LV 1871: Taking Virtualization to the Next Level

Specializing in life insurance, annuities and disability income protection products, LV 1871 (Die Lebensversicherung von 1871 a. G. Munchen) is a mid-sized, Munich-based insurer with over 8,500 independent agents throughout Germany. Fitch Ratings gives the company an A+ rating, and agents have given top marks to the quality of the company’s products. Nevertheless, LV 1871 competes with many large players in its industry and also understands that the quality of what it does must go well beyond its products.

“Ninety-eight percent of our business processes are built on IT processes,” said Alexander Triebs, LV 1871 Infrastructure Project Manager. “Processes for business and infrastructure for IT must be fast, reliable, flexible, robust, cost effective and efficient.”

That wasn’t the case seven years ago, when LV 1871 had a mix of hardware and systems in its data center, that included a classic inventory management system running on legacy hardware, a mix of Unix, Linux and Windows systems running front-end and back-end operations, and physical storage where an average of 600 new insurance policies per day were added to its existing repositories. “We wanted to grow and to separate ourselves from our competitors in our market,” said Triebs. “We also recognized that we wouldn’t be able to grow the way we wanted to if we tried to implement new sophisticated products in COBOL and ASSEMBLER, and running our systems the way we had always run them.”

Triebs and his team began talking about implementing other programming languages like Small Talk, Java and Delphi—and about a first-stage virtualization plan for the data center that would migrate LV 1871’s legacy system to an IBM System P AIX (Unix) platform, where additional data center system consolidations (through virtualization) could be performed with the System P as a target platform. At the same time, many of the data center’s services were being outsourced to contract IT providers. LV 1871 wanted to regain control over its IT processes by in-sourcing them, which would give it greater ability to reduce its IT costs.

“Moving our set of programming languages to more flexible languages was a major effort for us that took nearly seven or eight years,” said Triebs, “But we recognized that undertaking it was the only way that would assure us an IT infrastructure that would support the growth we wanted and expected in future years. We had internal expertise in Unix systems, and we knew that with rapidly changing markets with changing products and regulatory requirements, that the old code would not provide the agility. We also knew that we wanted a mid-range, industrial-strength Unix system (in this case, AIX) to take on this kind of workload.”

LV 1871 succeeded in virtualizing its servers and moving its application code base, but it felt it could take virtualization even further to benefit the business. “Our storage was principally network-attached at this point,” said Triebs. “We believed that we would improve both IT cost ratios and our computing performance for the company and its agents if we virtualized storage, and that this would also improve our capabilities of failover and data mirroring.”One of the data center agility concerns was the time and effort it took to re-provision storage with LV 1871’s network-attached storage orientation. “With the SCSI-attached storage, we did not have a dynamic infrastructure that would allow us to quickly re-provision storage when we needed to,” said Triebs. “Instead, we had to concern ourselves with the model of the hardware, the space required, a buy decision and finally implementation. This end to end process could take as long as six to twelve weeks, and it was an expensive use of internal IT resources.”

LV 1871 made the decision to invest in an initially more expensive SAN (storage-attached network) solution that would pave the way for a virtualized storage framework, where different storage media could easily be tiered into fast access, more expensive disk and lower access, cheaper disk—with deployments and provisioning being accomplished in a matter of minutes, not weeks. “The virtual storage backbone not only reduced our internal costs and our speed of response, but it also allowed us to improve our failover and backup mechanisms for our two separate data centers,” said Triebs. “Data mirroring between the two data centers now takes minutes.” The virtual storage backbone has dramatically improved performance. Virtualization has given LV 1871 vendor independence as well, which lends more flexibility to buying decisions. Additional return on investment (ROI) is being seen in the new tiered storage strategy with its reduction of wasted storage space. “It costs us roughly $20/GB for faster access, tier one storage, while tier two, slower access storage costs around $8/GB,” said Triebs. “In our new tiered storage structure, we find that only one-third of our data is constantly accessed and needs to be on tier one, and we have organized our data this way. This is a primary area where we are realizing data center savings.”

Along the way, Triebs and his staff learned valuable lessons about working with virtualized server and storage infrastructures. “One was a simple practice to remember to delete virtual machines when they were no longer needed,” said Triebs. “On the data side, it is also imperative to think about technologies like data deduplication before doing backups, so you do not store extraneous data. Finally, when you consider going to a LAN/SAN virtual infrastructure as your backbone, you need to consider architectural concepts, such as a split fabric with the use of virtual LANs (VLANs)—and when it comes to security you want your DMZ and LAN to be separate from each other, and hosted on separate hardware.”

With the groundwork for its data center set, LV 1871 continues to move forward with IT that supports its products and services. The results so far? “We provide basic services to our customers (e.g., Storage, SAN, System p (AIX) and System x (ESX) with a 7x24 availability of six 9’s (99.999999 percent) that we are meeting for possible transaction processing uptime,” said Triebs. “Our aggregate service level uptime for over 3,000 services that we provide to over 30 separate customer service groups, and that also includes inventory management and customer relationship management (CRM), is 98.5 percent concurrent availability, which means 99.9 percent for each service group. The entire operation for basic services (Data Center Infrastructure: Storage, SAN, LAN, Server-hardware and Operating Systems) is manned by only three IT staff in our data center….We have found that by systematically virtualizing, consolidating, reworking job control and backup, and establishing metrics and performance targets - thatwe can economize staff and infrastructure costs.”

See more on this topic by subscribing to Network Computing Pro Reports Hyper-V Takes on VMware (subscription required).