Changes between Version 4 and Version 5 of CRON

04/21/10 03:01:14 (12 years ago)
Seung-Jong Park



  • CRON

    v4 v5  
    4040The cyberinfrastructure, CRON, provides integrated and automated access to a wide range of high speed networking configurations. Figure 1 shows how CRON can be reconfigured to emulate such optical networks as NLR (National Lambda Rail), Internet2, LONI (Louisiana Optical Network Initiative) configuration, or purely user-defined networks having different networking characteristics such as bandwidth, latency, and data loss rates. Moreover, users can dynamically reconfigure whole computing resources, such as operating system, middleware, applications, based on their specific demands. Due to the automated and reconfigurable characteristics, all types of experiments over CRON will be repeatable and controllable. This reconfigurable feature of CRON coincides with that of GENI – Programmability
     44'''2.   Rationale of CRON[[BR]]'''
     46Although many statewide or nationwide or global high speed optical networks have been deployed, only a few groups of scientists and researchers can access the networking resources due to the following reasons:
     48Uncertainty: There are discrepancies between the expected and the actual performance experienced while developing new application softwares over high speed networks. And the reasons of the discrepancies can be the computing environment, networking environment, or the application software itself. Networks, especially in the early phase of deployment, can show varying levels of quality, such as bandwidth/throughput, latency, jitter and loss rate. The uncertain cause of these problems prevents the application developers from identifying their origin; they cannot tell if the problem source is the network or the software. It is critical for application developers to be able to identify the origin of problems so that they have a chance to solve them. If the problem is in the network, it is usually hard to find its origin as the application developers need to check all potential problems of all the resources utilized for development. This dramatically increases overall development time. Application developers should be able to concentrate on debugging only their software or end-systems when an unexpected level of performance is observed. To make this possible, they need to be able to decouple the networking environment from the computing environment. CRON is a decoupled networking environment that is guaranteed to provide a stable level of network quality because it is localized and fully controlled.
     50Scarce and Expensive Networking Resources: Even when optical networks achieve a level of stability that guarantees an error-controlled environment, they are still very high-priced entities that support only a few concurrent users at each optical link. As the demand for dedicated high speed networks increases, the number of users (including both researchers and engineers) will exceed the capacity of the networks. Therefore, it is critical to serve the increasing number of users with cost-effective networking services that provide a development environment that is equivalent to the real networking services. The proposed CRON will be able to accommodate these increasing user demands with smaller investments than real multi-million-dollar optical networks cost.
     52Research vs. Production Networks: Although users are ready to use a dedicated networking environment requested by them, the users at the application layer are not allowed to modify built-in configurations of other layers, such as transport protocols and other software configurations of end workstations which are connected to the dedicated networks. This constraint will limit the scope of experiments of users. Although they have a chance to modify the configurations, they can interfere with other users at different layers, who share the end workstations. To prevent this interference over production networks, CRON will provide an independent and separate virtual networking and computing environment to each user so that each user can reconfigure any software parameters fit for user’s own experimental purpose.
     54'''3.   Architecture of CRON'''
     56[[Image(CRON:Connection Diagram.png,30%)]]
     57[[Image(CRON:Connection Diagram 1.png,30%)]]
     59Above figure shows the architecture of the CRON, which consists of two main components: (i) hardware (H/W) components, including a 10Gbps switch, a 1Gbps switch, optical fibers, network emulators, and workstations; and (ii) software (S/W) components, creating an automatic configuration server that will integrate all the H/W components to create virtual networking and computing environments based on the users' requirements. All components are connected at two different networking planes: a control plane connected with 100/1000 Mbps Ethernet links and a data plane connected with 10 Gbps optical links. To allow access from outside networks, such as Internet2, NLR, and LONI, and to connect to external compute resources, the switch at the data plane has four external 10 Gbps optical connections that will extend the capacity of CRON and integrate it with existing networks and projects for the purpose of cooperative scenarios shown in the Figure.
     61Due to the high cost of the network hardware emulators, we will connect them to only four optical links that will be allocated to satisfy demands over 5 Gbps of bandwidths. The remaining links will be connected to machines running software emulators. More precisely, we utilize the NIST Net emulator. This emulator can only support up to 5 Gbps of bandwidth due to computing overhead on the CPUs of the machines they run on
