Do you have a project coming up and are in need of technical support?
Call on us to design a Solutions based approach to your project.
TSS has managed network project installations from as small as a single server
upgrade to entire enterprise infrastructure upgrades. All our projects are
handled by a team of experienced network professionals who will design, manage
and install every aspect of your update/ upgrade. We focus on milestones,
and turn over complete network documentation at the end of the project. |
After my contracts with GE Capital ITS, and the start of the economic downturn, I decided to take a longer contract and full time opportunity working for a company in Middlebury, CT.
They were once a leader in ATM hardware, and was poised for eithor a heck of a recovery or complete collapse. I took the position to get into some longer term and more stable work, and ended up in one of the greatest technological challenges of my career.
One would think that they would have taken care of it's IT network (since telecommunications was its primary business) but the network was anything but. They contacted me to fill a network admin position. The network consisted of a mixed environment with old Cabletron hubs, and CrossComm bridges for the WAN. The network infrastructure was undocumented, consisting of 21 remote sites connected via straight frame (64-128kps) links. The corporate and engineering Lans were converged, causing severe bandwidth and resource constraints. In fact, some of the engineers actually had used modems to dial into RAS, simply because performance was much more reliable!
The internal LAN was severely unorganized, with servers sitting on racks and no one knowing their function. IT had recently recabled the server farm to increase the reliability of the links. High priority servers were given 'home runs' into the Cabletron SmartSwitch 6000 core switch. There was no attention to detail, not one single patch panel or management arm. Cable ties were used to move unknown cables out of the way. In a nutshell, the LAN was a mess!
|
PROJECT: CORE LAYER SERVER REORGANIZATION
My first priority was to document the LAN, establish a monitoring station and to begin investigating link reliability. While a daunting task, I jumped right in, and through 5 minutes of a Sniffer capture, I decided to start replacing the physical infrastructure immediately. This benefited me enormously, as it helped to immediately and definitively document the LAN infrastructure, to remove the foot and a half (literally) extra obsolete Thick LAN and token ring cables, and to define a strategy for the LAN reorganization.
Click on the photos below for a larger image.
|
Picture of typical server rack. This was engineering and CAD file storage - determined as critical. The green Cat5 lines were from IT recabling the servers prior to my arrival. |
Some other racks. Note the Green Cat5. |
Same rack from above right. Note the difference in the hub with connectivity. The remaining servers were given 'home runs.' Non essential servers were given to the hub. |
Close up of the same rack |
This was a secondary office. Towards the bottom of the rack is a CrossComm router. The Sun box you see on the shelf had a script on it to ping the router for availability! Notice the hubs below it. I pulled over 35 pounds of unused cable (actually weighed) out of this rack alone. This link was constantly plagued by 37% unavailability, 90%+ link utilization (768 kps). |
Same rack after I reorganized the room. Existing budget only permitted replacing the router only, so I placed a 2611 with 4 Ethernets each going to the existing hubs. The link was upgraded to a full T. IT monitoring of the sun box showed an immediate improvement with 99.99% availability, and 65% sustained link utilization. |
Unfortunately, I didn't get all the photos after the reorganization. I led a team of 5 - 7 tireless individuals who put in many hours nights and weekends, in gutting out the existing unused cabling, installing all the new cat5e in shielded conduit (sub floor) to patch panels, dedicated 'home runs' to all L3 and L4 critical servers, replacing all other hubs with switches, as well as organizing the color scheme. We used Green for home runs, Red for backup, Purple for monitoring, Black for cross-overs, Gray for intra-patch and White for remote offices.
The end result of the server reorganization was to up the server bandwidth to switched 100 Mb (previously most servers were on a 10 Mb L1 blade of the Cabletron). We removed
approximately 100 L1 (hub) ports from the server farm and replaced them with dedicated L2 (switch). All bridges were reconfigured to true L3 (routing). Monitoring via HP Openview and Spectrum Element manager was initiated. Vlans were used to further isolate the departments, though that project was never completed.
Server availability went from 86% to 99.99% within 6 months. User throughput was severely increased, and the response was overwhelmingly positive. IT trouble tickets declined over 50% in that time - so much so that IT was reorganized and divisionalized into Level 1, 2, 3 and 4 support! Broadcast storms ceased to be a daily nightmare. |
PROJECT: CORE DISTRIBUTION AND ACCESS LAYER REORGANIZATION PROJECT
Once we completed the Core layer and Data Center Reorganization project, we tackled the distribution and access layers. Traffic throughput had already been significantly improved, so changes to the distribution layers went mostly unnoticed.
The corporate and engineering LANs consisted of 19 Cabletron L1/L2 SmartSwitch devices running in 'Securefast mode'. Strangest devices I have ever come across. 12 of the SmartSwitches had 4 switches ports, each with 12 Layer 1 ports or hub'd. Each of the four switch ports shared 10Mb of throughput, with 100Mb back to the core with no redundancy.
I started to meet with the engineering and CAD development group department heads, as well as countless users. Imagine 40-50 engineering users, each with a Sun workstation and a pc, some also had a laptop, vying for 40Mb of throughput! Typical user response was less than 56kps to the desktop during peak times! Engineers actually used modems to dial into the network from their desks because of increased speed and throughput via RAS FROM THEIR DESKS!
I immediately started certifying their drops, to find them far below Cat5 specs. In the end, I decided to rewire most of the users to bring them up to 5e specifications. Unfortunately the budget wouldn't allow a one for one rewire project, so we replaced the
SOHO hubs on their desks with SOHO switches. We then started to replace the Cabletron SmartSwitches with Cisco Catalyst 3548 switches.
|
Return to our previous projects
|
