The Right Hardware

The Right Hardware

Purchase

There are three different ways of obtaining your workstations and servers. Each as their own and advantages and disadvantages. The first is to build all of your machines from scratch. This can be cheaper in terms of actual cash outlay for the hardware. However, you also need to consider that a lot more time will be spent putting the pieces together. This can mean that the actual cost is higher.

The second alternative is to buy whatever computer happens to be on sale when you need it. This decision saves the time of putting together the machine as well as saves you money. The downside is that you may end up with as many different configurations as you have machines. This means additional administrative work to manage all of these different configurations.

Finally there is drawing all of your hardware from a single vendor. This helps standardized your machines, while at the same time shifting a lot of the work to the vendor. In many cases you may not work with the vendor directly, but rather a distributor or reseller. In essence this has the same advantages and disadvantages of working with the vendor directly.

Which of these methods you choose will depend a lot on your unique circumstances. Although building machines from scratch may not be the best solution for smaller companies, you may discover a financial advantage in purchasing then all from the same vendor (i.e. economies of scale). In order to get the best deal, you’ll need to negotiate with each vendor individually. What additional services are provided vary dramatically from company to company. Because it requires a fair bit of administrative work, let’s look at some of the issues involved in dealing with a single vendor.

You can obviously save yourself a great deal of administrative work by simply buying the same brand and model from your local computer superstore. However, this defeat the purpose and loses many of the benefits of going directly to through the vendor or through one of their official resellers.

The first thing that you will be missing is having a personal contact within the company. Although it is possible to get an individual account representative when buying through a large chain, you typically do not get the same level of personalized service. This is important issue in that part of what you’re trying to do is to shift much of the responsibility onto your supplier. It is possible that the local retailer is in a position to support you directly (e..g in-house tech support), so it is worth investigating.

There is a large variation in the sales program and services the vendors will provide. This not only depends upon the vendor themselves, but also of the size of your company. You are obviously in a much better position to make “demands” of the vendor if you’re planning to purchase several hundred machines. However, even in smaller units you have a great deal of negotiating power. Remember there are dozens of computer vendors. If the first one cannot (or will not) meet your needs, the logical alternative is to go somewhere else. Depending on your area, you may even find companies which will even build the machines for you according to your exact specifications.

One thing that is often forgotten is that buying the computer outright is just one alternative. Some companies will rent the computers to you. Others will lease them. Still others have programs in which you can buy the computers at the end of the lease. Although this is slightly more expensive then buying the computers outright due to the amount you need to pay in interest, you can end up saving time end money, because the vendor is responsible for getting the computer repaired (just like a leased car). Depending on your contract, the vendor may be obligated to provide you with a replacement while your computer is being repaired.

Another benefit in purchasing from the vendor is that you commonly get quantity discounts. This does not mean you’re required to buy in large number of computers at once, rather you need to buy them within a specific time frames such as six months or a year. In addition, the software may also be part of the deal. Normally, you can’t lease the software, but you can spread out the purchase price of the software over the lifetime of the lease. (although this is less of an issue with Linux)

One thing to look at the leasing contract is whether or not hardware upgrades are included in the lease. One possibility is that you can upgrade the hardware for a moderate payment. This keeps you up-to-date without making you go bankrupt.

One important thing to keep in mind is that standard machines from brand-name companies are often not as standard as one might think. I once worked on a server from DEC (first Compaq and then HP) which had a defective drive in a mirror set. When the replacement drive was sent, we discovered that it was smaller than the original although it had the exact same part number, the drive was from a different manufacturer who had a slightly smaller size than the other (although both were sold as 4 Gb). If the drive had been bigger, it would not have been a problem. However we could not simply recreate the mirrors set after adding new drive. Instead, we had to install and format the new drive, copy all the data to it, and use it as the base for the new mirrors set. This meant that some of the space on the older drive was lost, but more importantly it cost us great deal of down-time.

This example is not intended to demonstrate problems with brand-name computers, nor is it intended to put DEC/Compaq/HP in a bad light. Instead, it is intended to point out that there are things which you need to look out for. In this case the new drive was not an adequate replacement for the old one. If the vendor is not willing to provide an appropriate replacement, then you might want to consider looking for a new vendor.

