GigaOm Radar for Data Center Switchingv2.0

1. Summary

A decade ago, data centers were much smaller and applications were typically huge monolithic affairs deployed using the classic client-server model. The core-distribution-access architecture model for networking served those applications well in data centers with north-south data flows because the application was hosted on a single server that communicated with clients outside of the data center.

In contrast, applications nowadays are built using microservices, which run on different servers within the data center. Microservices must communicate with each other to achieve the functionality of the wider application, requiring more east-west traffic than their monolithic counterparts. As such, the microservices of today’s applications require a change in data center architecture, namely to leaf-spine deployments with two switching layers.

Another change that is seeing increased adoption in the data center switching space is the use of disaggregated solutions. These types of solutions exchange integrated hardware and software provided by a single vendor for bare-metal hardware appliances and separate network operating systems (NOSs).

This approach allows data center operators more flexibility and control over the networking infrastructure. While we don’t expect to see all data center infrastructure shifting toward disaggregated solutions overnight, vendors that focus only on integrated solutions may well be left behind.

With new architecture models and disaggregated solutions, along with the increased demand for data and the consolidation of data center operations resulting in fewer and fewer players, the key characteristic to assess in any solution is scalability. In this context, scalability is highly dependent on how vendors support design, deployment, and operation activities.

Intelligence and automation around activities for Day 0, 1, 2, and beyond will differentiate the market between leading vendors and laggards. A highly scalable data center solution can support features such as one-touch deployments, artificial intelligence (AI)-based network troubleshooting and diagnostics, and batch policy configurations.

Finally, today’s application-first perspective positions networks as a support function, reframing our approach from a bottom-up view—network to application—to a top-down view—application to network. Combining development operations (DevOps) methodology with network operations, NetDevOps entails remote provisioning, configurations, and networking policies to support application performance.

This GigaOm Radar report highlights key data center switching vendors and equips IT decision-makers with the information needed to select the best fit for their business and use case requirements. In the corresponding GigaOm report “Key Criteria for Evaluating Data Center Switching Solutions,” we describe in more detail the key features and metrics that are used to evaluate vendors in this market.

2. Market Categories, Deployment Types, Professional Services, and Portfolio Physical Characteristics

To better understand the market and vendor positioning, we assess how well data center switching solutions are positioned to serve specific market segments and deployment types (Table 1) as well as the types of services offered by vendors and the physical characteristics for a data center switching portfolio (Table 2).

For this report, we recognize the following market segments:

• Cloud service provider (CSP): These can either be public or private cloud service providers that operate large data centers and offer infrastructure-, platform-, and software-as-a-service products.
• Communications service provider (telco): These are telecommunications service providers that offer connectivity and other IT services, often operating their own data centers for their internal use cases or to sell cloud-like services.
• Public sector: These are regional, national, and multinational government bodies that have greater requirements for security and privacy.
• Edge service provider/content delivery network (CDN): These organizations host hundreds of small data center-like points of presence, increasingly offering compute and storage capabilities.
• Small-to-medium business (SMB): Smaller organizations have different switching requirements, often needing smaller and cheaper appliances with less complexity to manage.
• Large enterprise: Larger organizations typically require support for large and business-critical projects in complex architectures.

We also recognize four deployment models for solutions in this report:

• Proprietary integrated appliance: These appliances are made up of a vendor’s proprietary hardware, running a proprietary NOS.
• Open integrated appliance: This criterion includes vendors who offer their proprietary hardware running a pre-packaged third-party NOS.
• Bare-metal hardware: These are just hardware appliances with no NOS installed, and customers are able to bring their own NOS.
• NOS: Some vendors can offer their NOS separately, and customers are able to supply their own bare-metal hardware.

Table 1. Vendor Positioning: Market Segment and Deployment Model

For this report, we recognize three types of services: professional services, managed services, and technical support. These services can be offered by the vendor, by a certified third party, or by independent third parties.

