--File Server --Network device that provides files on demand --Fits the request/reply format --part of the client server --Application Server --provides services directly --Ex: SEQL Server and Exchange Server --Not a server that serves up applications but an application that responds directly --Connection Oriented vs. Connectionless Communication --Connection oriented begins with an attempt to communicate and a go ahead response whereas connectionless does not require a go-ahead before it begins sending information --Example: Phone conversations require a connection to be made before any information can be communicated so they are similar to Connection oriented communication; US Mail is like connectionless because there is no need to establish connection to send information - there is also no guarantee communication will be received --Transceiver types --transmitter and a receiver --couples NIC to medium --usually built-in, but also externally available --physical connection between the medium and the computer --Troubleshooting common answers: --Netware -> frametype --Token Ring-> Ring speed or protocol mismatch --Ethernet -> transceiver type --IRQ availability --there must be an interrupt available for card to use --available interrupt must be compatible with card --***MEMORIZE THE IRQ TABLE*** --Select and appropriate networking medium --based on cost, distance limitations, number of nodes --wireless is chosen by number of remote users and difficulty of laying cable --Fiber: --greatest distance --greatest bandwidth --least amount of external interference --greatest cost --2 nodes to any pice of fiber is PTP (except FDDI) --difficult to install --easy to damage --requires special trainings --Coax: --best for nodes with little change -> 10Base2 --thinwire is more flexible, no need for hub --thickwire is best for backbones and medium to long hauls --more immune to interference --choice between thicket and thinnet comes down to cost and distance --5-4-3 rule --5: maximum number of segments --4: maximum number of repeaters --3: maximum number of active segments --Cable lengths: --100 meters (TP) 2 max devices --185 meters (10Base2) 30 max devices --500 meters (10Base5) 100 max devices --Devices = pointers, repeaters, terminals --TP is PTP from hub --Token vs. Ethernet --Token Ring is a guaranteed chance to transmit information while Ethernet is probabalistic (not guaranteed) --Token: regulates the medium --IEEE specs on BNC, T-connectors, cable types: --802.3 --RG 58AU - Ethernet 10Base2 Impedence = 50 Ohms --RG 62 - Arcnet Impedence = 75 Ohms --RG 8 - thicknet Impedence =50 Ohms --Where Repeaters, Bridges, Routers, and Brouters sit on the OSI model --Repeaters -> Physical; only change media types --Bridges -> Data -> MAC; only change media types --Routers/Brouters -> Network; change physical media and architectures --as they get smarter they move up the model --Gateways: specialized devices that work at whatever levels they have to allow protocols to talk to each other --Switch --smart, fast, hub --makes electrical isolation paths --every path effectively has the full benefit of all bandwidth --easier and cheaper to install --Protocol choices --IPX -> Netware access; easier to manage than TCP/IP --IPX/SPX is the same as NWLink --good general purpose solution --not necessarily restricted to NetWare --is routable, second best choice for LANs --TCP/IP -> internet access, intranet use --requires heavy management of IP addresses --NetBEUI -> Basic, small Microsoft/LANmanager --AppleTalk -> Macs --DLC -> Directly attached printers and mainframe access ***Not routable*** --SLIP vs. PPP --SLIP is same as TCP/IP --connection protocols --SLIP is older and used in UNIX --not fully standardized --doest have authentication or encryption --only about IP --higher overhead --does not allow assignment of IP addresses --PPP --built as a standardized SLIP --can negotiate passwords and encryption --allows for dynamic IP addressing --supports multiple protocols: NetBUIE, TCP/IP, IPX --NetBIOS naming system --15 character name limitation --no minimum --NDIS and ODI --ODI eliminates need for custom driver for every protocol/NIC combination --vendors create cards to ODI standard --NDIS 3.0 -> multiple protocols and NICs --NDIS 3.1 -> adds plug-n-play to 3.0 --NetBIOS universal naming convention: \\servername\sharename\resource --avoids need to assign a drive letter --Establishing an administrative plan --network management --account management --security --any other specifics --fault tolerance --**memorize RAID levels** --NTFS --deploys all advanced security --allows for setting of permissions and share access --enforces local and network security --Differences between user level and share level --share is password/group oriented --user level is based on permissions --Results of performance monitoring --CPU% --page faults --& of disktime --CPU que length --Disk que length --If CPU is at 100% options include: --get a faster CPU --add another CPU --split apps across multiple servers --If disktime is to big options include: --get a faster disk --split disks --High disktimes and pagefaults are often due to memory problems --CPU que length --number of processes waiting for CPU time to run --shouldn't be higher than 1.5 --to large might indicate a need for multiple processors --Disk que length --number of processes waiting to be written to disk --Domains --group of computers that share a common accounts database --workgroups are federations of computers without an accounts database --create a domain with a PDC --only one per NT domain --multiple BDCs --accounts database holds: --individual computer accounts --individual use accounts --groups --primary purposes of BDC: --faster access --fault tolerance --if the PDC fails the BDC still operates but the account database is frozen until the PDC is restored or a BDC is promoted to PDC --Four domain models --Single --only one domain and PDC --all users and machines belong to the one domain --BCD is not required, but recommended --Master --multiple domains connected to each other --trust: one domain accepts the other domain's reliability --allows first domain users to log on to second domain --does not have to be two way --allows for easy local administration --easily scalable --can have local BDC of trusted PDC without actual connection --Multi-master --takes over when Master is no longer scalable --account domain is split into multiple PDCs with two way trusts --resources choose a master to trust --scaling beyond 100,00 limit --difficult to keep masters in sync --Complete --every domain trusts every other model --number of trust relations grows quickly --only good reason for a complete trust is the merging of two previously separate domains --really only a temporary solution
|