| TBD Networks

Storage Latency is the #1 Suspect

Building a VMware-based private cloud comes with a specific set of storage challenges resulting in prevalent storage latency issues. Approximately 70% of performance issues in VMware environments are driven by storage latency. This due to three reasons:

Limited visibility
From the storage side, it is hard to distinguish storage IO from different VMs. From the VMware side, however, the resource competition that different VMs are creating when they hit the same datastore is hard to see. Therefore placement decisions are frequently made based on available disk space not based on the actual usage patterns and available IOPS or storage throughput. "We placed the VM on this datastore because it had free space ..."

VMware's focus on CPU and memory
Using VMware it is easy to precisely tailor the amount of CPU and memory resources allocated to each VM; storage is a different matter. While the required disk space is easy to provision, it is impractical to properly allocate each VM the correct IOPS (Input/Ouput Operations per Second) and storage throughput (in MB/s).

High impact of resource competition
The technical nature of disk IO results in high impact on performance and utilization. If storage latency peaks at hundreds of milliseconds or even seconds, performance becomes visibly sluggish and billions of CPU cycle can be wasted.

Concrete Recommendations

Analysis is just the start. Other software products just point out problems, vPanorama creates solutions by calculating recommendations using smart algorithms for comparing thousands of usage charts and bottleneck patterns.

Which resource needs to be addressed first?
Even when everything is virtual, real issues happen on physical resources. Typically the reason is that too many VMs hit the same hardware resource - either host-side or storage-side. vPanorama analyzes in real-time which resources suffer from resource competition and allows you to focus on those resources that are impacted by real bottlenecks. Other solutions only point out resources that experience high load. But high load is not a problem; quite the opposite: you want to utilize your resources, therefore high load is exactly what you want. What is undesirable are bottlenecks (i.e. situations where VMs do not receive the capacity they need) and uneven load distribution.

Which VM to move?
Once resource bottlenecks are found, it is crucial to identify the root cause. Which VMs are creating the bottlenecks? Typically this is not the VM that consumes the most resources. What matters is the usage pattern. vPanorama analyzes patterns and finds the VMs that compete. It calculates a score for each VM and ranks all VMs by score.

Where to move it?
Now we know which resource is bottlenecked and which VMs contribute most to the problem. Next comes the identification of target datastores or hosts. vPanorama compares the usage patterns of all possible targets (ideally those for which vMotion or Storage vMotion is possible) and ranks them a by predicted impact. Now solving the issue is just a click away.