We generalize the notion of customer-dedicated, application-hosting servers in web server farms to a logical, distributed counterpart. The virtual servers so defined break traditional, static boundaries separating resources dedicated to individual (organizational) customers (such as physical servers or static partitions thereof) and open up new degrees of freedom for optimized and shared use of server farm resources. This approach offers an additional benefit to the utility-computing theme that substantial savings in peak-handling redundant capacity are possible if the independent customers share the same base of resources (– the probability of facing simultaneous peaks goes down “multiplicatively”). The benefit derives from the observation that the meaning of customer load is only determined dynamically, and that redundancy has to be built at the level of each static resource unit dedicated to a customer (a unit may individually get overloaded, even while other units are not). By co-allocating fractions of independent, virtual servers on the same physical server and keeping the boundaries between the co-allocations dynamic, the redundancy requirement on the physical server is shared and brought down for a given peak risk. The server farm we describe offers a further benefit of supporting customer-specific Service Level Agreements (SLAs) in terms of simple, hit-rate ranges of end-user generated load. Customers need not be bothered about the deployment and configuration details inside the server farm, but simply need to specify the functionality sought by them and a (load-based) price. Resource deployment, management, and operation are then the responsibilities of the server farm, for which the customer is billed for actual usage, as per an agreed upon SLA. We describe the architecture of the web server farm and discuss its management and operational details.
By: Vikas Agarwal, Girish Chafle, Neeran Karnik, Arun Kumar, Ashish Kundu, Johara Shahabuddin, Pradeep Varma
Published in: RI01006 in 2001
LIMITED DISTRIBUTION NOTICE:
This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.
Questions about this service can be mailed to reports@us.ibm.com .