How planning for a disaster helped The Sullivan Group go green
The Sullivan Group is a Human Resource Outsourcing company based in Savannah, serving clients primarily in Georgia and South Carolina. While providing PEO, staffing, internet job boards and insurance services, we made the strategic decision to convert upgrade our PEO software at the beginning of 2008. We had been operating on the same platform since the PEO was formed in 1996, so the implementation process would affect every aspect of the business.
This change initiated a complete over-haul of our server infrastructure. It allowed the Sullivan Group to become more proactive in planning for disaster recovery and other outages that could affect our business and clients. It also allowed us to retire many of our old servers and provide a better working environment for our internal staff and clients. The costs for implementation have very quickly paid off.
With the Sullivan Group based in Coastal Georgia, we are vulnerable to seasonal hurricanes. Recent Hurricanes have missed us by a few hundred miles; they have hit the coastline north of us, near Charleston and south in Florida. Our luck won’t hold forever. Disaster recovery is critical to keeping our business running and our clients happy. We service clients all over the region and would not directly be affected if a hurricane hit Savannah. Therefore, our systems need to remain operational. Our new software platform provides expanded web access so our clients and their employees have access at anytime, day or night.
Enter server virtualization. What is server virtualization? Imagine a car engine with all its moving parts and functionality sitting on top of a chassis. Server virtualization is very similar to the chassis in a car and server virtual machines are like an engine. Whether it’s an engine for a Ford Mustang or a BMW, the chassis will remain the same. With virtualization, the hardware remains the same across each of the virtual machines (VM). Virtualization provides an emulation of the host hardware to all guest operating systems within one physical server.
There are a number of different vendors now offering virtualization software. The entire IT industry has been transformed in the past couple of years. This technology went from being very limited in its use to very widespread. With any new technology the skill levels of the external consultants and internal IT staff need to be considered. The software has been primarily used to virtualize network servers but desktop virtualization is another area experiencing huge growth. Implementing a desktop virtualization solution was not something we initially planned for but in the course of the upgrades to our data infrastructure it was an easy transition. When we need to quickly test desktop configurations it is a huge time saver. Our management staff has also realized the benefits of desktop VM’s and we have moved the entire Sullivan Group, including our remote offices, to using virtual desktops. All the desktops can be managed from a single console, and we can easily backup the entire desktop image and make copies of it. The virtualization software that we chose is based on open source software and this resulted in a by-product that is robust with enhanced security, high availability, high performance, and is comparably inexpensive.
In addition to the value, virtualization offers near bare metal performance with low overhead of resources since the VM hypervisor is just a thin a layer of software that actually runs the virtualization. This type of performance trickles down to each of the VMs; the RAM, processing power, and network speeds increase. Another benefit and driving force was the expansion of our disaster recovery plan. With our bi-weekly VM backups and our nightly data backups, we can restore to another physical server host with minimum downtime in the event there is a major business failure. We also utilize high availability features which allows for uninterrupted end user experience if there is a component failure. For instance, if the network card fails in one of the hosts, the software will redirect the VM to another working host without the user noticing a “blip” in network connectivity. These products allow us to provide a seamless production environment for our users.
Our server infrastructure was similar to other small businesses. As we grew, we repurposed our legacy hardware or bought new hardware when the need presented itself. We had 15 separate servers running different applications on hardware that was purchased over 5 to 10 years. Our backup policy kept our data safe and secure but if we had any type of large scale business disaster, it would have been difficult to quickly become functional again. It would not have been easy to pick up and move 15 physical servers to our back up location.
Our new network infrastructure consists of four physical host servers with 10 virtual servers spread across these hosts. The ability to create virtual servers whenever one is needed is a great advantage to our IT department. We can test new builds of software and create test environments with just a few clicks of a mouse. The flexibility that virtualization brings to our business is critical for our success in moving forward. It is hard to imagine now how we worked with such a rigid legacy network and server environment. Just one of our physical hosts currently supports over 40 virtual desktops. We have four physical legacy servers left to convert to VMs. The current system is stable and runs multiple web servers allowing our clients direct access to the new PEO software.
With the new system in place we can essentially “pick up and go” when a hurricane is approaching or disaster hits. Being able to load the VMs on an external hard drive makes our disaster recovery plan incredibly portable. We can (if necessary) walk out the door with VMs in hand and set up at our office in Atlanta, or at the office of a company with which we have an informal space-sharing agreement in Savannah. Without virtualization, none of this is possible.
We’ve been asked, “Does it really work?” and we proudly say, “Yes. Yes, it does.” We had only been using the new system for a couple of months when on one Thursday afternoon we experienced a failure of a RAID card in one of our brand new servers. The virtualization and HA software performed exactly how it was designed to work. The HA software transferred the users and the data access to the mirrored server and the users did not experience any type of interruption. The software we chose allows the transfer to happen automatically with no intervention or manual steps required by the IT staff. Often in IT, when a problem occurs, fixes are done in a hurry so users can carry on with their work. These fixes may or may not be a permanent solution depending on the amount of time the IT staff has to correct the problem. With this secondary mirrored host running the application, our IT staff can take their time to diagnose and fix the problem while production continues uninterrupted. This HA technology gives IT the time they need to properly resolve the issue.
Implementing virtualization software allows companies to reduce the number of servers running in their data closet;, this change helped The Sullivan Group go green. Prior to virtualization, we were experiencing “server sprawl” which is the industry term for companies with servers that are generally underused and wasting space and energy. It is generally recommended that new software be implemented on a separate server. Even a small company can house 10 or more servers after a few years. Smaller companies commonly do not consider the power and cooling requirements that is required for a data closet. With the reduction in the number of physical servers, we have seen the temperature in our server room decrease which has cut down the power needs and AC usage. Also, reducing the number of servers reduces the amount of physical space required in our IT closet. We can grow our business without the expense of more power to cool the room or even building a larger room.
The Sullivan Group