We hope you enjoyed the Nov. 1 DemoFriday with Cyan, Accedian, and Colt. The presentation covered how network operators can leverage open, multi-layer SDN solutions to deliver services more efficiently across multiple technology domains, including an MPLS backbone or a third-party network. It also explored how Colt, a major European Ethernet service provider, implemented SDN solutions within their Ethernet Multi-Service Platform (MSP) network to lower costs and enable more automated end-to-end service delivery.
After the demo, presenters Abel Tong, Cyan director of solutions marketing; Ken Gold, Accedian senior product manager; and Mirko Voltolini, Colt director of design and architecture answered questions from participants. An expanded version of that Q&A follows. Watch the full presentation, or check out the teaser video and other resources below.
Why would I choose either Y.1564 or 2544 over the other?
Accedian: While both RFC 2544 and Y.1564 can be used for Service Activation Testing (SAT), they do have some fundamental differences. First off, RFC 2544 has been around longer than Y.1564 and is used more widely, especially in North America. Originally intended to test Layer 2 Ethernet devices in isolation, it is now used to stress the network to ensure it meets specific Layer 2 transport requirements. Y.1564, on the other hand, has been designed to test the service layer performance rather than the Layer 2 transport. Being able to set up and test up to eight concurrent service flows, each with a separate endpoint, better addresses the customer service environment; however, it is not yet as widely adopted as RFC 2544 for SAT. In the end, both protocols address the basic need of providing a birth certificate to back up a premium Carrier Ethernet service offering.
Colt: Colt uses Y.1564 for both internal acceptance test and customer-facing test portals to validate our services against the offered SLA. A significant benefit is that every single test frame is time-stamped and measured for Layer 1 and Layer 2 performance reporting/testing purposes.
Do these tests validate all aspects of the service, end-to-end?
Accedian: This depends entirely on the service being offered. Both will test the Layer 2 transport performance of the service. However, Y.1564 will more closely emulate the actual service environment, including VLAN tagging, CoS marking, etc., and may provide a more accurate view of performance of each individual service.
What is Colt’s level of involvement in the ETSI NFV project?
Colt: Colt is one of the founding members of the ETSI NFV initiative and has been an active participant to the collaborative work carried out by the now 150+ members. We contributed to the creation of the initial concept and to the preparation of the first specification recently released. We firmly believe the industry needs a radical change with the adoption of a standardized approach in hardware manufacturing. Today’s heavily customized paradigm is not a differentiating factor for service providers but a limiting factor, increasing time to market and holding back service innovation. Telecom manufacturers needs to focus their innovation effort into the software layer to provide service operators with flexible service platforms.
What is the biggest issue holding back the adoption of carrier SDN?
Colt: Two factors: lack of technology maturity and the need to preserve the investment in the current technology environment. SDN-based solutions (mostly based on “overlay” approaches) are available for adoption in a data-center environment but have not yet reached full maturity for widespread adoption in the WAN. In addition, it is not cost-efficient to fully replace the existing technology environment with SDN-based solutions.
Colt’s on the path toward giving customers “self-service” control over their WAN bandwidth. When would you expect that type of service to be broadly available in the industry?
Colt: There is a strong push from our customer base, and the whole industry is rapidly moving toward adoption. The main challenge is not technology but embracing a different business model, where the customer can control consumption of network services in a similar way they are able to do with cloud-based IT services. How rapidly this will happen depends on our willingness to disrupt the market; we are among some early adopters.
Where is OpenFlow in this demo?
Cyan: OpenFlow is a protocol that an SDN controller would use to program packet-forwarding instructions into an Ethernet switch. SDN is a methodology for deploying networks. SDN takes advantage of advances and scalability in computing to provide a simpler, more agile networking environment. The demo showed how Colt is leveraging SDN today to achieve end-to-end network control, service automation and service agility. The demo did not show the OpenFlow protocol.
In the absence of a unifying SDN standard, how do you ensure that your SDN solution is interoperable with other vendors’?
Cyan: Cyan is very much committed to standards, is actively monitoring the progress of many standards development organizations, and will adopt the relevant set of standards. In the meantime, Blue Planet’s southbound interfaces are mediated through element adapters (EAs) that allow the management and control of existing network elements through CLI, SNMP, OpenFlow, etc. In the absence of a standard, the EAs mediates between Blue Planet’s unified network model and a multi-vendor environment.
Colt: This is still to be decided.
What types of services does the system provision?
Cyan: Colt is currently using Blue Planet to orchestrate Ethernet services across a multi-domain, multi-vendor network. However, in the WAN, Blue Planet provisions wavelength, OTN, TDM and packet services.