• Professional services – These services include procurement, design, and deployment of switching services, as well as installing appliances, cabling, and hardware removal.
• Managed services – Here, the vendor is responsible for firmware updates, incident response, security patching, disaster recovery, reporting, and other types of operational activities.
• Technical support – This includes services such as level 1-3 help desks, on-site resources, drop-in maintenance teams, onshore/offshore support staff, support hours (24/7, 9-5, follow-the-sun), and technical documentation.

We also recognize the following physical characteristics for a data center switching portfolio:

• Port speeds and densities: This assesses a vendor’s range of switching hardware with respect to the speeds supported, from 1G to 800G, and the number of ports supported per line card and appliance.
• Rack sizes and fixed and modular appliances: This criterion looks at the types of appliances offered by vendors, including rack sizes and fixed and and modular appliances.
• Power usage and cooling: We assess a vendor’s portfolio based on its power consumption and efficiency, which in turn affects the total cost of ownership.
• Leaf-spine: Here, we evaluate the vendor’s range of switching appliances that support leaf-spine architectures, which enables east-west data center traffic.
• Core-distribution-access: We assess the vendor’s range of switching appliances that support core-distribution-access architectures, which enable north-south data center traffic.
• Top-of-rack and end-of-row: Here, we evaluate a vendor’s range of switching appliances to support top-of-rack and end-of-row use cases.

Table 2. Vendor Positioning: Types of Services and Portfolio Physical Characteristics

3. Key Criteria Comparison

Building on the findings from the GigaOm report “Key Criteria for Evaluating Data Center Switching Solutions,” Table 3 summarizes how each vendor included in this research performs in the areas we consider differentiating and critical in this sector. Table 4 follows this summary with insight into each product’s evaluation metrics—the top-line characteristics that define the impact each will have on the organization.

The objective is to give the reader a snapshot of the technical capabilities of available solutions, define the perimeter of the market landscape, and gauge the potential impact on the business.

Table 3. Key Criteria Comparison

Table 4. Evaluation Metrics Comparison

By combining the information provided in the tables above, the reader can develop a clear understanding of the technical solutions available in the market.

4. GigaOm Radar

This report synthesizes the analysis of key criteria and their impact on evaluation metrics to inform the GigaOm Radar graphic in Figure 1. The resulting chart is a forward-looking perspective on all the vendors in this report based on their products’ technical capabilities and feature sets.

The GigaOm Radar plots vendor solutions across a series of concentric rings, with those set closer to the center judged to be of higher overall value. The chart characterizes each vendor on two axes—balancing Maturity versus Innovation and Feature Play versus Platform Play—while providing an arrow that projects each solution’s evolution over the coming 12 to 18 months.

Figure 1. GigaOm Radar for Data Center Switching

As you can see in the Radar chart in Figure 1, all vendors featured in the report are positioned in the Platform Play hemisphere. In the previous version of the report, we also featured NOS providers such as IP Infusion, which were placed in the Feature Play half as they did not offer any proprietary hardware but only integrated with bare-metal hardware providers. For this version, we excluded these providers, featuring them instead in a separate GigaOm Radar report.

All vendors featured in this report offer both hardware and software, with hardware portfolios composed of more than one product such that they are all positioned in the Platform Play half.

Within the Platform Play hemisphere, vendors are equally distributed between the Innovation and Maturity quadrants. Vendors in the Maturity/Platform Play quadrant have well-defined offerings for both hardware and software. Those featured in the Innovation/Platform Play quadrant also have some differentiating features, such as proprietary traffic optimization developments or advanced features for intent-based networking and automation.

5. Vendor Insights

Arista (DCS-7000 Series)

Founded in 2004, Arista is a key player in the data center switching space, delivering an open, standards-based, elastically scalable automated switching solution. Arista’s data center switching portfolio includes the 7000 series switches and the Extensible Operating System (EOS).

Arista ranks high on most metrics defined in this report, for both hardware and software capabilities. With respect to hardware, the DCS-7000 series switch portfolio includes both fixed and modular systems with port speeds ranging from 1 to 400GbE, supporting leaf-spine, top-of-rack, and end-of-row architectures.