The steps to take when a machine or component needs to be replaced should be formally documented. Every administrator should be in the position to replace or repair machines as needed, this is not to say that every administrator has the ability to get out and fix damaged cards. Instead, the administrators should be a position to get the machine repaired. Your company may be different in that there is a special group which is responsible for repairing PCs. However the principal stays the same. As long as defined procudures are followed, they should be able to get the machine repaired.

Another important aspect of standardization is defining exactly what information will be kept, where it will be stored and in which format the data will be in. If different groups used different formats to store their information, sharing becomes difficult and in many cases impossible. This may be as simple as creating an database to which everyone has access, but you also need to ensure that the databases are all compatible.

Part of this is determining who has what access to what information. Although everyone on your help desk (if you have one) should at least be able to read this information, is not absolutely necessary that they be able to update or change it. If multiple people are all allowed to change this information, then the procedures also need to be documented.

Depending on the number of users in your company, you probably will end up with several different kinds of standard workstations. Each will depend on the type of work that is done on that machine. For example, you might have machines that are used solely to input data into databases and others which are used for your technical applications. By defining a standard you are likely to find that repairing or replacing the machines is much easier. You know which components go into that machine so you can simply order that exact same part. You know what the standard workstation is so you do not need to waste time trying to figure out what the replacement machine should need.

Because your business depends on getting work done, it is not unreasonable to expect timely deliveries of your computers. Therefore, you should consider including penalties for late delivery in the contract. Some vendors can take six weeks or longer to deliver the machines. If you plan delivery on a specific date and that date is not met you may be out of some money. Why shouldn’t the vendor take responsibility?

In one instance we were repeatedly promised delivery to one of our branch offices by a specific date. Since the installation was being done by administrators from the head office, plane and hotel reservations needed to be made and work schedules needed to be adjusted. The day before the scheduled departure the vendor announced they would not be able to deliver on time (although we had ordered six weeks earlier). Since we planned well enough in advances, we got a substantial discount on all our airline tickets. However, two days before the flight the tickets were neither refundable nor could they be exchanged. This was money that was lost due to problems within the vendors organization and therefore they should be responsible for the damages.

Fortunately, this contract had penalties for late delivery. The contract went so for as to stipulate that all machines had to be up and running for the contract to be considered fulfilled. For each machines that was not running the specific penalty would be applied. Here we had ten PCs that were already delivered, but the server was not. Since this was a Windows NT domain and we needed the server to do the installation and allow the user to to work, it meant that neither the server nor any of the workstations were operational. Bad planning on the part of the vendor resulted in a 10 fold penalty.

Even if you do not lose money because of non-refunded airline tickets, you do lose time and sometimes money when deliveries are not made on time. If users are waiting for the machines, they cannot work as effectively as they should be able to. You have increased administrative costs for the time you spend tracking down the status of the order. If the machines are replacements for existing computers, you may have the problem of different users running different software. These problems can be compounded when you have a large number of machines spread out across multiple deliveries. You have the extra administrative burden of insuring everything is delivered properly.

In such cases, it is in your best interest to have the contract stipulate what things cause a contract violation and what the penalty is in each case. And should also stipulate what your responsibilities are, for example, making arrangements with an ISP.

When deciding on a vendor, one thing to ask is how long they will guarantee the delivery of specific equipment. That is, how long will they ensure that a specific model is available. It defeats the purpose of standardization if you get different machines with each order. One vendor I have worked with changed models every three or four months. With a 6 to 8 week delivery time, there was a high probability that orders made two months apart would have machines that were not the same model.

An argument in favor of the vendor might be that this was necessary in order to keep up with the technology. To some extent this is true, but it makes managing standard PCs extremely difficult. If you are contractually obligated to purchase a specific number of machines within a specific period of time it is not unreasonable to expect that be vendor is obligated to provide you the same product during the lifetime of the contract. Besides, in a business, you are usually less concerned with being on the “bleeding edge.”

