithreaded to take advantage of multiple CPUs. What most of them advise is that performance reduces with an additional CPU. A two-CPU system offers 1.5 times the performance of a single processor machine, three CPUs offer about 1.8, and four CPUs only double that of a single CPU. Similarly, while more is the memory, greater is the operational ability, storage can be enhanced by Ultra Wide SCSI drive controllers and mission-critical data can be stored by using RAID 5 arrays, and there are hot-swappable drives that are also available. Moreover, today, fiber optic-based computer peripheral architecture named ‘Fiber Channel’ beyond the basic PCI and S or VME bus devices, is available to deliver throughputs up to 125 Mbps.

Vendors say selecting an actual server for purchase is not as difficult as it may seem, especially with most of them looking at bringing in more compatibility and better features. There are only three significant factors. One, the budget; two,platform preference; and three, the level of traffic the server is designed to handle. And the key considerations to all this are bandwidth, flexibility, higher availability, industry-standard technology, and consolidation. The higher bandwidth requirements can be addressed through higher I/O, and appliance servers and server blades; flexibility can be addressed through modularity, clustering, etc, industry standards are being arrived at for deploying compatible technologies; and server consolidation to effect lowering of the total cost of ownership.

Choosing the Right Server

Ch. Srinivas Rao