Arista ranks high as well for Day 2+ operations, offering a variety of features, such as real-time state streaming for network telemetry and analytics; cognitive analytics using machine learning (ML) models to generate network recommendations and insights; turnkey automation for Day 2 configuration management and network-wide change control, such as automated upgrades, network rollback, and network snapshots; and NetDevOps workflows to integrate into a broader continuous integration (CI) pipeline.

CloudVision is Arista’s single point of control for the data center switching hardware and EOS. It enables automated provisioning, change management, and compliance, as well as network-wide telemetric data capture, analysis, and recording. Arista offers CloudVision as an on-premises virtual or physical appliance or as a cloud-hosted service.

The vendor also ranks high on interoperability and openness because its switching series offers a number of programming and automation interfaces. These include command-line interfaces (CLIs), eAPI, OpenConfig, and YANG models with gNMI, RESTCONF, and NETCONF support. Further, there is an EOS software development kit (SDK) for native-component development, which lets customers program hardware directly for traffic engineering purposes and monitoring.

In addition to the integrated hardware and software, Arista also offers EOS for use on a select range of bare-metal switches from Quanta Technologies (QCT), EdgeCore, and Celestica in 10/25/40/100GbE connectivity options. CloudEOS can be deployed both as a virtual instance and in a container environment.

Strengths: Arista ranks high on most metrics described in the report, providing highly performant switching hardware and software with comprehensive capabilities.

Challenges: Arista’s advanced automation and management capabilities are provided by the CloudVision platform, which must be purchased as a subscription per device or based on volume licensing. This may constitute a considerable increase in the overall price of the solution.

Cisco Systems (Nexus Series)

Cisco continues to be a household name in the networking space, and its data center networking solution displays comprehensive capabilities across all metrics defined in this report. The Cisco Nexus platform consists of fixed and modular switches, delivering automation, programmability, and real-time visibility. Switches from Cisco’s Nexus 3000, 7000, and 9000 series are deployed according to standard architectural guidelines ranging from 1 to 800GbE. As such, Cisco ranks high on the physical format and performance key criteria. The vendor’s comprehensive hardware and software product suite also ranks high on the end-to-end portfolio evaluation metric.

Nexus series switches can operate in either of two modes—Cisco Application Centric Infrastructure (ACI) or Cisco NX-OS. The ACI is a software-defined networking solution for data centers that provides centralized automation and policy-driven application profiles. The vendor ranks high on Day 1 automation, as Cisco ACI supports a declarative control system through which end users can state their preferred configuration while the platform creates adequate infrastructure policies.

NX-OS is Cisco’s NOS, which provides the capability to use foundational Layer 2 and 3 technologies as well as modern technologies such as VXLAN, with a Border Gateway Protocol‒Ethernet VPN (BGP-EVPN) control plane, segment routing, Multiprotocol Label Switching (MPLS), and automation.

To support holistic management of NX-OS deployments, Cisco offers the Nexus Dashboard Fabric Controller (NDFC), which reduces provisioning and deployment times through automation, and offers graphical operational visibility of topology, network fabric, and infrastructure to reduce troubleshooting time and effort. Further, it can eliminate configuration errors and automate ongoing changes in a closed loop with templated deployment models and configuration compliance alerting with automatic remediation. It also provides a real-time summary of the health and topology of fabrics and switches.

The vendor ranks high on the NetDevOps evaluation metric because the data center switching solution provides automation via an application-driven policy model, including centralized visibility with real-time application health monitoring, and support for automation tools such as Chef, Ansible, and Puppet. The NX-API supports a common programming language across Nexus switches. Python scripts, Bash shells, and Linux containers can also be used to develop customer applications.

Strengths: Cisco has over 30 years of experience designing and building networking equipment, making Nexus a market-leading set of data center switches. The vendor has an extensive portfolio and can support most use cases required today.

Challenges: Many of Cisco’s advanced capabilities are delivered through applications such as the NDFC, which must be purchased separately, increasing the solution’s total cost of ownership. Moreover, the products are complex and comprehensive, requiring long learning curves or support from third parties.

