Discussion Locked

Quit I'm Sinking!!!!!! Net Admin Project

0 helpful
3 replies
Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.

I am a systems administrator working for a 75 Million/yr. Manufacturing company. There are approximately 200 local and remote users that you support along with several servers. As the systems administrator Iam responsible for identifying and solving a wide variety of problems. Many problems can be addressed with little or no expense, however others require a capital expenditure. The CIO and CFO must approve all capital expenditures. Many of the other senior management have input as well.

Based on the problems identified below I must:
 Articulate the problem
 Identify a solution
 Cost justify the solution
 Develop a project plan
 Develop a project budget

Note: You are not required to propose Microsoft only solutions but solutions must be accurate and justified.

My company has a very active 24-hour engineering division. The engineering div. requires access to a networked CAD system and a schematic archive imaging system. The engineering div. is directly responsible for 25 Million/yr. in revenue. There are 40 engineers with an average salary of $130,000/yr.

The server that supports the CAD and archive systems is several years old. The processor averages 85%, and the storage space is at 90%. I have determined that the server has 90% annual uptime. Within the last three months; I have had to restore from backup twice due to server failure. The down time was 8 hours per incident. I attribute the extended restoretime due to a backup system that require me to span across three tapes. The network is 10/100 Ethernet.

The engineers have complained of slow response time. They advise that the average engineer spends 30-40% of their time waiting for the server to serve up files. In the past year along they have had to recreate 76 man-hours of work due to server failure.

Please give me help resolving these problems.

3 replies

by wstankich

This is where I would start, in my humble opinion, from what you've stated. First, you have %30-%40 waste of your engineers time waiting on slow system or network performance. This equates to a 2.08 mil loss per year. In my opinion that is a pretty big loss. I would start by consolidaating protocols, if possible. Ex - TCP/IP, SPX and NetBUI ->> to TCP/IP, WINS(server), DNS(server). This will eleviate excess broadcasting on the wire. TCP/IP is also more routable, as opposed to SPX/IPX which is not. Second, use layer 3 switches like Cisco maybe, and put all engineering PC's and servers on the same switch or group of switches since the switch will maintain it's own routing table and the packets don't have to go to a server to get routed. Mosttraffic will then stay with in the switch or switch group. Third, make all machines on the switch or switch group, including printers being used by the engineers, 100mps/Full Dup. If this doesn't help response to a satisfactory degree, then I would look at my engineering software and database software to consider if there was a faster performing alternative or any tuning that could be done. Last, I would consider server hardware, especially disk I/O. Disk I/O issues, I believe, will also kill asystem. Evaluate for disk contention. If this is an issue, technology has made huge advancements in disk I/O. A vendor suggestion is Winchester disk arrays. As far as justifying cost, your losing approx. 2.08 mil a year and $23,750 in re-work. If you baught a SUN E450, Winchester Disk array, a few new Cisco switches, your only looking at approx. $150,000...that should be an easy sell.

by vt100

I think you were pretty good at:
1. Articulatng the problemm.
2.Cost justifying any solution that will work.
As to what you can do to transition the workload to a new server I may propose this:
You can purchase another server with the kick ass capasity, put in the same network, buy Computer Associates product called SurviveIT, install corresponding components on the original and the new server, Switch the new server to a primary and collect the bonus for saving the day. You may also consider using a clustered server for the destination machine, but I suggest consulting with Computer Associates if it would work first. If it would, go for it, it'll pay in the long run.

Regards

Vladimir

by rzan

Sound like you have a similar situation I faced where I currently work. I am a systems admin for a State Govt Agency responsible for Managing 7 Toll Bridges and 2 Toll Tunnels in the New York City Area. We also have an Engineering dept which uses CAD software and has had latency problems with files. First of all, to stop wasting backup time spanning multiple tapes, purchase a 35/70 Gigabyte DLT Library Tape drive and get away from those stupid 4mm DATs altogether. The DLT Library usually holds 14 tapes plus 1 cleaning. One tape holds an average of 70 Gigs of data in a compressed state, which should be more than enough for one server. Load up this baby and dont worry about changing tapes for 2 weeks, it will use the next blank one as needed. As far as your CPU utilitization and drive capacity, I would suggest getting a powerful departmental server, Pentium 2 or 3, 550-700Mhz with at least 6 18.2GB Ultra SCSI 2/3 drives in a RAID 5 array. Your bandwidth problem may a result of the protocols you use. Avoid using Netbeiu and Appletalk, as those create excessive broadcast packets. You do not have to be in a pure IP environment, IPX/SPX is fine if you have legacy systems which require it. Also consider installing switches instead of hubs which can provide 100Mbit full duplex to the workstations. The server of course must be running 100Mbit full duplex as well. If you are worried about server uptime, consider using clustering software to create reduncacy between your servers.

This discussion is closed to new replies.