     63All components of CRON, except the remote access clients, are located at LSU, which already has established 10 Gpbs optical paths to LONI and Internet2. Although all computing and networking resources are located at LSU, CRON provides virtual networking and computing environments to users from different physical locations.
     66== Major Accomplishments ==
     681.      Sep 2009        -Finished hardware purchase including servers, racks, 10Gbps network cards, 1Gbps control switch and 10Gbps data switch and connection cables.
     702.      Oct 2009        -Finished the OPS server, including user file service, mailing service. Set up the mail relay with CCT through CCT’s mail exchange server.
     723.      Nov 2009        --Finished the testbed Control Network, including Cisco 3560E switch configuration.
     744.      Dec 2009        -Finished the BOSS server, including web service, accounts management, DNS service, and PXE DHCP configuration.
     765.      Dec 2009        -Finished CRON testbed cluster movement from testing machine cabinet to the machine room of the first floor of the Frey building
     786.      Jan 2010        -Finished Memory File System (MFS) including both FreeBSD based and Linux based MFS. Database configuration and webpage configuration are made successfully. Updated the FreeBSD based MFS from version 6.2 to 6.4 and then to the latest version of FBSD7.2 MFS. Successfully customized Fedora 8 image to be compatible with CRON project hardware. However, due to the SUN Fire x4240 hardware advanced feature of SAS disk buss and RAID controllers, the Emulab FreeBSD 6.x version generic image can’t be used.
     807.      Jan 2010        -Finished network access for BOSS and OPS server, both within LSU and outside Internet, including LSU wireless access.
     828.      Feb 2010        -Finished Serial line control network, testing on the serial line cable because the Sun Fire x4240 machine has special pinout serial port. Asked the Comtrol company’s official technical support and got the right cable pinout design. Currently, the serial control network could work, but has some wrong display problem. Asked CCT IT support to help on making such serial cable on test.
     849.      Mar 2010        -Finished CRON system version upgrading from Utah Emulab stable source branch.
     86'''Outreach activities'''[[BR]]
     88The CRON project already envisages close cooperation between LSU and Southern University via a shared testbed, thus uniting two institutions from an EPSCoR state which struggles to overcome historic disadvantages related to education and poverty levels. Due in part to the large African American population of its state, LSU has a history of minority recruitment (one of the co-PI's Ph.D. students is in fact African-American). We expect to hire several more minority graduate students to work on networking research problems using the CRON infrastructure. Furthermore, the two PIs regularly teach the graduate High-Speed Networking Course (CSC 7502) as well as the primary undergraduate course in the LSU computer science department on computer networks (CSC 4501). We plan to incorporate a laboratory component in these courses relating to CRON - by allowing students in these courses hands-on access via experiments for utilizing the virtual networking and computing environment of CRON, we will enable a better understanding of the benefits of virtualization and network reconfigurability. We plan to crosslist this course with Southern University thereby allowing students there to also participate in and understand virtualization research via class experiments. Thus through this collaboration, the two institutions will become better equipped to train future generation of engineers for the network computing industry, an industry that is expected to experience rapid growth over the next several years. The LSU computer science department also hosts an annual summer workshop for incoming undergraduates called Computer Science Intensive Orientation for Students (CIOS).  We plan to use the workshop to introduce freshman computer science students to the challenging field of networking research through CRON presentations and on-site demos.
     90'''Other Contributions''' [[BR]]
     92Our goal is to actively involve minority institutions and underrepresented students in this project. As a preliminary step towards this broader outreach, we have already established an experimental testbed connection between LSU and Southern University through LONI to perform networking experiments as shown in Figure 4. Southern University is one of the nation’s largest Historically Black Colleges & Universities. Thus the broader impact of CRON at Louisiana State will be to enable underrepresented minority institutions, which suffer the last mile problem and cannot access high speed networking resources, to perform experiments and researches. We also actively plan to publicize and advertise research opportunities in CRON at our annual departmental workshop for incoming freshman undergraduates and interested high-school students. This annual workshop in which the PIs participated in 2008 is called Computer Science Intensive Orientation for Students (CIOS) and covers basic introductions to key aspects of computer science as a discipline and a profession.  In 2010, we plan to use the workshop to introduce freshman computer science students to the challenging field of networking research through CRON presentations and demos. We also hope to actively increase undergraduate and high-school student participation in the project through summer internships as part of the LSU Chancellors undergraduate research program which involves mentoring honors undergraduate students.