Dell Technologies (PowerSwitch Series)

Dell Technologies has long been prominent in the enterprise IT space, with the vendor catering to the data center switching market through the PowerSwitch S and Z series. Dell’s data center switching appliances can be deployed as integrated hardware and software solutions, standalone NOSs, or as bare-metal appliances. PowerSwitch Open Networking switches represent Dell’s latest disaggregated hardware and software solutions for data centers.

The PowerSwitch product suite ranks high on the physical format and performance metric, with the data center solutions offering 10/25/40/50/100/400GbE deployments in top-of-rack, middle-of-row, and end-of-row architectures. With high-density ports ranging from 25 to 400GbE and a broad array of Layer 2 and 3 features, the S and Z series meet the growing needs of today’s data centers.

With respect to software, the PowerSwitch series can be operated using Dell’s own networking OS10, which provides scalable Layer 2 and Layer 3 Ethernet switching capabilities with quality of service (QoS) and a full complement of standards-based features.

The choice of an NOS can simplify fabric orchestration and automation in data centers. Dell Technologies ranks high on interoperability and openness because the PowerSwitch series supports alternate operating systems, such as IP Infusion’s OCNos and VMware NSX. The vendor leverages the Open Network Install Environment (ONIE) in PowerSwitch for zero-touch installation of alternate NOSs. Dell also offers the Enterprise SONiC Distribution, a commercial version of the open source software with management enhancements, testing, and validation across select PowerSwitch models. A wide range of open source and Linux-based applications and tools enable customers to optimize and manage their network more effectively.

Dell ’s Fabric Design Center (FDC) is a cloud-based application that automates the planning, design, and deployment of network fabrics that power Dell compute, storage, and hyperconverged infrastructure solutions. This makes the vendor rank high on the Day 0 and Day 1 key criteria. The Fabric Design Center automates the planning, design, and deployment of network fabrics for Dell compute and storage infrastructure solutions, translates business intent to network designs and switch configurations, generates logical and physical network views for network planning and topology decisions, provides PowerSwitch details such as bills of materials, network diagrams, and cabling diagrams, and simplifies and automates new fabric deployments.

Strengths: Dell boasts a good portfolio of switching hardware as well as a highly open solution. This makes its switching solutions very flexible and able to support many use cases. The Fabric Design Center’s capabilities provide great Day 0 and Day 1 capabilities.

Challenges: Dell ranks lower on the NetDevOps metric, as it does not currently support integration with CI/CD or infrastructure as code (IaC) tools.

Extreme Networks (SLX Series)

Extreme Networks’ data center switching solution includes the SLX switching series, which is capable of supporting data center designs for leaf, spine, border routing, and data center interconnect (DCI). The SLX 9150 with 1/10/25/40/100GbE is ideal for leaf deployments, while the high-density 100GbE SLX9250 with 100/40GbE is most suitable for spine deployments. Further, the ultra-high-density 100GbE SLX9740 switch supports up to 80 x 100GbE ports. However, Extreme Networks does not currently support 400GbE ports, which are becoming readily available from multiple other vendors featured in this report.

The SLX switches can be managed using multiple methods, including REST, the NETCONF management interface, and a CLI for manual configuration. In terms of centralized management, the ExtremeCloud IQ Site Engine (XIQ-SE) offers a comprehensive unified management solution to manage users, devices, and applications.

Extreme Networks ranks high on Day 0 and 1 activities, enabling automated configuration and services provisioning, such as creating databases, provisioning the IP network fabric, and adding assets into the fabric.

Extreme’s SLX series runs SLX-OS, a Linux-based virtualized operating system with advanced switching features and support for the REST API with the YANG data model, Python, and NETCONF. Based on Ubuntu Linux, it comes with all the advantages of open source and provides access to commonly used Linux tools. Thus, the vendor ranks high on the NetOps evaluation metric.

The Extreme Fabric Automation solution streamlines and accelerates the deployment of an IP fabric. The application runs as a service on the Integrated Application Hosting environment of SLX switches and uses open API-based programmable interfaces to make provisioning, deployment, and automation of IP fabric networks easy.

