Strong network design experience: - quantifying & qualifying requirements
The client/server network design goals I use include minimizing cost, maximizing quality of service, and maximizing growth capabilities. A very good design strategy for ethernet LANs is to keep average utilization below 20% and co-locate servers with their LANs. This strategy maximizes throughput to the server/s and minimizes delay for users. Current low cost network technology allows us to bring the necessary bandwidth to users and servers via high speed ethernet switches, wiring hubs and router/s. The measure and adjust phase is continuous - I use nettools to gather SNMP data from hubs, routers, servers to calculate utilization of LANs, servers, and WANs. I then make adjustments to network topology, router buffer efficiency, and protocol efficiency.
Integration & interoperability of multivendor equipment
When setting out to build an enterprise internetwork supporting various logical (TCP/IP, IPX, DECNet, Appletalk) networks, you must select those vendors that have proved interoperability at both the physical/logical level and the network level. I like the INTEROP conference demonstration floors and have found that Cabletron wiring hubs, cisco and Network Systems routers, and Artel ethernet switches work well together.
Workstation/server (including UNIX) sys. administration
I designed a universal file system (using NFS ) for clusters of workstations. The achieved goal was to make the file structure look the same for each workstation. I did all the usual sys. administration such as set up accounts, domain name serving, sendmail, backup/s, and diskquotas. My emphasis was on developing cshell scripts to automate the repetitive functions.
Both small and large projects
I developed an SNMP monitoring and data collection scheme for all routers. This data was used to plot daily useage of all LAN and WAN links. I used MicroSoft excel spreadsheet to generate the plot files and NCSA Mosaic to make the plot files available to any user with Mosaic. RMON statistics are also used to generate the LAN utilization plots in real time. When I designed and managed UTRC's network upgrade, I began with a benchmark of the current network. The network consisted of a single ethernet backbone running through five buildings and some wiring hubs - the utilization was 30% with delays of several seconds for terminal session users. The client/server network design goals included minimizing cost, maximizing quality of service, and maximizing growth capabilities I selected a star network topology for the new network - at the center is a very high speed cisco 7000 router (there is also another acting as backup). The spokes are multmode fiber connecting the router to the wiring hubs located in the five buildings. This configuration allows me to run ethernet over the fibers today and positions the network for an ATM configuration upgrade. Each separate LAN spoke runs 15% utilization and one second delay with its own Novell server.
Installed and configured cisco LAN hub router/s, Network Systems FDDI routers, and several cisco WAN synchronous and framerelay links. All router configuration files are stored on a network host for ease of modification and backup. I designed and configured UTRC's Internet security access procedures which include access control list filters on the input/output router links. Installed ATM testbed at UTRC - includes a switch (FORE System's ASX-200) and several SGI workstations
Topology options for constructing LAN, MAN, WAN LAN
Topology options include bus (ethernet), ring (FDDI, token ring), star (ethernet), mesh (ATM). MAN topology options include (FDDI, SMDS, ATM). WAN topologies include dedicated, switched or dialup links at various speeds and committed information rates.
Demonstrated state of the art technology experience
In addition to hands on experience with these technologies I teach two courses on networking for Learning Tree International two times a year. Each course is four days in length. The courses are; Hands on TCP/IP Networking, High-Speed Networks: ATM, SONET, FDDI, SMDS.
Physical layer (mm & sm fiber, UTP, TIA 568)
mm fiber is used for distances up to 2 km depending on the protocols used (ethernet is restricted to the 1500 meter end to end propagation delay while FDDI and ATM can use the full 2 km). sm fiber is good up to 20 miles for both FDDI and ATM and UTP level 5 has been acknowledged as the media of choice to the desktop up to 100 meters from the hub. UTP level 5 will handle ethernet, FDDI, and ATM (tested up to 155 Mbs). Note that the available bandwidth in fiber is in the Gigabits and network research laboratories are demonstating the capability to multiplex data from various sources on/off the fiber at these speeds using the next generation switch (wave division multiplexor). I am currently involved in managing an ATM campus project - the first step is determining if we can use ATM SONET over multimode fiber for distances greater than 2 km.
Low level LAN protocols (IEEE802.1-6, 100BaseT, FDDI, ATM)
These protocols use MAC (media access control addresses) to uniquely identify a network host no matter what higher level protocol (TCP/IP, IPX, DECNet, Appletalk) is being used. IEEE802.2 specifies the logical link control while 802.3 specifies 10BASE5(Standard Ethernet), 10BASE2(Thin Ethernet), 10BASET(Unshielded Twisted Pair). 802.5 specifies Token Ring where the links are point-to-point. FDDI is a token passing protocol that includes two counter rotating rings (one is for failover) - 802.5 provides for 20 Mbs and FDDI specifies 100 Mbs. 100BaseT is the specification for 100 Mbs ethernet over copper or fiber. Unlike the shared medias (ethernet, token ring, FDDI), ATM is a cell switching bandwidth scaleable media. Currently available bandwidths are 45 Mbs (T3 WAN), 155 Mbs SONET LAN / WAN, and 622 Mbs LAN. I have set up an ATM testbed at UTRC consisting of a FORE switch (ASX-200) and several workstations using mm fiber and UTP L5. I am now trying to extend the ATM LAN beyond 2 Km. I also designed and managed the deployment of a special purpose (visualization, distributed computing) FDDI ring.
