Q&A: Busting Through Backup Bottleneck

IT service provider Concentus was called in to help a school client make sense of all its VM backups. MSP Today reached out to Fernando Sabio, CCNA, Principal, Concentus to find out more by posing 11 questions.

MSP Today: This was a three-way deal. Can you explain the roles of Veeam, Gridstore and then yourselves, Concentus?

Sabio: Concentus was/is a managed service provider (MSP) to an independent, non-sectarian day school for boys in grades 3-12.  Concentus collaborated with the school's technical team to deploy a turnkey virtual machine (VM) environment that would virtualize the cross-campus critical applications including its Microsoft (News - Alert) SQL database and file server systems.  Veeam was chosen by Concentus as the backup software for all the VMs.  Together, Concentus and the school's IT team found and brought-in Gridstore's Grid, software enabled storage platform.  Gridstore was selected because it could uniquely deliver a storage solution optimized specifically for the specific requirements of Veeam.

MSP Today: The Veeam solution was built from scratch for backup of virtual environments. What can it do that traditional storage tools adapted for virtual environments can't?

Sabio: Veeam allows for whole server recovery, formerly "bare metal restores," while also allowing for file level restoration from the same backup data, as well as being able to replicate key VMs to an alternate vSphere host for quick DR turn.

MSP Today: How did the combo of Veeam and Gridstore address the performance issues?

Sabio: While the ability to protect and restore virtual machines instantly is one of the innovations that make virtualization so valuable for businesses, it also presents new challenges - particularly in the area of performance.  In order to take advantage of some of the great features in Veeam – such as dedupe, with changed block tracking and instant restore – random I/O performance becomes critical. Because each Gridstore vController [virtual controller] can optimize specifically to the workload it services and operate directly in the host, a significant amount of random I/O can be handled locally. The result is random I/O performance that is typically 2X (News - Alert) or greater than traversing the network for every random I/O.  Moreover, Gridstore delivers massively parallel I/O and striping of data across many storage nodes to ensure the highest possible throughputs during critical restore operations.

MSP Today: What does this say about future scalability?

Sabio: With regard to Gridstore's Grid, Concentus and the school do not have to worry about forecasting capacity needs in the future for their VM backup environment.  Through its industry unique software enabled storage platform, the Grid allows them to start with exactly what they need and add more capacity as required without disruption to their operations.  Each storage additional node instantly adds not just capacity, but also I/O and processing power.  This differs from traditional storage that scales within the same physical system and forces you to squeeze larger and larger backups through the same sized pipe eventually resulting in missed backup windows.  The Grid allows you to start small and incrementally grow, in step, with the growth of your storage requirements- and because the Grid also adds I/O and processing power in addition to capacity with every storage node, the school can be sure their backup windows will be met both now and in the future.

MSP Today: How virtualized is the school environment?

Sabio: The vast majority of the school's mission critical application infrastructure is now virtualized. 

MSP Today: Does this demonstrate that virtualization today is mature enough to handle mission critical and I/O intensive apps?

Sabio: With Gridstore's Grid solution, the answer is "yes."  In the event you need to restore a VM, the Grid can deliver as many parallel I/O channels as you need to maximize your throughput when you need it the most. And while your Grid grows more powerful, it also grows more fault-tolerant because there is no single point of failure.

MSP Today: What is the backup architecture? How many tiers? Is it all on-premises?

Sabio: The school also off-sites encrypted copies of its data, in addition to the local Veeam backups.

MSP Today: Why was this architecture chosen?

Sabio: After evaluating its options, the school and Concentus discovered the Grid - Gridstore's storage solution that has been optimized for Veeam VM backups. Using Veeam Backup & Replication, the school would have the file-level granular restore capabilities it needed in a Windows-centric environment. When combined with the Grid, the school could benefit from stable backup and replication with an optimized storage target that was future-proof, and the value of a "Pay-as-you-Scale" flexible pricing model to achieve both cost and performance efficiency.

MSP Today: What lessons did you learn that could apply to other MSPs?

Sabio: Like all client sites, the weakness is local utilities – ensure enough UPS runtime to span the most common brownouts and leave enough time for graceful system shutdown.

MSP Today: What infrastructure technology did Concentus replace at the school? What was the general set up?

Sabio: This was the school's first virtualization effort (no previous virtualization technology was replaced).  Deployment was as follows:

1. Veeam's backup and proxy host solution was set-up on the virtual machine and connected to the host.
2. The Grid's vController was installed on Veeam's backup machine to deploy, configure and manage the storage nodes via the Grid's GridControl.
3. Veeam's backup solution now viewed the Grid's storage pool as a mounted volume as a target for VM backups.

MSP Today: Why was the new infrastructure on-premises as opposed to more heavily pushing cloud services?

Sabio: For the most part, the reason was because the price of high speed and high SLA multiple ISPs combined with the inability to bend the cost curve in pay-per-GB/Mailbox of cloud services was a very real hurdle (common for organizations that exceed 500 users).  Also, the latency of "on-LAN" vs. "hosted" is noticed by shops that move all their storage and apps fully to the cloud.