Operating out of a modern 165,000-square-foot facility, Coastal Sunbelt Produce has emerged as the leading food-service distributor of produce and dairy in the U.S. Mid-Atlantic region. According to Director of IT Richard Kreuzburg, Coastal Companies has experienced a great deal of growth and plans to expand its infrastructure considerably. With people using home laptops, iPhones, Android devices and iPads—anything they can get on the network—they needed more insight into the health and performance of the server applications that were running and to get a reliable measure on the end-user experience of the virtual desktops.
Coastal has a blended environment of VMware Horizon with View desktops and vSphere-based virtual servers. The combined workloads were accessing a mirrored pair of NetApp storage systems using the iSCSI protocol and, as a result of the storage system trying to satisfy the demands of the combined workloads, datastore latencies started to climb. This situation is not unusual in a virtualized environment because even if the virtual servers themselves were doing mostly sequential read and write operations, by the time they get multiplexed together by the hypervisor, the traffic ends up looking like random 8KB reads and writes. Such a situation is a worst-case scenario for rotating media, which prefers large sequential accesses for maximum performance.
The network-sniffing tool the company was using showed that everything was running smoothly with no bottlenecks, but that couldn’t have been further from the truth. Although the LAN may not have been stressed to its tipping point, login times were tripling and people were starting to complain, while server applications were failing to meet their SLAs. The network-monitoring tool that Coastal was using “let us know that the network was up and running, but it could not pinpoint the bottleneck,” according to Richard. They needed more insight into the health and performance of the entire end-to-end infrastructure including servers, switches, applications, client devices and users to ensure that server applications and VDI desktops would meet the expectations of their users.
Richard used the Xangati VI, VDI and Network Suites concurrently for a clear view into the entire infrastructure and to identify any resource-contention storms. By factoring together metrics from the entire infrastructure, using root-cause and predictive analytics as well as prescriptive remediation, Coastal Companies gained a clear view into the entire infrastructure and could easily identify any resource-contention storms.
While encountering a real issue, a conventional siloed network-monitoring tool that the company was previously using showed everything that was “green.” But the Xangati dashboard, which factors together metrics from the entire infrastructure, indicated that datastore latencies were regularly climbing from a healthy 10ms to over 1,000ms. Richard used root-cause and predictive analytics, which clearly showed that the controllers of the NetApp storage system were being overwhelmed and were not able to deliver the necessary IOPS to meet the demands of the virtual servers and the virtual desktops.
He moved the virtual desktops to a second NetApp device to see the performance of the virtual servers move higher and higher, meeting their SLAs consistently. This result gave Richard insight into higher-level concepts including virtual desktops, the users of those desktops, and the client devices and networks being used to access the virtual desktops. He could now verify that the end user’s experience was snappy and responsive. Every time a threshold was crossed from healthy to sluggish, a recording was captured that clearly indicated the cause of the resource-contention storm. On the basis of this data, Richard was able to identify over- and underutilized resources to guide him in rebalancing workloads to get the most from his existing infrastructure, as well as to help him justify the purchase of additional hardware resources when necessary. Also helpful were predictions of the exact time that a particular resource will start to exceed its traditional thresholds, enabling future problems to be averted.
CFO Now Sees VDI’s Needs
Richard was able to clearly show his CFO that a second dedicated NetApp storage system was needed for the VDI desktops. Richard’s reports showed clearly that adding additional hypervisor host servers or beefing up the network switches would not improve the overall performance.
As many an IT professional would attest, once the virtual-desktop limit rollout is proven and end users are happy, the journey has only begun. More often than not, the virtual desktops will roll out more widely so more users can get the benefit of any device anywhere and anytime, while IT can secure corporate data and benefit from central administration. The new architecture will ensure that users are not inconvenienced as additional virtual desktops or virtual servers are added.
Richard was able to work with his CFO on the basis of data rather than just throwing more hardware at the problem. In many cases he was able to rebalance existing resources rather than make new investments by using the insights provided by the analytical reports.
“Having a history of your IOPS, latencies and throughput on our virtual desktops and using that as a baseline, you can look at your potential capacity and gauge what your new capacity should be in the future,” said Kreuzburg. “When you leverage root-cause and predictive analytics you get an accurate picture of what you need—and that results in faster and better IT purchase decisions.”