Akorri Expands Storage Management For Virtual Machines

BalancePoint manages across both VMware and Microsoft virtualized environments and a "Predictor" can foresee storage problems that could lead to application workload outages.

Image Gallery: 8 Online Storage Solutions
(click for larger image and for full photo gallery)
Akorri, a third party supplier of tools for virtualized environments, has added a storage "Predictor" to its BalancePoint tool in the 3.5 release of the capacity management software. Akorri does something that only a third party can do. It manages across both the VMware and Microsoft virtualized environments.

The 3.5 release includes analytics for virtual machines running under Microsoft's Hyper-V hypervisor as well as those under VMware's ESX Server. BalancePoint is both a "VMware-ready" management tool able to work with vSphere 4 and one that plugs into and extracts information from Virtual Machine Manager in Microsoft's System Center, according to Jim Comstock, VP of marketing.

The addition of a Predictor for Storage module to the tool means it can foresee storage problems that could lead to application workload outages. Predictor can spot demanding workloads that tend to tie up storage I/O, slowing other workloads. It issues an alert on the disk contention, allowing interventions that better balance the workloads, Comstock said in an interview.

BalancePoint also issues an alert on a disk that is being stressed by its workloads and warns against increasing demands on that disk.

The 3.5 version of BalancePoint supports storage systems from IBM, HP, Dell EqualLogic, and EMC and has extended that support to NextApp V-series, HP SVSP and Hitachi Data Systems Dynamic Provisioning.

BalancePoints is a capacity manager and infrastructure monitor operating below the level of network management or application performance management. It establishes performance indexes for servers and storage, whether in use as physical or virtual systems.

"It builds a model of the time it should take for a standard I/O operation -- where the milliseconds are consumed across the infrastructure. It can do this across hundreds or thousands of processes in the infrastructure," said Comstock. When a process is starting to deviate from its norm, BalancePoint can "see where the bottleneck is occurring" and point to a device or distinct step in the process as the probable cause, he added.

Bryan Snyder, a senior systems engineer, joined Millennium Pharmacy Systems, a supplier of pharmaceutical management software to long term care facilities, two years ago. It was his job to bring virtualization to its data center.

He's gone from zero virtual machines to 45 running on 16 physical hosts under ESX Server. In the process, he's moved away from running Microsoft SQL Server on a cluster to running it in a set of virtual machines. When he experienced and SQL Server outage on the former cluster, it took 32-40 minutes for his database system to restore itself and start running again, a bothersome wait for the database system that served Millennium's external customers. Now he can recover the system in three minutes, he said.

To help ensure that his database systems stay up and running, Snyder implemented BalancePoint six months ago. "With Akorri talking directly to storage, it knows what logical unit numbers (usually each disk has a unique identifier, a LUN) are serving which virtual machine. I can drill in anywhere along the chain," he said.

"I can also see how memory is being used by both the VMs and the host machine," he said. Memory is a chief constraint in virtual machine operation and seeing the amount used is a key performance indicator, he said. Since he implemented BalancePiont, he has better distributed his resources among virtual machines to make more efficient use of CPU cycles and memory on a host server. He sums up the increase as going from 50% use to 75% use, he said.

"In the old days, I knew of a problem when someone came to me and told me about it. It could take several days to clear up what the problem was," he recalled. Now it takes a few minutes to isolate the area of the problem, and with luck, an hour or less to resolve it.