The vendor ranks high on interoperability and openness, as the solution can work well with hypervisor providers, storage solutions, and security partners.​ Extreme Networks can support proprietary or third-party applications to be hosted on its own appliances to offer better support for security, monitoring, troubleshooting, or extended network functionality, without a separate hardware device.

Extreme Fabric Automation is able to provision data center fabrics based on BGP, EVPN, and VXLAN, the application validating the network topology before configuration to ensure accuracy. For Day 0 activities, the solution can configure a management IP address per switch manually or automatically via ZTP. For Day 1 activities, the solution can provision a fabric with the assets that need to be included (VLANs, VRFs, Port Channels, Ports) and End Point Groups (EPGs).

Strengths: In addition to the vendor’s comprehensive portfolio of switching hardware, its support for NetOps methodologies through interoperability, openness, and automation of Day 0 and 1 activities makes Extreme Networks a good choice for supporting modern data center requirements.

Challenges: While Extreme Networks ranks high on many of the criteria described in the report, the vendor can further improve its capabilities around automation for Day 2+ activities. Similarly, despite a good portfolio of switching hardware, Extreme does not currently support 400GbE port speeds.

HPE Aruba Networking (CX Series)

In 2015, Hewlett Packard Enterprise (HPE) acquired Aruba Networks, a leader in networking gear such as wireless access points and network access infrastructure solutions. Since the acquisition, HPE Aruba Networking has expanded into the data center networking space with the CX series of switches.

As part of its DC portfolio, HPE Aruba also offers the Fabric Composer, a software-defined orchestration solution for simplifying and accelerating day-to-day operations and for provisioning leaf-spine design across rack-scale compute and storage infrastructure. Using the AOS-CX API, HPE Aruba’s Fabric Composer orchestrates the CX series switches into a single entity—the fabric—which significantly simplifies operations and troubleshooting by automating various configuration and lifecycle event processes.

One of HPE Aruba’s differentiating features is its Distributed Services Switch, which provides stateful firewall services at every top-of-rack configuration. The firewall service can be managed through Aruba Fabric Composer to configure the firewall and set up segmentation and microsegmentation.

HPE Aruba’s portfolio of switching hardware includes a range of appliances. On one end is the CX 8000 series, a switching solution consisting of flexibly sized switches ranging from 12 to 48 ports of 10/25/100/400GbE suitable for leaf-and-spine data center networks. And on the other end is the CX 10000 series switch, co-developed with Pensando Systems, which is the industry’s first system that uses a hardware-accelerated programmable processor (Pensando P4) to deliver stateful services inline, at scale, with wire-rate performance.

As a single operating system for all HPE Aruba switches, AOS-CX delivers a consistent operator experience, simplifies network design, and enables management tasks across data centers and remote edge infrastructure. As well as the standard CLI, AOS-CX has an intuitive WebUI that simplifies and standardizes tasks while shortening the learning curve for network engineers who are new to HPE Aruba DC switches.

Strengths: HPE Aruba’s CX series caters to modern data center requirements through advanced tooling such as the Fabric Composer and the Distributed Services Switch. The vendor also ranks high on the switching and tunneling key criterion.

Challenges: In terms of deployment models, HPE Aruba does not currently offer bare-metal hardware or open integrated appliances. However, AOS-CX is available as a standalone NOS.

Huawei (CloudEngine Series)

Huawei has a comprehensive portfolio of data center switching solutions and holds the largest network market share in China, as well as maintaining a strong presence in Africa, Asia, Europe, and Latin America.

The CloudEngine series features switches that support top-of-rack and end-of-row architectures, with ports supporting speeds up to 400GbE. While the devices support modern leaf-spine architectures, the portfolio of products is still tailored for core-distribution-access configurations.

In combination with Huawei’s Versatile Routing Platform (VRP) operating system, the CloudEngine series switches enable high-performance Layer 2 and Layer 3 switching capabilities that support an elastic, virtualized, and high-performance network. As part of its Layer 3 capabilities, the CloudEngine series offers features such as BGP-EVPN, which can run as the VXLAN control plane to simplify VXLAN deployment.

