in a manner more usually associated with very high cost SMP architectures. RAIS (“Redundant Array of Inexpensive Servers”) achieves this by turning a cluster of independent servers into a single large server running applications across a virtualised network of nodes. This virtualisation of the network offers full redundancy, reliability and flexibility in a way that, prior to the introduction of RAIS technology, had been unattainable. This ‘virtual shared architecture’ provides extreme scalability, allowing the seamless scaling of capacity without the attendant re-coding difficulties associated with changes to established cluster based installations. Most importantly the RAIS system operates without any changes to source and computer codes, making RAIS quick to implement and easy to manage.
the philosophy of data and disk management which today represents an almost universal approach to server disk operations. RAIS does the same for cluster based network management and will have the same far reaching consequences across the world of enterprise computing. RAIS models for servers the same capability as RAID does for disks, offering variations of scale and redundancy depending on the configuration chosen. Therefore, just as RAID stripes and mirrors data across multiple independent disks, RAIS stripes and mirrors memory and code of an application program across multiple independent nodes of a cluster. The applications see only a single logical shared memory which functions as a binary compatible Symmetric Multiprocessor (SMP) system.
Waratek achieved this breakthrough with the invention of an entirely new memory architecture that executes a single target application in multiple parallel memory states simultaneously. The Waratek RAIS software re-writes a target application’s instructions and code in real-time to execute the target application across the virtualised network. This is done completely independently of (and is invisible to) the application. It is this autonomous operation of the RAIS software that permits applications written for SMP systems to execute on a RAIS array without alternations or recompilations of source or object code.
From a software stack perspective, the RAIS software sits as a layer between the application and the underlying servers and is installed on each server in the cluster. In a similar way as the RAID Controller, the RAIS Controller manages a RAIS array in configurable levels of striping, mirroring, and parity configurations, functionally identical to RAID levels 0, 1, 5, and 1+0.
The development of the RAIS technology represents an important advance in systems design and architecture; for the first time in enterprise computing history RAIS delivers the efficiency and binary-compatibility of SMP based architectures at a fraction of the costs associated with such systems. Furthermore, it does so with without the limitations, inefficiencies and difficulties inherent in managing cluster based architectures.
management of enterprise computing systems. By removing the limitations and the operational problems associated with cluster based systems, the introduction of RAIS will have far reaching effects across the whole of the IT industry. Literally, RAIS will enable the full utilisation of low cost x86 based server clusters offering scalability, resilience and ease of operation to a level long sought after, but never envisaged as being truly achievable by conventional technology approaches. As RAID proved to be the Holy Grail of data and disk management so RAIS will have the same significance to the world of network management.
The demands of high through-put enterprise computing are achieved by one of two marketed hardware solutions: clusters or SMPs. While SMPs offer a ‘shared architecture’ solution, clusters, by their nature utilise an ‘unshared architecture’ approach which carries with it several very distinct and ultimately limiting complications. While clusters are relatively cheap to purchase they are inefficient and complex to manage and use, these difficulties translating into high levels of direct management cost and downtime. Conversely SMPs, whilst clearly avoiding the problems inherent in clustered networks do so only at a price - an extremely high price - which places SMP technology outside the reach of many organisations and applications.
Now for the first time, RAIS introduces a third alternative computing architecture that will revolutionise network computing. With the RAIS alternative the undesirable elements of both clusters and SMPs are eliminated while the strengths of each are amplified.
when faced with scaling up computing power. Whilst highly desirable, this solution must always provide a framework of continuous operational reliability and dependability. RAIS achieves this along with many other benefits and attributes which are summarised below:
- Reduced capital cost - A RAIS virtualised network operates from multiple low cost x86 server platforms, a far cheaper solution (by several orders of magnitude) than moving to SMP technology. Moreover, based on equivalent memory and processors characteristics, RAIS systems will out perform SMP systems.
- Reduced complexity and operating costs - Because RAIS creates a single logical environment the complexity and error proneness inherent in writing and managing an application on clustered servers is eliminated.
- Increased IT flexibility and agility - Network configurations can be easily adjusted to suit peaks of operation and the increasing demands of particular applications (e.g. the rapid growth of a newly introduced subscriber based service).
- Scalability - Scaling existing applications on clusters is time consuming, expensive and difficult. RAIS is fully scalable and because it operates without any changes to existing code it is fast and inexpensive to implement.
- Resilience - RAIS enables continuous operation with universal redundancy - if a server fails the application doesn’t stop; RAIS automatically distributes the existing workload among the remaining servers.
- Uninterrupted change out of servers - An important consequence of redundancy inherent to RAIS is that RAIS not only directs workloads away from a failed server, but it enables the hot swapping of hardware without interruption to the operation of an application.
- Heterogeneity – A key feature of RAIS is the ability to cluster different configurations of servers, either by memory type or even by processor type, removing the headache (and cost) when upgrading the system or, for example, when changing out component servers for maintenance/repair.
- Compatibility – The automated nature of RAIS makes it a simple proposition to adapt additional applications to run on the clustered server.
- Low impact on existing staff skill requirements – RAIS removes the need for additional or specific high level skills. RAIS is easily assimilated into existing daily operations so developer teams are left to concentrate on their specific applications, having no need to worry about dealing with operating code and scalability issues. Indeed they do not even need to know that RAIS is present.
- Reduced power and cooling consumption – Multiple x86 servers have much reduced power demands compared to SMP hardware. At the same time cooling demands are much lower creating more flexibility in the installation environment.