Due to the frequent changes in the computer market, it might not be possible to completely guarantee the availability of machines over the course of the year. Therefore, you’ll need to work closely with your vendor to see that they come as close as possible.

We next come to the magic phrase total cost of ownership (TCO). Eventually the computer will need to be repaired. The question is what kind of service are you getting from the computer discount store as compared to the major vendor? Even if you are not buying hundreds of computers at a time of you can still get good support from the major vendors. Obviously, the quicker and more detailed the support the more you are liable to pay.

However, I need to warn you here. Just because you’re buying workstations or servers from one of the major vendors it does not mean that what you’re getting will be consistent. One prime example is the example with the hard drivers mentioned above. In such cases, you cannot always be sure that the components are consistent even within a particular model. That is the description of two machines may be identical. However, they could have different motherboards, different video cards, or different hard disks. This makes mass installations of machines extremely difficult. For example if you are expecting a particular video card to be in a machine, the installation may not work correctly because of the different video card.

Another advantage of having a single vendor is that you have a single point of contact. If something goes wrong, for example a computer breaks down, you know exactly who to call. On the other hand, if you have machines from different manufacturers, you need to first check to see her the manufacturer is of the computer in question. In addition, if you have warranty or maintenance contracts, you’ll need ten separate contracts, one for each of the different vendors. A side benefit of having a single vendor with a single maintenance or support contract is that you build up a relationship with the company. Depending on the company, this can meet to sometimes getting more benefits that are then are defined in your contract.

One thing that we have found out is vital when ordering hardware is that the invoice should be as detailed as the original order. For example, if you ordered and Adaptec 2940 PCI host adapter , then this should be listed on the invoice. In some cases I have received invoices that simply stated that there was a SCSI host adapter. This is not enough, especially later when you need to make warranty claims. Also, simply stating that there is 32 MB of RAM is not enough. The invoice should state exactly what you get like 60 ns EDO RAM.

Sometimes buying hardware is more than just finding the fastest or largest for the best price. Often, it is a question of “piece of mind.” It’s like accident insurance. You know to be careful, but sometimes accidents happen. That’s why it’s called “accident” insurance. In essence, you are betting against yourself. However, the costs of being wrong are not worth the risk. Therefore, you pay a little extra each month, “just in case.” Even if you never have to use it, there is the piece of mind of knowing that if something were to happen, you would be covered.

I take this same approach to hardware. For example, I will never again put an IBM hard disk in any machine. In the space of about a year, I had three different IBM hard disks, all with similar problems. The first was a SCSI drive. It repeatedly reported read and write errors. I brought it to my dealer, who kept it for several days to “test” it. After three days of not being able to work, I got the same drive back with the report that the dealer could not recreate the problem.

When I put it back in the machine, I got the same problems. At that point the first thought was a defective controller or cable, but the drive reported the same problems on a completely different machine.

Next, was an IBM IDE drive. I brought it home, pluged it in and it wasn’t recognized at all. By that time the dealer was closed, so I had to wait until the next day. In this case, they could easily recreate the problem, since it wasn’t recognized on their machine either. They gave me a new one which I installed. A couple of months later the same read and write errors started appearing as with the IBM SCSI drive (also on different machines).

One day when I turned on the machine, the system reported that it could not find the primary hard disk (which was supposed to be the IBM drive). Although I could boot from a floppy and access the hard disk, I could not read anything from the root directory. When I ran CHKDSK.EXE (It was a Windows machine), I was left with about 3 directories that were not damaged. All others were now in the root directory with names like DIR00001.

Needless to say, this forced me to re-install my system, dozens of applications and reconfigure the entire system. This cost me two days (actually two evenings) until I was back at a place where I could work effectively. However, it was more than a week until I had re-installed all the applications.

Two months later, it happened again. The primary hard disk was not recognized and I could not access the root directory when I booted from a floppy. A low-level format revealed a dozen bad tracks. Obviously after the low-level format I had to re-install the system, my applications and reconfigure everything.

