Redefining networking for IP storage
By Sean Ong August 6, 2015
- One way of using IP storage is to leverage the existing data network
- But shared networks cannot handle the growth of business-critical workloads
AS business critical workloads doubles nearly every two years – driven by virtualisation and big data analytics – storage requirements have skyrocketed and are now measured in petabytes instead of terabytes.
Although organisations are still relying on Fibre Channel for their mission-critical applications, they are also turning to IP (Internet Protocol) storage like iSCSI (Internet Small Computer System Interface) and NAS (network-attached storage) for their business applications, due to the fact that they can be deployed more quickly and easily.
A very common way of using IP storage is to leverage the existing data network. But shared networks simply cannot handle the growth of business-critical workloads on IP storage as virtualisation, storage capacity, and the adoption of new storage technologies like flash continue to grow.
Application, storage and virtualisation companies are starting to recommend, and enterprises have experienced for themselves, that the deployment of a separate network dedicated to IP storage in the data centre is the best option to achieve greater reliability and more deterministic infrastructure.
To successfully leverage IP storage for business-critical applications, the following need to be considered:
- Achieve service level agreements for business-critical applications running on ultra-low latency solid state drives;
- Gain disaster recovery solutions that deliver local replication performance over long distances; and
- Benefit from unified management of IP and Fibre Channel storage networking.
These storage networks need to be based on flexible, open network architectures that deliver a simplified infrastructure and help fulfil the promise of IDC’s Third Platform compute model, with low, deterministic latency, guaranteed delivery, a smaller administrative domain that is easier to troubleshoot, fewer configuration compromises, better fault isolation, and less complexity to upgrade and maintain.
Dedicated networks for IP storage
To achieve the required SLA (service-level agreement) for business-critical applications and to take better advantage of the read-and-write characteristics of flash storage, co-mingling IP storage traffic with other data traffic like email, video, web and voice traffic in the same network will defeat those objectives and SLAs.
The impact is that all the other data traffic will make the network highly unpredictable and difficult to troubleshoot should there be IP storage issues.
With a dedicated network for IP storage, business-critical service level agreements can be addressed – particularly with resilient, high-performance Ethernet fabric, in Brocade’s view, to deliver predictable performance and high resiliency.
Ethernet Fabrics allow for easy provisioning and operations of the IP storage network due to built-in automation capabilities, resulting in minimal configuration and intervention as multiple switches in a Fabric can be configured and managed as a single logical element.
Within a Fabric, multi-pathing between switches provide storage-class resiliency with non-disruptive failover after a path or link failure.
IP storage extension for disaster recovery
A key concern with IP storage is disaster recovery. The ability to provide local replication performance over long distances between geographically-distributed data centres – with strong encryption – will alleviate this concern.
With SAN (storage area network) extension switches, not only can enterprises meet the disaster recover requirement, but up to 50 times more data can be moved to meet recovery objectives.
In addition, WAN (wide area network) links can be maximised through protocol optimisation technology, and downtime can be minimised to overcome failure of WAN links, latency, packet loss and security challenges.
Compared with using a native replication application, it is possible to now deliver 320 megabytes per second of replication data over long distances, with 25 milliseconds latency and 0.1% packet loss.
Unified storage network management
To gain unprecedented network visibility and insight across the storage network, unified network management allows monitoring of IP and Fibre Channel health and performance from a single dashboard, simplifying network monitoring and providing alerts and integrating network data into VMware vRealize to maximise VM (virtual machine) performance and availability.
These capabilities allow pre-validation and troubleshooting of the physical infrastructure to streamline deployment.
With the appropriate technology, IP storage can visualise network health and performance, increase instrumentation and granularity for real-time visibility and actionable insights, and simplify monitoring with the deployment of pre-defined policies, rules and actions.
By automatically detecting and recovering from errors, providing early warning of potential problems and minimising downtime with faster troubleshooting, common network problems are reduced and availability is increased.
In conclusion, for larger environments, there are compelling advantages to having a separate physical network for IP storage traffic, including low, deterministic latency, guaranteed delivery, a smaller administrative domain (which is easier to troubleshoot), fewer configuration compromises, better fault isolation, and less complexity to upgrade and maintain.
Simplifying the data centre network design by using a dedicated IP network for storage pays dividends. This is why leading application vendors recommend this as a design best practice.
Sean Ong is the country manager of Brocade Malaysia.
The KSS principle: Keep storage simple(r)
Flash is fast, but the real business benefit is yet to come
Brocade out to shed SAN player image
For more technology news and the latest updates, follow us on Twitter, LinkedIn or Like us on Facebook.