|
Exelgen Discovery as based in the old Tripos buildings in Bude, Cornwall - inheriting a complex network of fibre-optic, gigabit ethernet, and server rooms full of non-functioning servers needing a complete infrastructure rebuild, and ongoing system administration.
Infrastructure Rebuild
Working on a tight budget, the goal was to get the production facility operational once more utilising where possible the existing network. This was a big build, the facility includes miles of fibre-optic cable, ten times that of copper, scores of dead servers, and over a thousand network drops.
Here’s how it went:
Step 1 - provide internet, E-mail, and web site
Staff needed E-mail and internet as soon as possible, an expedited ADSL installation coupled with a small workgroup switch and basic Netgear ADSL router linked to a quality, combined ISP-provided hosted mail, domain, and web site.
Step 2 - re-network the building
Fibre and copper links were tested, and then linked with core/edge L2/L3 modular HP network switches - providing fail-over multi-gigabit links between the buildings. Over 100 network drops were connected to the switches - for the various laboratories within the buildings. Layer 3 routing was accomplished with HP switches, and Cisco and Netgear routers providing a blend of performance and security. VLANs, trunked from building to building subdivided the network’s broadcast domain.
Step 3 - provide network services
A Windows Server Domain was established providing File and print sharing, security, and network services needed by client PCs, such as DNS, DHCP, and WINS. Remote access was provided and linked using RADIUS to allow staff to access their files remotely. Network printers were installed to allow quality printing from any location. A 1.5 TB external SAS storage array was tested and commissioned providing extremely high-performance file access. A high-performance SQLServer database was installed and configured for access to the line of business application, ChemCore.
Step 4 - prepare for failure
Backup schedules were setup to use external USB devices and the storage array to cater for a variety of disaster recovery scenarios using Veritas BackupExec and Netbackup Pro. The RAID 1 System drives of the 5 core servers were imaged with Acronis TrueImage, and a second Windows DC was added for Active Directory resilience. Servers were moved to UPS-protected power.
Step 5 - Build some PCs
With over 70 decommissioned PCs from which to chose, a mass-build and rollout of PCs was undertaken to provide identical Desktop PCs configured with all the needed Office and scientific software. Matching hardware was chosen and tested using system-stressing burn-in tests and benchmarks to ensure the PCs would be reliable and fast.
Step 6 - Thin Client PCs for the Laboratories
Using a PC in a laboratory is impractical, so dozens of thin-client PCs were used linked to a new, beefy HP Citrix Server using Citrix Metaframe XP.
Step 7 - Link the Sites
Using Cisco routers, and IPSEC, the other laboratories in Bath and Camelford were linked so that the three sites could seamlessly share files, data, and access the Chemical Inventory database system.
System Administration
We design network and server solutions to be reliable, and two years later, the approach taken in the Exelgen Discovery network rebuild has proved reliable and resilient, with no unplanned network downtime in the period other than:
- British Telecom ADSL problems - what’s new ?
- A power surge that fried one server that was not UPS-protected; it was quickly rebuilt from backup and no data was lost
The server’s and automated systems are mostly self-monitoring and just run smoothly and our system administration involvement is weekly remote monitoring for any signs of problems, almost all of which are caught automatically by the various firewalls, anti-spyware, and antivirus.
|
