Alliance Sets a High Bar for I/O Vendors
September 9th, 2011 By: Industry Perspectives
Jake Smith is a member of Intel’s Data Center Group focused on virtualization and cloud computing technologies. Currently working to drive standards and interoperability of cloud computing, Jake has experience in introducing new technologies in the server, storage and mobile markets.JAKE SMITH
At Intel, we spend a lot of time thinking about input/output issues. We want to be sure we deliver I/O products that eliminate bottlenecks, reduce bandwidth contention among virtual machines, improve quality of service, and cut power consumption.
Given this focus, it’s always good to work with people who share the same goals. That’s the case with the Open Data Center Alliance, a consortium of global businesses that recently released a new usage model for I/O controls. This publication stems from the Alliance’s mission to develop usage models to drive interoperability, flexibility, and industry standards for cloud and next-generation data centers.
Shifting World View
The Alliance’s usage model for I/O controls is positively challenging Intel’s view of the world. The model focuses on making network and storage resources fully manageable. It advocates the use of guarantees, limits, and QoS tools in I/O management.
Here’s a relevant excerpt from the model: “The intent is to provide bandwidth partitioning to bandwidth allocation by task, VM or time of day, using priority scheduling and bandwidth throttling. The key here is development of performance targets by type of I/O and being able to guarantee these performance targets.”
To drive toward these goals, the ODCA spells out usage requirements for monitoring, service-level agreement (SLA) metrics, APIs, timeslice monitoring and control, and I/O reservations. As an Ethernet provider, it’s our job to consider these requirements in our product and technology roadmaps. There are things we must do in the silicon layer to enable the desired capabilities in the software management layer.
For example, the usage model spells out the requirement to “aggregate workload I/O consumption by hour, day, week, etc.” To make that possible, we have to build the certain capabilities and instrumentation into our silicon. This work at the hardware level enables the aggregation of workload I/O consumption.
Rising to the Challenge
For Intel and our Data Center Group, the release of the Alliance’s usage model is an opportunity—and a challenge—to ensure that our current 10Gb and 1Gb Ethernet technology, and our future 40GbE technology, have all the right stuff. That includes not only protocol flexibility but also controls and instrumentation that system administrators and data center architects can leverage to avoid bottlenecks, decrease power consumption, and increase performance.
In our role as a technical advisor to the Open Data Center Alliance, we listen carefully to the concerns and requests from the Alliance’s members. In our role as a company that builds silicon and solutions for I/O, our job is to deliver next-generation products and technologies that take the Alliance’s usage models into account.
The Alliance has set a high bar for I/O vendors. We at Intel think that’s a good thing—for both technology users and technology providers. Tell us what you think….we always like a good challenge.
To learn more about this new usage model for I/O controls, download the PDF from the Open Data Center Alliance.
Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.