What portion of the graphic is considered “The Controller?”
Cyan: I believe the question is referring to the Blue Planet architecture diagram shown in the webinar. The diagram has a large horizontal box labeled, “Blue Planet software platform.” This box is the controller. The controller provides “network services” including management and orchestration. And it is the controller that maintains the end-to-end network view. Blue Planet does so through element adapters, which communicate to network elements with CLI, TL1, SNMP, Openflow, etc.
Can you also make Y.1731 measurements?
Cyan: Yes. ITU-T Y.1731 is the standard mechanism being used to measure delay, delay variation, and frame-loss metrics for Ethernet services. These metrics are aggregated and displayed through Blue Planet’s Planet View application.
Accedian: Yes, ITU-T Y.1731 is also supported in the Accedian MetroNIDs and MetroNODEs, as well as RFC 5357 (TWAMP). However, these protocols address continuous performance monitoring of the service (Y.1731 at Layer 2 and TWAMP at Layer 3 rather than service activation testing.)
Can you do performance monitoring at both PW & LSP levels?
Cyan: Yes. Planet View is a general-purpose, real-time monitoring and SLA assurance tool. Planet View provides real-time monitoring for Ethernet and non-Ethernet services. While most customers use Planet View for Ethernet services, Planet View is also being used to monitor optical power, OTN errors, ATM, IP SLAs and more.
Can you perform path optimization with PCE?
Cyan: Yes. The default path computation is based on shortest-path. However, Colt also uses other criteria for path optimization.
Colt: Yes. We can create fully flexible routing templates manipulating all network segments (ring instances, ring directions, core path/tunnel used).
Can you give a quantitative or qualitative capex-saving ratio?
Cyan: The savings are significant. In fact, Colt’s Modular MSP network has several elements that lead to significant capex savings. First, Colt has converged IP and Ethernet service delivery into a single network. A converged network reduces equipment and infrastructure and also simplifies the network. Second, Colt has chosen all-Ethernet transport. Ethernet switching is significantly less expensive than the equivalent IP routed solution. Third, when delivering IP services, Colt deploys a simple Layer 2 CPE and virtualizes the Layer 3 CPE functionality (vCPE). The vCPE allows Colt to deploy the same cost-efficient CPE for all services, eliminate “custom” configuration at the CPE, and ultimately turn up services faster. Finally, Colt’s multi-vendor strategy allows vendors to compete for business in every part of the network. Vendor competition ensures Colt has fair market pricing for their equipment.
Colt: Colt has set a target capex reduction of 50 percent and is on the way to achieve it through the various network transformation initiatives.
Same for the “faster innovation” cycle. Any quantitative numbers or analysis?
Cyan: Speed of innovation can vary greatly depending on the specific type of innovation. For example, consider the two applications — (1) bandwidth on demand for end-users and (2) cloud bursting data center services. These applications are very different in scope and complexity. In either case, Blue Planet is a key enabler for faster innovation. Blue Planet provides a single “pane of glass,” open northbound APIs, and end-to-end network control of a multi-domain, multi-layer, multi-vendor network.
Colt: Today it takes at least 18 months to introduce a new network connectivity solution based on a monolithic, single-vendor architecture (with CPE, edge, core, and network management system provided by the same manufacturer). We are aiming to reduce the introduction of a new service component to less than six months across all our network geographies following the modular multi-vendor approach.
How does the legacy OSS interact with the controller and app intelligence?
Cyan: Legacy OSSs interact through Blue Planet’s northbound APIs.
It was mentioned earlier that the wizard automates service creation, but that certain details can be tweaked later. How does Colt make sure that future automated changes, such as a speed change, don’t override or affect any custom tweaks and/or cause a service outage?
Cyan: The wizards shown in the demo are used for service creation. The wizard essentially transforms user-definable templates into real services. Wizards are not currently used to modify services after creation.
Is OpenView able to carry all of the management information up to the Blue Planet controller?
Cyan: The Blue Planet SDN platform has element adapters that communicate either directly to network elements or with EMS systems. Blue Planet can collect network information from HP OpenView and would likely do so via an SNMP interface. I am not aware of HP OpenView being used in Colt’s network.
Is there a way to schedule and automate the RFC2544 test?
Cyan: Test scheduling and automation is area that Cyan is actively working on. Look for a Blue Planet solution, including an application as well as northbound APIs, coming in the near future.
