Monday, January 29, 2024

Juniper unveils AI-Native Networking Platform

 Juniper Networks unveiled its AI-Native Networking Platform, unifying all of the company's campus, branch and data center networking solutions with a common AI engine and Marvis Virtual Network Assistant (VNA). The idea is to leverage end-to-end AI for IT Operations (AIOps) for deep insight, automated troubleshooting and seamless end-to-end networking assurance. 

Juniper says its AI-Native Networking Platform provides the simplest and most assured Day 0/1/2+ operations, resulting in up to 85 percent loweroperational expenditures than traditional solutions, demonstrates the elimination of up to 90 percent of network trouble tickets, 85 percent of IT onsite visits and up to 50 percent reduction in network incident resolution times.

“AI is the biggest technology inflection point since the internet itself, and its ongoing impact on networking cannot be understated. At Juniper, we have seen first-hand how our game changing AIOps has saved thousands of global enterprises significant time and money while delighting the end user with a superior experience. Our AI-Native Networking Platform represents a bold new direction for Juniper, and for our industry. By extending AIOps from the end user all the way to the application, and across every network domain in between, we are taking a big step toward making network outages, trouble tickets and application downtime things of the past,” states Rami Rahim, Chief Executive Officer, Juniper Networks.

The new AI-Native Networking Platform includes two new enhancements to Marvis, the only AI-Native VNA in the industry driven by Mist AI, with proactive recommendations and self-driving operations plus a conversation interface (using GenAI for some use cases). With the following new capabilities, Juniper customers and partners get even more automation and insight:

  • Marvis Minis: the only AI-Native Networking Digital Experience Twin, uses Mist AI to proactively simulate user connections to instantly validate network configurations and find/detect problems without users being present. Minis simulates end user/client/device/app traffic to learn the network configuration via unsupervised machine learning and to proactively highlight network issues. Data from Minis is continuously fed back into the Mist AI engine, providing an additional source of insight for the best AIOps responses. No manual configuration is required, as Minis is always on and can be deployed automatically during periods of low network usage (e.g. at midnight on weekends) or via trigger events (e.g. after a network configuration change). Also, unlike conventional digital twinning solutions and synthetic testing, Marvis Minis integrates directly with the network, eliminating manual monitoring and analysis via external sensors, clients and applications.
  • Marvis Virtual Network Assistant (VNA) for the Data Center: Juniper is introducing the first and only AI-Native VNA for the data center, delivering the best insight throughout the entire data center lifecycle across any vendor’s hardware. For example, issues with data center cabling, configuration and connectivity from any vendor’s hardware are surfaced in the Marvis Actions UI (from Juniper Apstra) with suggested proactive actions. Additionally, the Marvis conversational interface (CI) allows IT teams to pose direct queries and get simple, seamless insight into the data center product documentation and knowledgebase using GenAI.

Juniper intros QFX switches based on Broadcom's Tomahawk 5

Juniper Networks introduced new data center switches based on the most advanced Broadcom Tomahawk 5 silicon for 800GE:

  • QFX5230-64CD switch is a high-radix class switch, dedicated for high- bandwidth network switching devices supporting up to 64 × 400GbE, 128 × 200GbE, 256 × 100GbE, 64 × 40GbE, 256 × 25GbE, 256 × 10GbE ports in 2 U form factor. The switch is designed for AI data center deployments and spine and super-spine roles within IP and Ethernet VPN– Virtual Extensible LAN (EVPN-VXLAN) fabrics. The additional Remote Direct Memory Access over Converged Ethernet (RoCEv2) capabilities of the QFX5230-64CD support IP storage deployments where instead of relying on deep buffer switching, the QoS mechanisms such as priority-based flow control-distributed services code point (PFC- DSCP) and explicit congestion notification (ECN) deliver high performance for the storage workloads. Support for ZR/ZR-M optics makes it suitable for edge and data center interconnect (DCI) use cases. 
  • QFX5240 800GbE switch (64 port 800GbE) is a next-generation, fixed-configuration platform designed for spine, leaf, and border switch roles. The switch provides flexible, cost-effective, high-density 800GbE, 400GbE, 100GbE, and 50GbE interfaces for intra-IP fabric connectivity as well as higher density 200/400GbE NIC connectivity for AI/ML use cases. It’s 51.2Tbps unidirectional throughput meets the bandwidth requirement of AI/ML workloads and storage systems with latency in the range of 700-750ns. Remote Direct Memory Access (RDMA) is the de-facto data transfer technology used in AI/ML workloads, and it uses Remote Direct Memory Access over Converged Ethernet v2 (ROCEv2) for transport at the network layer.

