Managed Service Provider Enables All-flash Storage Platform | Tintri

0 0

All-flash platform enables large MSP to differentiate its cloud offerings

Envision Technology Advisors

Company Background

Envision Technology Advisors is a managed service provider that delivers best-in-class technology solutions. Its company-wide dedication to customer service has now made it one of the most comprehensive, business-oriented technology companies in New England. Envision’s specialties include: cloud and managed services, desktop and data center virtualization, wireless infrastructure design and implementation, networking support, and website design and development.

Monolithic Arrays, Subpar Performance

Envision Technology Advisors had been running its internal and client workloads on NetApp storage primarily, as well as a specific customer workload on a dedicated Xiotech SAN. “We were getting frustrated with the management of our bulky, previous generation storage arrays,” reported Jeff Wilhelm, Envision Technology Advisors’ Chief Technology Officer. “The challenges fell into two primary categories. One was managing all of the LUNs, fiber channel switches, and all the things that come along with the monolithic storage arrays of the past. And the second piece was dealing with sub-par performance. Where do I put this disk? How big are my LUNs? What type of RAID am I using? How am I handling the SATA vs SAS vs flash tiers? Those were the challenges we had with both the NetApp iSCSI and the Xiotech fibre channel attached arrays.”

Looking for a Better Storage Solution

“We wanted to find a new storage technology that was NFS-based, one that simplified the stack, had good reporting capabilities, and provided a single pane of management— one without all of the complexity of 3-tier solutions,” noted Wilhelm. “We were getting tired of all of the complexities that come with traditional storage arrays.” After reviewing the storage options, Envision purchased its first Tintri T540 hybrid array.

Faster Workloads

Envision is now running all of its client and internal workloads on Tintri, including a broad mix of database applications and marketing data solutions. “We are running Red Point, SQL Server, and Pentaho’s business intelligence tool on Tintri,” noted Wilhelm.

“As soon as we moved everything over to Tintri, we saw significant improvements in performance. Jobs that were taking 5.5 minutes dropped to 1.5 minutes on the VMstore T540. Later, when we upgraded to the Tintri T650 and the T880 systems, we saw even greater improvements due to the amount of flash in those larger systems.”

Comparing Clouds

Envision Technology Advisors decided to compare the performance of its cloud offering on Tintri to several other cloud providers in 2014. “We compared the performance on the Tintri hybrid arrays to Terremark, Amazon Web Services, Rackspace, and two or three other cloud providers,” Wilhelm shared. “We used synthetic benchmarks, so we could measure performance equally across all platforms. Thanks to Tintri, our storage I/O performance was higher than every other co-location and service provider’s platform that we tested.”

Moving Up to All-Flash

Envision Technology Advisors purchased its first Tintri T5060 all-flash array in 2015. “We didn't test any other all-flash arrays,” Wilhelm admitted. “We’ve been so happy with the Tintri hybrid systems that we didn't want to change vendors. Tintri readily provides systems for evaluation and testing with the assumption that you're going to be so blown away that you're going to want to buy it at the end. And that was the case with us. We decided to see if the all-flash system performed as well as advertised, and they sure did. Jobs that used to take several minutes ran in 48 seconds on the all-flash arrays. Clearly, keeping all data in flash was really excelling with our large, data analytics-based workloads. When you’re reading a large file from a traditional file server, it’s frequently sequential I/O. That’s something that even monolithic storage arrays can excel at. But when you're talking about business intelligence, you’re pulling data from many different places on the database—there's lots of random I/O. The Tintri T5060 array really blew the performance out of the water due to its all-flash architecture.”

The Tintri all-flash array is now servicing many of Envision’s internal workloads as well as some customer workloads.

“Because of the way our environment is built, we can have full segregation between customer workloads with Tintri,” Wilhelm explained. “We operate a software-defined data center, and enforce segregation at multiple layers. With Tintri, our customer workloads are running side-by-side with our internal production workloads on the same box with full segregation between them.”