Low level point-to-point protocols
V.xx async specifies the protocols for analog modems to achieve bit rates higher than 2400 bps (i.e. 6 data bits per sample rate yields 2400 x 6 = 14,400bps) and error correction MNP levels. SDLC (IBM) includes the clocking signals between CSU/DSU in order to achieve data transfer rates such as 56Kbs, 128Kbs, etc. PPP was designed to solve interoperabilty between router vendors - lately PPP has been adopted by terminal server vendors to provide async. dialup TCP/IP connections from other terminal servers or remote PCs / MACs. Frame relay is packet switching without the error control overhead, two types of virtual circuits are supported. They are permanent virtual circuit (PVCs) and switched virtual circuits (SVCs - currently under development). Frame relay is designed as a NISDN (9600Kps - T1) for LAN to LAN.
Network protocols (TCP/UDP, IPV4&6, DECNet, VINES IP, SNA)
The transport control program (TCP) is the part of the TCP/IP protocol that is responsible for full duplex, reliable (flow control) and error free delivery of data for multiple virtual connections. TCP is a connection oriented protocol. The user datagram protocol (UDP) is a connectionless protocol (unreliable). UDP has no overhead of virtual circuit maintenance - is suitable for applications that need message type responses (database request/response, network time protocol, and broadcasting such as RIP). IPV4 contains the multicasting standard for IP routers and IPV6 contains the standard for host automatic discovery of its IP address from a router via broadcast. Note here that MicroSoft has recently announced it DHCP automatic ip address discovery which works. DECNet is Digital Equipment Corporations' networking protocol - it is has both hop based and cost based routing but its addressing structure is severely limited with only 64 separate areas. -network services (DOD suite, sockets, streams, DNS, NFS, SQL, SNMPV2) The DOD suite (TCP/IP) mandated a common set of protocols in order to promote interoperability and competition. TCP/IP is a group of protocols which include transmission control protocol (TCP), user datagram protocol (UDP), internet protocol (IP), file transfer protocol (FTP), terminal emulation protocol (TELNET), etc. I setup the primary DNS server for UTRC and directed a UTC wide DNS servers architecture so that the servers cooperate in naming and email delivery. I have installed and am using the V1 features of CMU's SNMPV2 - so far I have not found support for the encryption standard in SNMPV2.
Active LAN components (routers, switches, hubs)
I use SNMP & RMON to gather, plot, capacity plan for the overall network. I configured all the UTRC routers (cisco AGS+'s, 7000's, Network Systems 800 Mbs, Wellfleet) and am now managing two people who are taking that over. -active point-to-point components (FEPs, muxs, controllers, DSU/CSU) no experience here except for DSU/CSU's for sync.circuits connecting to router/s. -security (IEEE 802.10, Kerberos, IP firewalls) I designed an IP firewall system for UTRC's connection to the Internet using access control lists on our cisco router, a DMZ LAN with a packet-filtering bridge, an alarming and logging system, and proxies for telnet, ftp, Mosaic. Remote access from the Internet and through terminal servers all reference a SecureId server on the internal LAN - this means access is controlled via the user's SecureId card which has pin pad keys.
Proven troubleshooting skills: -general proficiency (LAN/sync. analyzers, TDRs, deductive reasoning)
I use our LAN analyzers and Fluke meter to capture and bust packets. I use them for traffic analysis - it was useful in finding an SMTP loop between two misconfigured sendmails. The Fluke meter is also useful in generating artificial loads on a LAN.
Principled in fundamental architectural concepts: -when to switch vs when to route
An ethernet switch operates at the network's physical layer and provides a large backplane bandwidth for dedicated 10 Mbs connections (for example, an 8 port ethernet switch will have at least a 40 Mbs backplane in order to support 4 dedicated conversations). Also the switch will include an ethernet bridge for connection to the enterprise network - hence all network devices running TCP/IP connected to the switch must be in the same network to minimize traffic. Routing uses the network's network layer and involves the creating of separate LANs - LAN broadcasting is efficiently controlled in this environment.
How to route (distance vector vs link state)
The distance vector is used to pick the path through the minimum number of routers. The link state information is used to select active links and skip the down links. In the case of TCP/IP's RIP routing protocol, routers exchange their routing information (their active links and associated networks) and build a topology of the entire network. If a link goes down in the network, the routers will nform each other and converge to the "new" topology. Since RIP does not have a notion of transmission cost based on link bandwidth, I have artificially adjusted link hop counts to accomplish the link biasing.
Advantage of switched vs permanent virtual circuits
I have used permanent virtual circuits (PVCs) on frame relay - this is not too bad from a network management point of view. ATM PVCs are very difficult to set up and manage - I would certainly not entertain this solution for a LAN! SVCs are established between network communicating nodes using a standard signaling protocol - the nodes establish a minimum and maximum bandwidth and quality of service (time critical data requires a higher QoS).
Advantages and disadvantages of flat vs hierarchical networks
A hierarchical network structure contains more network elements that a flattened one. Most corporate organizations are going through re-engineering which results in a flattening of the network. There are fewer network elements with greater logical network interconnection. The result is greater horizontal communication and interaction within the organization regardless of geographical location.
What functionality to expect of network components vs network hosts
I expect the following functions from network components - routing, bridging, switching, packet filtering, alarming, encryption, terminal serving, structured wiring (hubs), monitoring. I expect the following from network hosts - proxies, file servers (including Mosaic), email hub, application security, network access security, SNMP monitoring, network application servers (time, irc, DNS, SMTP gateways, directory server).
I teach for Learning Tree International twice a year
I have tested on the Myer Briggs Type Indicator as ENTJ - Extroverted Thinking with Intuition. I enjoy new situations and bringing about constructive change.
Test lab and/or simulation traffic modeling
I have configured a test lab with router, traffic generator, monitor host. Used to study maximum acceptable LAN ethernet collision rate.