Intelligent CIO Africa Issue 08 | Page 50

INTELLIGENT BRANDS // Cloud always better than fault tolerance, as any failure puts the infrastructure at greater risk of outage. 4. What cloud management platform do you use? Customer self-service through web GUI or API is part of the definition of cloud and is a base-level requirement from any cloud provider. Once again, enterprise data requires enterprise hypervisors, virtualisation management and orchestration tools. How comfortable would an enterprise be with a cloud provider experimenting with a complicated, cutting edge open-source software stack? Clients must demand a mature, fully developed product set with established security credentials. This needs to be purpose-built for cloud, upgradeable and supportable by certified engineers, and backed by a trusted vendor. 5. Do you offer the full suite of cloud services? Although there is an argument that enterprises should use multiple cloud providers to mitigate risk, it is also important that when it comes to cloud services the provider is a one-stop shop. Basic requirements include a virtual data centre, virtual LAN, and routing/firewalling service. Integrated self-service backup, archiving, and replication/disaster recovery services to geo-redundant data centres should be the norm for any leading cloud provider. Andrew Cruise, Managing Director, Routed be restricted in your choice of private or public connectivity into or out of your cloud environment, or paying a premium for Internet breakout due to vendor lock-ins. For performance (latency) or compliance (e.g. POPI) reasons, it is better to be located in-country. 3. What hardware is utilised within their infrastructure? Performance and availability SLAs with credit-backed penalties for breach of contract provide comfort to enterprises and are the minimum requirement from any reputable cloud provider. Although there is an argument that this should be sufficient, due diligence is also required to reassure a client that the SLA can be achieved. Similarly, minimum N+1 redundancy is a given (N+2 or 2N is much more preferable), but is not on its own sufficient. Enterprise workloads require enterprise hardware: for performance, clients should expect guarantees (in particular to counter ‘noisy neighbour’ in multi-tenant environments); for availability, only established vendors with battle tested products should be considered – failure avoidance is 50 INTELLIGENTCIO Consider the provider’s willingness and flexibility to connect to other service providers (e.g. storage services, Internet and other private connectivity services) either physically (open access) and/or virtually (using advanced routing and networking technologies). What services are offered to enable hybrid solutions, between a client’s premises and cloud; and what options are available for bare metal dedicated compute or isolated deployments? What overarching security solutions are available? 6. What additional steps do you take to protect my data? It’s all about a client’s data. Apart from the base requirements for redundancy and fault tolerance, and the additional paid for backup and disaster recovery products, what else is being offered for data durability and resilience? Is data kept in single instance only, or snapshotted locally, or even replicated across failure domains or geo-redundantly? What additional security options are available: disk encryption? Two-factor authentication? How can a cloud provider evidence that they care about the data as much as you do? n “While there is massive scope for growth and improved cloud adoption, decision makers clearly need guidance in terms of what to look for when finally taking the leap into the cloud.” www.intelligentcio.com