AT OETIKER+PARTNER AG we use wherever possible the same systems and concepts that we recommend to and provide for our customers ("eat your own dog food"). In this way it is possible to know firsthand where problems lurk and which improvements are necessary. We are thus experts on the systems and can support our customers even from a user perspective.
The IT system is supposed to be an "all-rounder":
- Developers need standardized reference systems on which programs can be implemented and tested under defined conditions.
- Our administration needs a stable current environment.
- Supporters need systems where the behavior of the customer systems is reproducible.
- The system should be usable location independently, so that the work location can be chosen freely.
- There must be free resources in the system, so that we can react flexibly and rapidly to new demands.
- The platform should be selectable as needed.
The system is built with the following components:
We use a TerraServer4 system provided by DALCO AG. The system has 26 disk slots running off four 8 port LSI controllers. The system is equipped with SAS disks with additional SSD disks for caching.
On this hardware we run an OmniOS operating system with the ZFS file system. This system provides us with redundant and secure (checksums) storage services to all our systems (SMB, NFS, iSCSI).
Our company head quarters in Olten is connected to the internet redundantly via links by UPC cablecom (250/15 Mbit) and Init7 (25/2 Mbit). For the internal network we use manageable switches and routers, so that we can intervene in case of problems.
The routing is configured so that we can route traffic over either link depending on the application. Because of the two providers currently only Init7 offers IPv6, we set up an IPv6 tunnel to our servers in the hosting center of Init7. This gives us IPv4 and IPv6 capabilities on all connections.
We use our own open source projects (MRTG, SmokePing, Extopus, RRDTool) to monitor all components to detect and fix irregularities quickly.
For our own use we operate the following servers:
- Linux Server ( Ubuntu ) with ThinLinc from Cendio as desktop Linux terminal server system
- Collaboration Server Zimbra for eMail/calendar/contacts Integration accessible on handheld devices and tablets
- Windows Terminal Server for MS Office and Windows applications
- Linux Server ( Ubuntu ) with database services ( MySQL, PostreSql ) and webserver ( Apache ) and with additional eMail services ( Dovecot, Postfix)
- about 20 to 30 virtual servers with a variety of environments for development, testing, and integration
Most of these systems run in a virtualization environment, mainly KVM and partly VSphere.
As desktops we use HP T610 thin clients from HP. The thin clients run via PXE and NFS of an Ubuntu diskless root system and operate without a local file system. The automatically started ThinLinc client then accesses our ThinLinc server and makes the terminal session available to the users.
The system has the advantage that the users' sessions can be accessed location independently. Thus, an employee can also connect from home or on the road in his/her session.
Traditional fixed-line telephony, mobile phones, eMail, SMS, etc. are available for all employees and customers.
In the support area, we operate a Request Tracker instance assigning to each request a unique tracking number. The processing of the request is thus documented and all information can be accessed as needed by another employee.
We use two systems for data backup:
- We operate our own Crashplan server on our systems. The client can be run on most popular platforms and is manageable by the server. Through block-level deduplication, the data quantity is almost constant even with frequent backups.
- Thanks to ZFS functionality, to send differential snapshots to a remote system, we can achieve redundancy of systems. On failure of a system, the corresponding service can be transferred to the remote system.
Locally we use ZFS snapshots for all other cases.
Impact and benefit
Due to the flexibility of the system it is possible to set up a test system or proof of concept in a few hours. The diskless system makes it possible to install updates without affecting the running operation.