Huawei also developed iLossless, a technology that prevents packet loss caused by network congestion. This algorithm prevents excessive data from entering the network, protecting the device buffer or link from being overloaded. It achieves this protection by controlling network flow and congestion. With flow control, the transmission rate of the sender is suppressed, enabling the receiver to receive all packets. On the other hand, congestion control is a global process involving all hosts and network devices for end-to-end management of any deterioration of network performance.

With the open programmability system (OPS) embedded within the VRP(OS) software platform, the CloudEngine solution provides programmability at the control plane. The APIs can be integrated with mainstream cloud platforms (including commercial and open cloud platforms). The OPS enables services to be flexibly customized and provides automatic management.

iMaster NCE-Fabric is Huawei’s network automation and intelligence platform, which implements unified control and dynamic scheduling of multiple local data centers and public cloud network resources, enabling rapid deployment of enterprise cloud services. As such, Huawei ranks high on Day 0, 1, and 2 activities.

Northbound, iMaster NCE-Fabric uses independent service provisioning user interfaces (UIs) to implement Layer 2 to Layer 7 interconnection with an OpenStack cloud platform and Kubernetes container orchestration platform. Southbound, it manages physical and virtual network devices—such as firewalls and physical and virtual switches—achieving heterogeneous network device management by loading drivers.

Northbound interfaces (NBIs) enable iMaster NCE-Fabric to intelligently convert user-centric service requirements into network configurations, perform simulation and verification, and deliver configurations in batches, making networks automated and intelligent.

Strengths: The solution has comprehensive switching and routing capabilities as well as extensive functionality for automating network design, deployment, and operation activities.

Challenges: Despite Huawei’s position as one of the world’s leading suppliers of data center equipment, possible ties to the Chinese government paint the vendor into a corner when it comes to entering the US market or other Western economies due to national security concerns. By association, Huawei’s ecosystem of US and EU-based partners is negatively affected as well.

Juniper Networks (EX and QFX Series)

Founded in 1996, Juniper Networks is one of the household names in the data center networking space, ranking high on most metrics described in the report. Data center switches from Juniper include multiple integrated products, such as the EX and QFX series, which are designed to meet the needs of data centers, small, medium, and large enterprises, cloud providers, and managed service providers. Its hardware portfolio supports data port speeds from 10 to 400GbE, and can support leaf-spine, core-distribution-access, top-of-rack, and end-of-row architectures.

Switching hardware is powered by the Junos NOS, which can power Juniper’s entire portfolio of switching, routing, and security products. The EX and QFX series switches deliver an extensive routing stack. And these switches can be deployed in a number of different network designs and fabrics. These options include IP fabric with EVPN-VXLAN and Juniper MC-LAG for Layer 2 and Layer 3 networks, giving customers complete architectural flexibility. Juniper solutions are generally led with the Apstra data center fabric management and automation software.

Junos is a modular NOS built on the principles of open programmability and intent-based automation. It provides clear separation among the control, management, and data planes. Junos OS supports a number of open source automation frameworks, including Puppet, Chef, Ansible, and Salt. Furthermore, Junos REST APIs allow users to connect securely to switches and execute remote procedures. By using the REST API Explorer GUI, users can experiment conveniently with any REST API and explore various formats and options, including JSON.

In addition to all the Junos automation features, the recent acquisition of Apstra provides a multiple-vendor management solution. This is an intent-based solution that allows network operators to automate entire data centers from end to end, integrating tasks such as group-based policies, as well as enterprise-scale analytics and operations.

Strengths: Juniper boasts a comprehensive and fully featured data center switching portfolio with its EX and QFX series and Security product lines. The vendor ranks high on a variety of metrics, including support and automation for network design, deployment, and operations.

Challenges: While Juniper has comprehensive capabilities for its data center switching solutions, these are achieved using products such as Apstra and Contrail, which must be purchased separately and will impact the solution’s initial cost to deploy.

Nokia (IXR Series)

Nokia has expanded in the data center networking market with the IXR series, offering a comprehensive portfolio of integrated hardware and software switching solutions that focuses on NetOps methodology.

