Application Layer Processing
ALP represents a dramatic new approach to application acceleration - accelerating application flows across multiple application tiers. Traditional application front ends (AFEs) accelerate content delivery from the web tier back to the customer location. ALP is the first solution that addresses application latency across all logical data centre application tiers. ALP therefore represents the only solution that addresses end-to-end data centre performance bottlenecks, regardless of location. Coupled with Maestro's award winning short life transaction (SLTTM) technology, (which already addresses all front-tier issues) ALP delivers levels of acceleration unmatched by traditional AFE vendors. ALP monitors and controls the flow of all requests into the application. Using definitions created using the Crescendo Rule Engine (CRE), ALP is capable of recognising which tier each application request is destined for. ALP also understands that different requests impose different processing loads on the application. In addition, ALP recognises that each tier in the application has an upper processing limit in terms of simultaneous requests. When a tier reaches maximum capacity, ALP's Admission Control mechanism queues requests within Maestro and only sends them to the application when processing capacity is available. The order in which requests are sent to the application is controlled through ALP's Request Scheduling algorithm, which sends requests to the application in a way which guarantees optimal performance. This scheduling prevents a phenomenon commonly known as head-of-line blocking, where some requests block others from being processed in a timely manner. This blocking can occur either between tiers or within a tier, between different software components (e.g. OS, application framework, application code, etc.) – both of which Request Scheduling circumvents. Finally, ALP's Application Insight mechanism provides monitoring and reporting for all requests. Monitoring is performed between Maestro and the application (either the entire application or only a specific process within the application), between Maestro and the client, and end to end. Performance thresholds can also be configured to monitor and report on application performance compliance.Core Components
ALP's sophisticated algorithms are based on unique components that are new additions to the Maestro Platform. These new mechanisms provide the engine that enables ALP to deliver its acceleration, optimisation, and monitoring capabilities:- Crescendo Rule Engine (CRE) – The CRE is an advanced hardware-based classification engine that can identify application flows across all protocol tiers. The CRE can classify requests at multi-gigabit rates and includes an advanced Regular Expression engine responsible for parsing HTTP headers and content.
- Dynamic Request Queuing – ALP's Admission Control mechanism uses new queuing mechanisms that build on the advanced buffering techniques of Crescendo's SLT technology, which is responsible for TCP acceleration, offload, and multiplexing. The new request queues are dynamically adjusted to accommodate all application tiers and request loads, also enabling ALP's Request Scheduling algorithm.
- Adaptive Learning – ALP's new adaptive learning mechanisms are responsible for identifying the application processing loads of all requests and recognising per-tier capacity thresholds. Adaptively learned metrics are applied to both Admission Control and Request Scheduling mechanisms.
Expected Results
ALP's multi-tier acceleration capabilities should yield an improvement of 5x-10x in application performance, depending on the application. This is in addition to the performance improvements garnered from SLT's offload and acceleration capabilities which can be up to 10x....Crescendo Multi-tier Application Acceleration
Other Products in this category:
|
Top