Both platforms offer Juniper Apstra intent-based networking, which delivers full Day 0 through Day 2+ capabilities for IP/EVPN fabrics with closed-loop assurance in the data center. Apstra is a fabric management solution that empowers organizations to automate and manage their networks across virtually any data center design, vendor, and topology, making private data center as easy as cloud. Apstra provides full Day 2+ operations assurance with multiple built-in intent-based analytics probes to assure your network is running as designed.

Juniper debuts 800GbE PTX Packet Transport Router

Juniper Networks introduced its PTX10002-36QDD router, a compact, high-density platform for use as core, peering, data center interconnect and data center edge, metro aggregation, and AI data center networking, for service and cloud providers, as well as enterprise networks.

The PTX10002-36QDD is a 2 U fixed-configuration 800GbE router with 28.8Tbps capacity for supporting 36x800GbE or 72x400GbE interfaces with inline MACsec. The platform is powered by Juniper’s custom Express 5 ASIC, which boasts 8M counters, 10M-plus FIB, 100,000-plus SR tunnels, 256 AES MACsec encryption supported on all ports, and wire-rate packet performance.

The PTX10002-36QDD offers a high radix routing platform with deep buffers and a cell-based switch fabric for spine or leaf in AI data center networking environments. AI data center networking capabilities include efficient, deep-buffered interfaces, scalable cell-based fabric design, VOQ scheduling, RoCEv2, Adaptive Load Balancing, 8M Counters, built-in IPFIX and INT-MD.

As a data center interconnect router, the PTX10002-36QDD features advanced EVPN, VXLAN, and data center connectivity including: EVPN/VXLAN, ECN/PFC, and in-band telemetry

The PTX10002-36QDD is also well suited for L2/L3 metro aggregation roles including: CFM/LFM – Y.1731; MC-LAG; hierarchical QoS; BIER; SRv6.

Arrcus brings Egress Cost Control to Multicloud

Arrcus introduced an Egress Cost Control (ECC) capability for its FlexMCN multi-cloud networking solution.

Arrcus' Egress Cost Control helps enterprises slash cloud connectivity costs by intelligently routing traffic to destinations with the lowest egress costs. 

This is especially useful for bandwidth-hungry applications such as distributed Large Language Models (LLM) for generative AI, Database Replication, High Performance Computing (HPC), as well as hybrid work. In these instances, customers routinely incur significant costs in their cloud bills because of hidden data egress charges. This is distinctly different from charges levied by a cloud provider when a customer exits the cloud permanently. 

"With Arrcus' ECC-enabled FlexMCN, organizations can now gain granular control over cloud egress costs to achieve significant savings," said Shekar Ayyar, Chairman and CEO at Arrcus. "We are committed to innovating and addressing the challenges faced by businesses in multi-cloud environments as is evident from this smart solution."

650 Group: Campus Ethernet switch NaaS market grows 60% Y/Y

The needs of enterprises are different now and will continue to evolve away from pure CAPEX consumption models, according to a new report from 650 Group, which provides insight into the NaaS market for campus/enterprise switching. While managed service providers historically took vendor equipment to offer solutions as a service, vendors offering the solution directly via the channel have been a relatively new phenomenon over the past two years. 

“NaaS is not a new concept for Campus Ethernet Switch networks. However, it has historically been implemented by managed service providers (MSPs) or local resellers,” said Alan Weckel, Founder and Technology Analyst at 650 Group. “The NaaS market for campus switching is still in the early stages of adoption but is showing robust growth. Existing vendors like HPE Aruba along with new vendors like NileSecure vie for first mover advantage. During 2023, several vendors enhanced their NaaS offerings and expanded solutions toward MSP partners. We continue to see vendors shift revenue away from traditional one-time CAPEX models towards subscription and NaaS to meet customer demand. Customers want options and flexibility as hybrid and remote work are here to stay, and applications reside in a multi-cloud world.”

650 Group’s Quarterly Ethernet Switch – Campus Switching NaaS report includes segmentation on a regional and worldwide basis and provides market data for the China and non-China markets. It also looks at the adoption rate of NaaS compared to Cloud Managed Solutions and provides additional adoption metrics. The reports will also examine the vendor’s current revenue splits between traditional CAPEX and subscription-related revenue. Future reports will include additional segmentation and vendors.