Spring 2009Your physical environmentToday’s computing environment has changed: moresmaller boxes such as the ones used in this class arebecoming more common. And even those are now beingpartitioned into smaller logical computers, such as withXen and vmware (changing the trend seen in the midto late 90s when large computers such as Sun Enterprise10000 class were being divided into a 8 or 16 “domains”.)Instead of seeing large boxes sitting on their ownCNT 4603Spring 2009in a computer floor, generally smaller computers are“rack-mounted”, and generally follow the convention ofspecifying their height in “u” notation:+ 1U == 1.75 inches+ 2U == 3.5 inches+ 3U == 5.1 inches+ 4U == 7 inches+ 5U == 8.34 (or 8.75) inchesCNT 4603Spring 2009Fortunately, the physical challenges are less with smallerboxes: they are less bulky, they weigh less, they (generally)consume less power and less cooling than the behemothsof old. Some challenges do grow: more power sockets arenecessary to handle more machines in a single rack.CNT 4603Spring 2009Maintenance? Should I buy a hardwaremaintenance contract?+ If you are buying essentially commodity equipment (i.e.,it is using standard low-end parts that can be easilypurchased from multiple vendors), then this questionbecomes less one for the system administration groupthan for the budget department. If you are purchasingthis equipment from a major vendor, you are likely toreceive a warranty that is sufficent (3 years being aboutCNT 4603Spring 2009the maximum that you want to plan to keep any pieceof equipment these days.)+ Onsite maintenance versus swap maintenance: onsitemaintenance means that the service provider will sendsomeone to work on your problems; with swap, you getto do the diagnosis (perhaps with the vendor helping onthe other end of a telephone line), and then the vendordispatches a new board or other equipment, generallyvia overnight mail.CNT 4603Spring 2009Safe board-handling+ It is best practice to use a grounding strap to minimizedischarge of static electricity.+ Reseating boards: many years ago, it was quitecommon to take an ordinary pencil eraser and cleanthe board contacts if a machine was behaving oddly.While USAH still recommends this (page 747), usuallyjust reseating a board is more than sufficient to correctthese type of problems.CNT 4603Spring 2009+ Simply pressing down on socketed ICs does still work,although finding a socketed IC is getting very rare thesedays.CNT 4603Spring 2009Monitors: going to LCDsFortunately, we have almost completely moved awayfrom old CRT technology, some of which we still havehere in room 016 on the floor ready to head to a newhome, and toward the use of LCD technology.These are much easier to move, and in the event of anearthquake (not likely here in Tallahassee!), are much lessdangerous if they fall on or near you.But for those times when you might still work withCNT 4603Spring 2009the old glass CRTs, keep in mind that they are relativelyfragile and also are not safe to open up. If they don’twork, there are not any system administrator-level partsinside of a monitor that are serviceable.CNT 4603Spring 2009MemoryAs USAH points on page 747, memory these days is acommodity item. Prices are very attractive (a gigabyteof memory from Newegg is currently around $15 to $30)as of this morning for commodity memory), and generallyyou should not buy memory even for a non-commoditymachine from the original vendor if you can avoid it.Memory is very static sensitive, and you should stronglyconsider using a grounding strap whenever you work withCNT 4603Spring 2009it.Removing memory is generally more challenging thaninserting it. Generally, it snaps in quite easily, but removalon some older machines was not as easy — the worst inmy memory was a box from Sun where I succeeded inbreaking more than one (expensive) memory board on aparticularly bad day.CNT 4603Spring 2009Preventive maintenance (also known as“p.m.”)In the past, preventive maintenace was a big job dutyof system administrators. Cleaning printers, cleaningtape units, vacuuming or blowing out dust: there wascertainly a plebian side to being the system administrator.Fortunately, there is less call for such mundania these daysexcept for dust problems.As USAH points out on page 748, some preventiveCNT 4603Spring 2009maintenance is done to combat human failings. Stackingbooks over vents is still all-too-common. Even rack-mounted systems are not immune to air problems. Theworst that I have seen was was when cardboard wascasually inserted into a rack system that had closeddoors. Until I happened to open the doors on the rackfor that machine (which was showing a significant over-temperature reading on its internal sensors), I had no ideawhy it was getting too warm.But the parts that fail most often are fans and powersupplies. Fortunately, these days it is becoming commonCNT 4603Spring 2009on servers to have N+1 separate power supplies, whereany one of them call fail and the machine will continueto run. Both failing fans and failing power supplies cannow often be detected by the computer hardware itself(though in the case of fans, it may be indirectly throughtemperature monitoring on some equipment.)CNT 4603Spring 2009EnvironmentAs does USAH, I recommend keeping your machineroom around 66 to 68 degrees Fahrenheit (19 or soCelsius) with about 30%-50% humidity. I have seen alarge machine room which was being kept around 78degrees, which I think was too warm.However, cooling equipment can be a bit worrisomeitself; if you have raised flooring, you may come in as I didone day to find that the cooling equipment had ventedCNT 4603Spring 2009water down into the raised flooring creating a lake (or atleast largish puddles as has occurred in the machine roomwe had in the first floor of MCH.)While monitoring temperature is quite common ona lot of computer equipment, as USAH recommends,having a separate temperature monitoring setup is a goodidea. USAH mentions a standalone product (PhoneticsSensaphone); another possibility is that other equipmentsuch as PDUs often also have their own environmentalmonitoring capabilities. (A PDU is a power distributionunit. These are very common in large server rooms.)CNT 4603Spring 2009PowerIn providing power in a server room, many places,including our computer science server room, provide UPSpower. Generally, the UPSs used in computer roomsprovide power conditioning in addition to emergency powerservice in the event of power