In terms of hardware, Nokia offers the 7250 IXR product suite, which consists of high-performance, high-density modular platforms designed for data center spines and data center wide area network (WAN) connectivity deployments. Similarly, the 7220 IXR suite provides a high-performance, high-density, fixed configuration platform for leaf-spine deployments in data centers. These two series of hardware products feature multiple chassis variants that support port speeds from 1GbE up to 800GbE.

The IXR series runs the Service Router Linux (SR Linux) NOS. SR Linux is an open, extensible, and resilient NOS for data center fabrics and switching, built on an unmodified Linux kernel. The vendor ranks high on the NetOps suitability evaluation metric, having designed the Nokia Fabric Services System as an operational toolkit and management system to improve and scale operations across the entire operational life cycle of the fabric. Due to the highly automated design, deployment, and operations processes, Nokia also ranks high on support for Day 0, 1, and 2+ activities.

For Day 0 and 1 activities, the network engineer needs to provide a few parameters, such as the number of racks and the number of servers per rack, and the system will auto-generate the rest of the configuration based on Nokia-certified design templates. The result is a standard BGP-based IP fabric design with details such as IPv4/IPv6 addressing, BGP configuration, and cable maps. This can all be validated using the Digital Sandbox before being deployed to the data center fabric.

For Day 2+ operations, the Fabric Services System constantly monitors the fabric by leveraging telemetry, comparing it with various intents, and analyzing the results to find configuration inconsistencies, faults, or other deviations that may lead to network issues.

Nokia offers the NetOps Development Kit (NDK), which enables application developers to take advantage of the underlying model-driven architecture of SR Linux. With the NDK, data center teams can develop new applications and operational tools in the language of their choice and have deep programmatic access to, and control over, the entire IXR switching system.

Strengths: Nokia 7200 IXR series switches deliver high-performance hardware with a fully programmable NOS built for NetOps from the ground up. The vendor ranks high on a variety of metrics, including automation and support for network design, deployment, and operations.

Challenges: Compared to other solutions, SR Linux is a newer solution on the market, so we expect its users to require comprehensive training to understand and make use of all the features the product has to offer.

NVIDIA (​​Mellanox Spectrum Series)

Since its inception as a graphics accelerator company specializing in video game development in the 1990s, NVIDIA has evolved into a key player in processing acceleration with composable, disaggregated resources that serve cloud and edge applications. After acquiring Mellanox Technologies and Cumulus Networks, NVIDIA consolidated its entrance into the data center switching market.

The NVIDIA Spectrum Platform is an end-to-end data center networking solution, including the NVIDIA ConnectX NICs, Bluefield DPUs, and the SN4000, SN3000, and SN2000 series of switch systems, specifically designed for leaf/spine/superspine data center applications. The switches range from 16 to 128 ports, with speeds from 1 to 400GbE, allowing the construction of any size data center with any desired blocking ratio.

For software, the Spectrum series supports network disaggregation, allowing the use of a variety of NOSs, such as NVIDIA Cumulus Linux, NVIDIA Onyx, SONIC, and native Linux OS. Cumulus Linux, based on Debian Linux, is a principal choice for using the Spectrum range. It is a modern NOS designed to build, automate, and operate web-scale networks both affordably and efficiently.

An integrated Spectrum hardware switch with Cumulus Linux software can offer advanced network virtualization, including single-pass VXLAN routing and zero-touch network provisioning. Moreover, it allows users to simulate a multiple-switch configuration in the NVIDIA Air infrastructure simulation platform to verify the hardware purchased before it’s installed on-site, and to configure an entire network within minutes.

NetQ is an intelligent network validation tool from NVIDIA that provides real-time visibility and troubleshooting to simplify unit testing development and accelerate adoption. In addition, NetQ provides operational intelligence and insight into the health of data centers—from the container to the switch and port—supporting a NetOps approach.

NVIDIA Air is a cloud-hosted network simulation platform that replicates a real-world production environment. NVIDIA Air is used to create a digital twin of your IT infrastructure so that you can use it for running full-scale network architectures with multiple NOSs, and validating configurations, features, and automation code.