Hybrid vs All-Flash Cost Comparison

“Cost is always a concern when you start talking about storage,” noted Wilhelm. “An all-flash array of course is going to be a bit more expensive than a hybrid flash array, or certainly more than a traditional non-hybrid flash array that’s just spinning disks. But when you talk about cost per IOPS or performance, the all-flash array was actually a better deal than the other arrays we looked at. You have to go into all-flash knowing what you're getting into. You are getting less disk space because you’re getting much higher performance from the disk. So if you’re going into it with a performance need—not a disk capacity need—then the cost per IOPS of the all-flash array can be highly beneficial for your workloads.”

A Great End-User Experience

“Our customers like to throw challenges at us, and we always try to over-deliver on their expectations,” noted Wilhelm. “One of our hosted customers is a very large data analytics provider. They have a file build that takes over two hours, since it churns through billions of records. It is an absolutely insane process. We didn't tell them we were moving their VMs to the all-flash array, we just told them that we were doing some maintenance over the weekend. So we asked them to run a traditional file build on Friday against their server instance on Pentaho, and then run the same job again on Monday. They sent me an email that said, ‘I don't know what you did over the weekend, but the process ran 300% faster. The time to complete went from two hours down to 28 minutes.’ That time savings translates directly into significant business value for this customer.”

Leveraging the Same Management Pane

“The beauty of the Tintri’s all-flash arrays is that they use the same management pane as all of the other Tintri systems,” Wilhelm said. “The Tintri Global Center interface makes system management easy, it enables us to make intelligent decisions based on VM placement. We were able to see that the VMs on our T650 would really benefit from being moved to the T5060. If we had disparate arrays—a NetApp, a Xiotech, and a bunch of other things—there would be multiple interfaces to learn, there would be no single pane of management, and there'd be no single reporting capability. We had already made an investment in the Tintri ecosystem, and this just took it a giant step forward and gave us better performance without having to climb a new learning curve.”

Relying on Tintri Auto Support

“We are an engineering-driven organization,” explained Wilhelm. “They are always a bunch of people doing things in our datacenter, so there will always be times when an engineer will accidentally bump a power supply or purposely pull something out to see what happens. Just last week we were provisioning a Tintri in our lab and had a single PDU outage. It turns out we had plugged in one of the really big servers that we just finished building for a large customer VDI project, and it took out one of our power distribution units (PDUs) that also touched the Tintri. Of course nobody noticed, because everything is connected into multiple PDUs. So the system kept chugging along until we got an email from Tintri support that said, ‘Hey guys, your power supply is unplugged again.’ Tintri’s Auto Support feature is great—they know when something is wrong in our staging environment even before we do.”

Gaining the Ability to Differentiate Service Offerings

The Tintri all-flash arrays are enabling Envision to distinguish its offerings from other managed service providers. “Companies in our industry are always looking for the least expensive way to deliver a solution,” noted Wilhelm. “A lot of our competitors buy inexpensive disks and hosts to get by with the best possible margins. But quite often, these margins come at the expense of customer satisfaction. We decided to take a different route, and leverage our investment in Tintri to provide higher quality services that extend far beyond what you can get with a typical managed service provider. We will continue to invest in Tintri because of its best-in-class performance, management, and reporting. Even though it does cost a little bit more to provide managed services on hybrid and all-flash platforms, it allows us to have a differentiated offering that we believe our customers are more satisfied with, and with Tintri Global Center, a single pane of management and visibility for us.”



Managed Service Provider, Solutions Provider, Consulting


Providence, Rhode Island and Boston, Massachusetts

Virtualization environment

VMware vSphere

Traditional storage: NetApp and Xiotech

VM profile

Customer cloud services and internal applications

Key challenges

Improve storage performance for key customer use cases

Tintri solution

Tintri VMstore™ T540, T650, T880 and T5060 all-flash arrays

Primary use case

Tintri is being used for all customer cloud services, data analytics workloads, internal server environment, and virtual desktop workloads

Business benefits

Decreased runtime for common reports by 70%

Obtained all-flash performance for IO-intensive data analytic workloads

Gained single pane of glass management across all systems

Enabled the creation of differentiated service offerings