Taking SDS to the next level

Bloxham Mill provides serviced office solutions to over 200 UK based companies; offering a broad portfolio of services. They approached Ethos Technology to help expand their data centre offering to provide an infrastructure service to their clients.

Bloxham Mill had identified a number of customers who had requirements where a public cloud Infrastructure as a Service (IaaS) just wasn't a good fit; as such, they had no plans to move workloads to the likes of AWS.

They each had core legacy applications which would need to remain in operation for many years to come; all of which had a requirement for the infrastructure to provide high availability, and the storage layer zero data loss. Prospective customers included a healthcare software provider, a finance house and a research institution.

This defined a requirement for synchronous data replication between two discrete data centre locations which until recently would have meant going down the SAN route (at significant cost). Currently, the customers' services are hosted in various places; either within their own core data centres or by other managed service/colocation providers.

Cost was also key; Bloxham Mill needed to differentiate their service offering from that of the various incumbents. To persuade clients to move to the new service both the technical solution and the price point it was offered at would need to be compelling.

To meet their prospective customer's needs, the storage solution had to support virtual and bare-metal workloads; block, file and object protocols, directory integration, and offer secure multi-tenancy. To provide all of this functionality with a single solution is a lot to ask.

SDS FITS THE BILL
Software Defined Storage (SDS) has been a reality for a few years now. SDS initially gained momentum supporting critical internal applications at the likes of Ericsson, Google and Facebook. Legacy approaches to storage weren't fit for their requirements so they went back to the drawing board and came up with new solutions to address some fundamental compromises of the commercially available technology.

In recent years, several excellent SDS offerings have come to market including VMware VSAN, Nutanix and Isilon, some of which even offer synchronous replication between data centres; until recently this was pretty much the sole playground of SAN. Each of the above solutions is a good fit for specific aspects of storage functionality; however, to fulfil all the identified requirements would need multiple solutions from one or more vendors. In contrast, Bloxham Mill found that vNAS from Compuverde meets all of their needs in a unified storage fabric capable of spanning multiple data centres.

TAKING THE METRO LINE
One of Compuverde's key differentiators is its ability to be deployed as a metro cluster; a simple, multi-site configuration that offers complete redundancy to support mission-critical data and applications. Within a metro cluster, data is synchronously mirrored and distributed to two independent physical locations.

The proposed design utilised a resilient, low latency (sub 5ms) WAN interconnect between primary and secondary data centres. One single file system spans both DCs, presented to users and applications via Ethernet. The storage keeps itself internally synchronised, horizontally throughout the cluster and vertically down to the physical storage so that all nodes have the same consistent view of the unified file system for all reads, writes and updates.

To meet Bloxham's future demands it was essential that the proposed storage solution could be scaled out easily and incrementally. In a Compuverde metro cluster, scaling is done simply by installing new nodes at the primary and secondary location. As a cluster expands, performance, capacity and IOPS all scale linearly. Auto rebalancing functionality ensures that storage is balanced within a cluster by dynamically moving data to the new nodes.

The design leveraged erasure coding to provide additional protection against data loss and also significantly increase the logical storage capacity of the cluster. This ensured that each data centre could tolerate the loss of two disks or nodes without disruption to service. The self-healing process on each non-affected node automatically kicks in to replicate the missing data simultaneously.

A Proof of Concept was run to validate the solution for Bloxham Mill. Initial testing was performed on one cluster of eight nodes providing a total capacity of 20TB; the cluster was then split into two clusters of four nodes, emulating a metro cluster configuration.

The list of Compuverde's available features is extensive so it was important that testing focused on Bloxham's identified requirements; functionality demonstrated included:

• Flexible scale-out capability; nodes were added to expand a cluster
• Emulated a metro cluster configuration with inter-cluster replication
• iSCSI and/or NFS Primary storage for bare-metal and VM hosts using multiple hypervisors; Hyper-V, KVM & VMware
• NFS and SMB file services
• S3 compliant object storage
• VMware integrated hyper-converged storage including live migration
• All flash and hybrid platform architectures
• High availability and data redundancy

The Compuverde platform 'sailed through' the testing and the solution has now been implemented in Bloxham Mill's primary data centre with initial customers using it for test and development workloads. Over the coming months, the secondary data centre will be brought on line facilitating the implementation of a metro cluster and along with it full DR capability; once in place customers will migrate production workloads onto the environment and enjoy significantly improved performance, availability and simplicity at a price point well below their existing services. Interest and initial feedback has been wholly positive.

Using Compuverde, Ethos Technology have helped Bloxham Mill build a storage foundation able to fulfil their clients' needs now, whilst also providing the capability to evolve incrementally as future capacity and use case requirements dictate. They've avoided many of the fundamental compromises which plague traditional approaches to storage with a solution which is incredibly simple to deploy, manage and maintain.

"We were very fortunate in that an introduction to Compuverde coincided with a data centre refresh," commented Ray Avery, Managing Director of Bloxham Mill. "Our business is built around providing impeccable service to our customers; we always endeavour to present a calm and professional exterior to our clients but at times there's an incredible amount of work going on beneath the surface. Compuverde has given us a fantastic platform to build upon, offering a level of flexibility and resilience which would have been impossible to achieve with a legacy alternative."

"It was essential that the proposed storage solution could be scaled out easily and incrementally. In a Compuverde metro cluster, scaling is done simply by installing new nodes at the primary and secondary location. As a cluster expands performance, capacity and IOPS all scale linearly. Auto rebalancing functionality ensures that storage is balanced within a cluster by dynamically moving data to the new nodes."

Published in: Storage Magazine UK

Published in: Cloud IT

Published in: Data&Storage Asean

More news
Top