For Day 2+ activities, NVIDIA NetQ provides real-time data for troubleshooting, visibility, and automated workflows, and a method to compare prior network configurations to configurations after network changes are made to eliminate risk of disruption, preventive validation.

Strengths: NVIDIA’s strategic acquisitions of Mellanox and Cumulus Networks positions the vendor strongly in the data center switching market, being able to deliver end-to-end data center switching solutions with powerful capabilities around interoperability, openness, and support for network operations.

Challenges: NVIDIA offers extensive software capabilities, but its portfolio of hardware products does not currently include devices such as modular chassis, WAN routers, and power over Ethernet (PoE) switches.

6. Analyst’s Take

The data center switching space is mature and features a wide range of vendors with comprehensive capabilities for both hardware and software. Innovations with respect to hardware keep up with the demand for increased data consumption, as most vendors are already providing hardware that supports 400GbE ports and some are starting to offer 800GbE ports. Similarly, most players in the space now offer hardware suitable for supporting architectures used for facilitating east-west data center traffic. With the core switching capabilities at this level of maturity, vendors are now focusing more on automation, orchestration, and interoperability.

Most developments are taking place in the software space around NOSs and with management platforms. The increase in the complexity and number of data centers, as well as the distributed computing infrastructure for edge platforms, requires switching solutions to support smooth design, deployment, and operations.

Openness and interoperability are other trends that need to be addressed by switch vendors. While integrated hardware and software solutions have been the norm and offer a consistent experience and a single point of contact, the resulting solution may be more expensive and lock in businesses with a single vendor. In contrast, disaggregated solutions may be more flexible and cost-efficient but could initially require additional skills and longer learning curves for the network operations team.

We expect that an increasing number of deployments will feature disaggregated solutions, leading to a more extensive networking ecosystem. Market share will become more evenly distributed among vendors because enterprises will have more possibilities to mix and match, depending on their investment capabilities and technical requirements.

As it currently stands, all vendors featured in this report can provide reliable and performant data center switching solutions, regardless of whether they offer integrated or disaggregated solutions. The key aspect to consider when selecting a vendor is their capabilities for supporting the deployment and management of increasingly complex environments that are geographically distributed. Especially when looking at edge deployments, it is important to keep costs manageable. Costs are not associated only with the price of hardware or software licenses but also with the number of network operators required to maintain the network. We therefore expect trends in the data networking space to focus on consistent management platforms and NOSs, while hardware can be either integrated or bare-metal appliances.

7. About Andrew Green

Andrew Green is an enterprise IT writer and practitioner with an engineering and product management background at a tier 1 telco. He is the co-founder of Precism.co, where he produces technical content for enterprise IT and has worked with numerous reputable brands in the technology space. Andrew enjoys analyzing and synthesizing information to make sense of today’s technology landscape, and his research covers networking and security.

8. About GigaOm

GigaOm provides technical, operational, and business advice for IT’s strategic digital enterprise and business initiatives. Enterprise business leaders, CIOs, and technology organizations partner with GigaOm for practical, actionable, strategic, and visionary advice for modernizing and transforming their business. GigaOm’s advice empowers enterprises to successfully compete in an increasingly complicated business atmosphere that requires a solid understanding of constantly changing customer demands.

GigaOm works directly with enterprises both inside and outside of the IT organization to apply proven research and methodologies designed to avoid pitfalls and roadblocks while balancing risk and innovation. Research methodologies include but are not limited to adoption and benchmarking surveys, use cases, interviews, ROI/TCO, market landscapes, strategic trends, and technical benchmarks. Our analysts possess 20+ years of experience advising a spectrum of clients from early adopters to mainstream enterprises.

GigaOm’s perspective is that of the unbiased enterprise practitioner. Through this perspective, GigaOm connects with engaged and loyal subscribers on a deep and meaningful level.

9. Copyright

© Knowingly, Inc. 2023 "GigaOm Radar for Data Center Switching" is a trademark of Knowingly, Inc. For permission to reproduce this report, please contact sales@gigaom.com.