Colt: Colt has developed a portal to allow the customer to launch a test. This allows the customer to repeat the same test Colt performed at service activation time using the same test values and allows them to compare results. We also offer the possibility to set up a Layer 2 MAC or Layer 3 IP source/destination loop for up to 24 hours at any network endpoints. Blue Planet will enrich our test portal with more functionalities for the next release.
Can I use the Blue Planet solution without the Cyan optical-packet transport platform to build a multi-vendor solution?
Cyan: Yes. While we would certainly advocate the deployment of Blue Planet customers together with Cyan packet optical equipment, Blue Planet can be used without Cyan packet optical equipment. Approximately 20 percent of Blue Planet customers today are operating networks that have no Cyan hardware.
Where’s the routing or re-routing decision made? At the control plane or data plane?
Cyan: I believe this question is a question about protection switching. Failure detection and the actual protection switch occur in hardware, on a network element. Hardware-based switching is required to guarantee sub-50ms switching. However, in order to guarantee overall network performance and each individual service’s performance for normal and failure conditions, working and protect paths both need to be engineered. Traffic engineering for working and protect paths includes consideration of resource management, path latency, and other factors related to optimal network performance.
Can the RFC2544 tests handle jumbo frames above 9,000 bytes?
Cyan: Yes. Accedian RFC2544 testing supports frame sizes up to 10,000 bytes. Cyan also supports frames up to 10,000 bytes.
Accedian: Yes. The RFC does not specifically exclude specific packet sizes, and all Accedian MetroNIDs and MetroNODEs support a maximum frame size of 10,240 bytes.
What kinds of customers are using the CPEs for line speeds above 1 Gb/s?
Colt: Large enterprise customers or wholesale providers can have connectivity requirements of 10 Gb/s for hub sites aggregating multiple remote spoke sites. In addition, 10 Gb/s CPEs are used for NNIs between carriers.
Does the MetroNID shown in the demo have any kind of SBC functionality, or is it just a dumb NID?
Cyan: The Accedian MetroNID in not “just a dumb NID.” Accedian provides traffic policing and shaping, hierarchical QoS, Ethernet service OAM, etc. We do have a product that is “just” an Ethernet NID and does not provide session border control (SBC) functionality. Note that while many networks deploy SBC at the network edge, SBC functionality can be virtualized and run as a virtualized network function (VNF).
Accedian: Very well stated. Thanks, Abel!
Will the same network be able to support and self-provision OTN/wavelength services (today and mid-term)?
Cyan: During the webinar, Mirko mentioned that this is something that Colt will be looking into in the future. Blue Planet supports A-to-Z provisioning for wavelength, OTN and Ethernet services. As Colt introduces these services, Blue Planet will be able to support them.
Colt: Colt is focused on providing end-to-end automation across all our service portfolio. OTN/wavelength services are also a priority and are being addressed.
Does Colt see a need for automated routing at the metro head-end to direct a Layer 2 service onto the transport core and bypass the MSER (as a further capex saving)?
Cyan: Today, Colt is already realizing benefits from bypassing routing. Ethernet services switch locally. In other words, Ethernet traffic from the metro only reaches the provider edge when necessary. However, there are other opportunities to bypass routing. Looking at traffic through the core (in a number of carrier networks, not just Colt). Much of that traffic is transit — i.e., not routed. Bypassing routers and transiting the traffic through a “hollow core” can offload the routers and boost efficiency. Offloading the routers is significant because it gives the routers headroom. This means that carriers can increase the capacities of their networks while maximally leveraging the investment in their already installed routing infrastructure.
Colt: For international carrier Ethernet services, we are building a full mesh of Ethernet PW tunnels through the core, to create a direct connection between Cyan Layer 2 metro areas (Cyan PBB tunnels end-to-end using a transparent Ethernet PW tunnel through the MPLS core). Each such single tunnel can carry hundreds of services between city A and B and is managed by Blue Planet end-to-end.
What portions of the control plane (as opposed to the management plane) are moved to Blue Planet for each of the vendor islands?
Cyan: Blue Planet provides network management and orchestration. Blue Planet maintains the network-wide view and makes the network accessible through a unifying information model and northbound APIs.
What are the SDN API protocols used in this network?
Cyan: Blue Planet northbound APIs are RESTful APIs over JSON. At Colt, alarm management is integrated via a northbound SNMP interface.
How does the SDN solution work for initial provisioning when a new network element, such as an EMUX/EMT, is deployed in the network?
