When virtualized applications experience performance problems, the complaints are always forwarded to IT administrators like Fritz Gielow. Fritz is the storage administrator for the County of Nevada in California, and he was having those “calls” come in too often. A change was needed. For Fritz, the answer to these storage performance problems was a hybrid flash array from Tegile. Since the Tegile solution, those complaint calls no longer come in because users and applications are getting all the performance that they need.
The County of Nevada is known for trees, clean air, hiking, biking, camping, and plenty of history, but like any other business, the County needs flexible IT services to keep the local government running smoothly. The County’s IT team services more than 25 county departments and other local government entities, each with its own unique IT demands. Its mission-critical applications include their finance system, Exchange mail system, the tax system and then there are individual departments with their own applications. Most of the applications are Microsoft SQL Server based—and all of them are virtualized.
After upgrading to VMware vSphere 5.1, these various departments started to complain about response time from their applications. It was going from bad to worse. For example, in SharePoint the user would click on a link and have to wait and wait for the attached file to open. Their finance system exhibited a similar behavior, and users were constantly complaining about the application being sluggish. The complaint calls were coming in fast and furious, all ending up at Gielow’s desk.
Troubleshooting any storage performance problem is always difficult. Add to it the layers of abstraction that virtualization introduces plus customized applications, and the task is almost impossible. Despite the size of the task, Gielow jumped in. He determined that the physical CPU and memory resources were not the sources of the performance problem, so he turned his attention to his storage system.
Their array was an EMC CX4-120 array, a hard-disk based system, and the County was connecting to it via iSCSI. Running an analysis on this system showed that it was exhibiting latency spikes of more than 20 milliseconds. With mounting performance issues, more frequent drive failures, the EOL status of the EMC and increasing capacity demands, the County decided to look for other options.
Selecting a New Storage System
Other than the desire to make the performance calls go away and keeping costs in check, Gielow did not go into the selection process with a long checklist of requirements. Each vendor was free to highlight its capabilities. The County did feel that the best way to address their performance problems would be to leverage flash storage in some way, so they looked at both hybrid and all-flash storage vendors. The County quickly ruled out all-flash. While it would eliminate the complaint calls, their performance sensitive applications were relatively small in capacity. They felt a hybrid solution that provided a large flash tier could solve most of their storage performance problems. As a result, they narrowed their search down to hybrid vendors.
Why Tegile Won
There were several key differences between Tegile and the other hybrid vendors that Gielow investigated. First, Tegile could provide both inline deduplication and compression. Having access to both of these data efficiency technologies was critical because of the amount of SQL Server databases that the County runs. Databases, in general, don’t deduplicate well, but most benefit considerably from compression. Another important aspect of Tegile data efficiency capabilities was their ability to provide deduplication and compression on both the flash tier and the hard disk tier. While HDD capacity is inexpensive, extracting maximum value from it was important.
The Tegile system that the County uses has 45TBs of capacity. As expected, their Server environment did not deduplicate well, but it did compress, saving about 30% of the SQL capacity requirement. The non-SQL data does deduplicate well and is seeing a 60 to 70% savings. The key though is the net impact of the combined technologies, on the 28TBs of data the County has, they are only consuming about 16TBs, a savings of 12TBs of capacity.
The second differentiation was Tegile’s active-active controller design. Most enterprise systems come with two controllers, but how they actually use those controllers varies. It is easier from a design perspective to make one of the controllers passive, meaning that it is in standby mode until the primary controller fails. The advantage of an active-passive design is that performance remains the same even in a failed state; the downside is that that second control lays idle when in a working state. Controller failure is very rare, and so the amount of time that a passive controller will sit idle is extremely high.
An active-active design enables both storage controllers to be active. Instead of assigning all the workloads to a single controller, they are split up between the two. Under normal conditions, the base system can support more workloads and deliver better performance. The downside is that if there is a failure, performance may suffer by as much as 50% until the second controller is fixed. Again, controller failure is a rare occurrence; most IT professionals like Gielow will choose to use the second controller instead of having it sit idle. If there is a failure, Tegile’s support organization can have a replacement controller out to the data center quickly.
The third differentiation point was the quality of the interface, both regarding ease of use and day-to-day monitoring. In the days of the EMC array, they were forced to perform many functions via the command line or wait until EMC could help them through the task. Today, with Tegile almost 100% of their administration tasks are done via the web interface. Gielow also likes how well the Tegile interface lets him know what is going on with the system. He can easily monitor current performance load, capacity utilization, and resource consumption.
The final and most important point of differentiation was the cost. According to Gielow the Tegile solution was just priced better than the competitors when comparing usable capacity after factoring in real-world experiences of compression and deduplication across the potential solutions. Gielow said it best; “Tegile’s technology was hands down the best, but the cost per GB was the deciding factor.”
In the final analysis, Gielow feels he has selected a solution that not only addresses the pressing performance needs of the environment but that also reduces storage operating costs. Users just don’t complain about performance anymore. He feels that this is a system that will last the County for years to come and enable it to build a better IT that will improve the service the County can provide to both its residents and visitors.