This taught me three important lessons. First, even if the data is on a different system and backed-up regularly, a crash of the primary hard disk means a lot of time is wasted reinstalling everything. Therefore, you need some mechanism to quickly reinstall and reconfigure your system. Some of these techniques we will get to in later chapters.

The second lesson I learned is to get a dealer that won’t waste my time “testing” the drive and then return it to me when they cannot find any problems. If the problems are intermittent, the dealer should expect that the problem cannot be re-created so easily. In my opinion, the dealer should simply give you a replacement and be the one who has to deal with the manufacturer. For private customers, the dealer may not be so willing to go out of his way. However, if you are business spending hundreds of thousands of dollars a year, they must be willing to make the extra effort.

The last lesson is never to put an IBM drive in my machine again. IBM “claims” that they have a large number of drives in all sorts of machines and do not have these problems. There are probably many of you, who are reading this who have not have had the same experience. However, I am not willing to take the risk, just like I am not willing to take the risk of driving without insurance. The consequences of being wrong are two high.

Every single IBM drive I have had has exhibited some kind of problems which put my data at risk. Even if a drive from another vendor costs $20 or ever $50 more than one from IBM, I am willing to pay it for the piece of mind of not having to worry about whether the system will boot the next time I turn it on or that the data actually gets written to the drive. Even though I know people who have not have the same problems, I want the piece of mind.

An additional advantage is the learning curve is much steeper with hardware from single vendor. This is because the construction of the machines and components are generally the same. Again, this is not an absolute. Because the hardware is similar with each machine, the administrators and users do not need to re-learn each time a new piece of hardware is acquired.

On the other hard, sticking with a single vendor can become a trap.Can you really wait six weeks for your machines? In one company where I worked, we stuck with a single vendor who could not promise delivery in less than six weeks. The IS manager thought it was more beneficial to have 100% standard PCs than having a PC at all. Even though the machines differed from country to country and sometimes even between deliveries, it was pretty much standardized within each office.

In one company they had the best of both worlds. PCs were grouped into classes with similar characteristics. However, being 100% identical was not as important as having a machine. Therefore, the company stuck with three vendors. If the first one could not deliver fast enough, they went to the second.

Note that this requires a little discipline on the part of the administrators. They needed to keep track of which machine had which configuration. However, with any of the asset management tools available on the market this is fairly easy.

Going to three vendors also has the added advantage of being better able to dictate prices. Obviously, if you must have the machine within specific amount of time, you will probably end up paying more. However, if you are willing to wait a little, you might be able to cut a good deal. For example, if one vendor says they can deliver in four weeks and another says they can deliver in two weeks, you might be able to get a good deal by waiting the extra two weeks. Remember: supply and demand. You are supplying the money, so you can make demands.

Repair

You might find it useful to develop a “maintenance pool” for your workstations. This pool consists of complete PCs as well as copies of the different kinds of hardware. Standardization of your machines means the you need to have fewer different models and different brands of hardware. This allows you to choose between replacing the entire machine or just individual components.

In some cases it is not always necessary to have exactly matched hardware such as hard disks. However, with devices which require a specific driver, such as hard disk controllers or network interface cards, it is in your best interest to standardize this as much as possible.

If your hardware pool is designed to swap out the defective machine, repair it, and then return it to its original owner, you can decrease the number both spares you need body using removable media. Your maintenance pool consists both of middle to high-end machines. Each with their own removable media drive. When you needed to exchange machines it is not matter who needs the machine, but you simply take one from the shelf and stick in the appropriate drive.

The downside of this is that it requires much greater standardization of hardware and there is the additional expenses of the removeable media drives. These have a much higher cost per megabyte as standard hard disks. Alternatively, you could simply swap out the hard disks and save on the cost. The downside of this is the additional time it takes to install the harddisk. On the other hand, as hardware prices continue to sink having a few extra computers on hand probably isn’t enough to make you go bankrupt.

If you have a maintenance contract with your computer vendor, you should really investigate how things are handled internally weigh all in advance of your first call. In some cases, the sales and support organizations may have the same company name, but are actually separate entities, which have a hard time communicating with each other. I have experienced it myself that after purchasing a large number of computers from one vendor, we could not support from them, because the sales organization had not yet gotten around to submitting the necessary paperwork.

Depending on how much the salesperson pushes the after sales service aspects of their company, you might want to try getting a penalty clause built and should the vendor not repair or replaced the equipment within the time they promise. When they say that a particular service contract “guarantees” a specific response time, what does this actually mean? If a tangible product such as a computer fails to meet the guarantee, you can always return it. How do you return a promised level of support?

Just like when the computer is not delivered on time, a non-functioning computer can cost you money in terms of loss of productivity. If you don’t pay for the computers like you promise the vendor can simply repossess it. What your options when the vendor does not meet their obligations? You couldn’t go to another vendor the next time, but you already lost the money. If the vendor wants your business, some kind of tangible penalty should be applied if they do not fulfill their obligations.

The repair or replacement of defective machines is another place where you can save time, at least from the users perspective. This is one of the strongest motivations for having standard PCs. Should the machine go down you can swap the machine in just a few minutes allowing the user to get back to work. You can then repair the machine or replace defective components at your leisure.

In some companies, assets such as computers are assigned to specific departments or even individuals. Therefore, you may need to return that computer to its original owner. However, if that’s not the case, the repaired machine can then be used as a replacement the next time a computer goes down.

If your workstations are standardized, it is much easier to keep a stock of spare parts on hand. Swapping a defective component is often as cost-effective as replacing entire machines, especially if you do not have the money for machines which are not regularly being used.

Regardless of how you obtain your hardware, you should always have spares on hand. This is not as easy when you get your computers for major vendors, as it is when you build the machines yourself or order them to be build for you. Often times when you order from major vendors, the description may simply be something like “3.4 GB SCSI hard disk”. This tells you nothing about the manufacturer. (Although with hard disks this is often less of a problem.)

However, with other pieces of hardware the difference becomes more critical. For example, if your network card fails you would probably want to replace it with the same kind of card. This enables you to simply remove the old card and insert the new card without having to install a new driver.

I need to emphasize the need to have the spares on hand as opposed to ordering them from the vendor. Even if your vendor promises a one-day turnaround, this usually means one business day. Therefore, if the hard disk dies on Friday, you may not be able to continue work until Monday. If Monday is a holiday, you may not get it until Tuesday. Your business week may not be the same as the vendors.

If your company uses standard computer configurations, there may actually not be any need to have spares of individual components. Instead you might want to consider having spares of entire machines. In one company I worked, we had a few spares for each of the different classes. When one computer failed, we simply replaced it with a comparable machine. Since all of the data was stored on the server, the only thing the user lost was local configuration settings. However, using NIS for users and group names and some cases NFS, even that was kept to a minimum. The result was we were able to replace the computer in less than 10 minutes. That meant the user was back to work in 10 minutes. If we were replacing components, they can take anywhere from a half an hour to two hours, or even longer.

Which method is most effective will depends on your company. The larger the company, the more likely you would want to employ a system by which you replace the entire computer. This is simply because there are more problems occurring in a larger company. You will need to make replacements more often. Therefore you will need to limit the downtime for the users.

If you do decide for systems where you swap components, I would recommend having several spares of each type of components. Once again, standardization of hardware is vital to make the system more efficient. You must be able to swap out like components. You do not know in advance when a component will fail and which component it will be. What do you do if you have only a single spare and the same component decides to break on two machines on the same day?

The odds are generally low for such things, but how much money do you stand to loose if you are wrong? If you have standard machines from a single vendor (or just a few vendors), the vendor may have bought a whole batch of defective parts. Remember, they are trying to keep their costs low and do no always put in brand names. The more components that are defective, the greater the likelihood that even more will have problems.

What you do with the defective computer will be determined by your company’s policy. In some companies, the computer is repaired and the old one is returned to the user. This is because there is a 1:1 assignment between computers and the users. In other companies, where there is no 1:1 assignment, there is no need for the old computer to be returned to a specific user. Therefore, it is returned to the computer pool and waits to be used as a replacement somewhere else.