Cyan: Blue Planet needs IP connectivity to manage new equipment. For example, in the demo, Blue Planet was able to cut an Accedian NID into a ring and provision services over the NID. However, IP connectivity was established first (outside the demo). IP provisioning can be automated, but for a number of reasons, we (Colt, Accedian, and Cyan) intentionally have not automated IP provisioning yet.
Colt: The service activation process is partly automated today. Blue Planet provides a ring access module which allows us to insert a new customer CPE into an existing access ring with other live customer circuits. A static IP address and the network port ERP enabled are required to be manually pre-configured on the Accedian CPE. Blue Planet then completes the CPE insertion automatically (Blue Planet performs a controlled traffic switch away from the insertion section, creates a new CPE object and inserts it on the ring, checks connectivity and auto-installs the correct firmware, configures a default network role template, and, once completed, the traffic switches back to normal state). At this point, the CPE insertion is completed, and Blue Planet is ready to create services (with manual or automatic steps).
In Q1 2014, we plan to introduce zero-touch provisioning (no hands-on configuration needed). Once a new CPE is inserted on a ring, a multicast beacon provides instructions on what management VLAN to lock in. Once the management interface is created, DHCP starts and Blue Planet will auto-detect the new unit, then the rest of the service activation already in place today is completed.
For the NIDs, are they on a separate VLAN? And are they intrusive or not, i.e. service affecting?
Cyan: First, VLAN S-TAGs are used to distinguish services traversing the same physical ring. Second, cutting a NID into a ring is “intrusive.” Since the cut-in is changing ring “topology,” the cut-in forces a sub-50ms protection switch around the cut-in. This may or may not affect other services depending on where the services are with respect to the cut-in.
Accedian: The MetroNIDs and MetroNODEs are managed using a management VLAN which is configurable by the carrier and can be either single- or double-tagged. In this way, the traffic is not “intrusive” on any specific user traffic. However, it is a VLAN within the overall link capacity and can be prioritized such that during congestion, lower-priority user VLANs may be impacted.
Can the NIDs assume a TX role independently?
Cyan: I am not sure I understand the question. NIDs demarcate services at the customer premises. NIDs bridge customer traffic onto standard ITU-T G.8032v2 Ethernet rings. As per the Modular MSP concept, Colt has ensured that functionality performed by NIDs are based on well known standards.
Accedian: I am not sure I understand this question, but the Accedian NIDs do not independently do anything that may impact the existing services. NID behavior is well specified in G.8032v2.
Does this architecture require both the service provider and infrastructure provider (fiber) to subscribe to SDN and the portal?
Cyan: SDN is flexible. Either or both providers would benefit by having Blue Planet. Neither is required to have it. Consider a mobile backhaul example. A wireless carrier may lease backhaul services from a local service provider delivering connectivity to a cell site. On the one hand, the backhaul provider could use Blue Planet to manage, orchestrate and virtualize their network. Blue Planet would aggregate performance metrics on the backhaul services. The wireless carrier could use Planet View to monitor the performance of all of their individual services. On the other hand, the wireless carrier may independently use Blue Planet to maintain network-wide visibility and monitor services end-to-end. Note, in this case, the backhaul provider would normally inhibit the wireless carrier from having direct access to the backhaul network infrastructure.
Can you comment on service-provider requirements to host this type of service orchestration platform (i.e. private & public access)?
Cyan: Blue Planet can be hosted either in a server-based or VM environment. The Blue Planet server is normally hosted privately, since it interacts directly with network elements. However, individual applications like Planet View are hosted in the cloud. In the case of Planet View, the Blue Planet server pushes performance data into the cloud. Planet View performs analytics and makes the data accessible to anyone, anywhere.
Colt: Colt hosts the Blue Planet platform internally. Self-service activation, service test, and service views are provided to the customers through a web portal.
What is driving the capex reduction in the Colt implementation?
Cyan: As mentioned in an earlier response, Colt’s capex reductions are driven by a combination of network convergence, an all-Ethernet metro architecture, network function virtualization, and competition between vendors. While capex reductions are certainly significant, opex savings are also important. In his presentation, Mirko talked about improving service turn-up times from 45 days to seconds. This service velocity equates to shorter time to realizing revenue. Also, having a “single pane of glass” for end-to-end service management simplifies Colt’s network management and troubleshooting. The abstracted MEF CE2.0 provisioning model gives Colt the ability to change vendors somewhere down the road without needing to retrain network operations personnel. And having a unified network model with open northbound APIs gives Colt DevOps the ability to optimize and customize their service delivery.
Colt: This is to be decided.
Find the SDN Products that are right for you on our DemoFridays! Watch videos and download podcasts and PDFs about leading SDN technologies: