| Previous | Table of Contents | Next |
The hardware initiatives we are pursuing to further improve the price/performance (P/P) of the AS/400e series comprise the final topic in this chapter. In recent years, customers have seen dramatic improvements in AS/400 P/P. With the introduction of the e-series, there was an average rate of improvement of about 60 percent across the product line, with even greater reductions for certain server models. In the past, when we talked about improved P/P, we usually meant that year after year the performance and capacity of a system improved with no change in the price. That view of P/P has changed as prices have rapidly started coming down.
I remember buying my first PC back in 1982. IBM had a program for employees to purchase some of the first IBM PCs. Mine had 64K of memory and two 5.25-inch diskette drives (those early PCs had no hard drives). With a monochrome monitor and a matrix printer, that PC cost me $4,500. Today, for that amount of money I can buy one of the biggest, most powerful PCs available. I know — I just did.
Of course, I could have bought a much-less-expensive PC and still had a system with very acceptable capacity and performance, which is exactly the problem computer manufacturers face today. A few years back, most of us were rapidly adding new applications that required us to add capacity and performance at a rate that was just keeping up with the P/P improvements. As a result, we continued to spend about the same amount every time we upgraded our computer systems. That has changed. For most of us, the P/P improvements have surpassed our needs for more capacity and performance. As a result, we are now spending less to upgrade our systems, whether we are buying PCs, AS/400s, or mainframes.
The challenge to all manufacturers in the computer industry today is to find ways to eliminate cost from the systems to keep reducing prices for our customers. In this section, we look at two initiatives the AS/400 is using to improve P/P: commonality and new I/O technologies.
The PC industry is driving the P/P standard for today’s servers. Intel-based servers running Windows NT have set the bar for all other servers in the industry to meet. The AS/400e servers’ P/P is very competitive against these PC servers. However, the bar is not fixed; it continues to drop. An advantage the PC servers have is common hardware, used by many vendors. To remain competitive, the AS/400e series also is moving to common hardware.
The most obvious indication of this move is the appearance of the new physical packages announced with the e-series. The RS/6000 uses these same boxes. Most of the components inside these boxes also are common between the two systems. Common parts mean lower inventory costs, lower field stocking costs, and even lower parts costs because of the increased volumes. To reduce manufacturing costs, IBM consolidated all manufacturing of both AS/400 and RS/6000 machines into Rochester and Santa Palomba, Italy. Common parts made this consolidation possible.
If we look further inside the new physical boxes, we see a newly designed Central Electronics Complex (CEC). The CEC consists of the main processors, the main memory, the power supply, and the I/O buses. Typically, as IBM introduces newer and faster models, these components are replaced. Previously, to get increased performance and capacity, customers had to replace major parts of their hardware, often replacing the entire box. With this more efficient CEC, they can obtain performance and capacity increases simply by plugging in new cards. IBM has designed these new physical packages to last through the Version 4 timeframe, which should reduce the cost of upgrades for certain e-series models. Still other models are designed not to be upgradable, to reduce their initial cost, similar to the approach the PC marketplace uses.
IBM is committed to using common hardware across all its system platforms, which reduces the development expenses on those components that are common to multiple IBM systems and also lowers the parts costs. A partial list of the common hardware includes processors, memory controllers, system buses, I/O adapters, power supplies, and packaging. We have already seen several of these common components in the AS/400e series: PowerPC processors, the 6xx system buses, PCI adapters, and SAN interconnections that will be used on multiple IBM systems. The bottom line for all this commonality is improved P/P for our customers.
Probably the biggest area for P/P improvements is I/O. The move to PCI and the ability to use industry-standard adapters in the AS/400e series is a big step, but it is only the beginning. Several initiatives are underway to lower I/O costs, but I want to look briefly at only two, both of which are related to reducing the costs and improving the performance of disks.
IBM has adopted a new Serial Storage Architecture (SSA) that has begun to appear in various IBM systems. The American National Standards Institute (ANSI) has produced a formal standard for this architecture. SSA defines a high-performance serial link for the attachment of input/output devices. This link has been optimized for storage applications such as hard-disk drives, host adapter cards, and disk-array controllers. IBM introduced the first SSA disk subsystems in 1996.
SSA has many advantages over existing parallel interfaces such as the Small Computer Systems Interface (SCSI). SSA has better performance, better connectivity, and better reliability. It also uses more compact cables and connectors. To facilitate migration and coexistence, SSA retains much of the SCSI-2 logical protocol.
The use of SSA to attach disks to the e-series in a later V4 release lowers the attachment costs and improves the performance. A typical loop configuration with one host adapter in a current SSA implementation can provide a sustained bandwidth of up to 73 MB per second, and higher speeds are becoming available. The physical connection can be either a copper cable for up to 20 meters or fiber optics for longer distances.
This serial link offers several advantages over the traditional parallel interfaces. The smaller, less expensive cables and connectors are especially useful for the smaller 3.5-inch and 2.5-inch disks. SSA provides better connectivity because a single host adapter can attach many devices. The serial interface can provide full-duplex connection to the devices for even higher performance. For higher reliability, error detection is improved via a cyclic redundancy check (CRC) rather than a simple bus parity check. Finally, the serial link is easier and less expensive to switch, a feature that will be especially important as the AS/400 implements switched and shared disks for continuously available clusters.
| Previous | Table of Contents | Next |