Stronger Cyber Defense with Advanced CISO-Level Protection
Stronger Cyber Defense with Advanced CISO-Level Protection
Sophos Gold Partner
Monitoring & Support
Deployment Options
Nationwide Coverage
Arista Cloud Data Center Networking Solutions are built around a leaf-and-spine fabric, Arista’s EOS operating system, and CloudVision for management, giving data centers the flat, high-speed network cloud providers and AI workloads actually need. Instead of traffic bouncing through layers of traditional switches, every device in a leaf-spine design sits a short, predictable hop from every other device, running at speeds up to 800 gigabits per second on Arista’s newest platforms. That matters more in the UAE right now than almost anywhere else. Dubai alone accounts for nearly 57% of the UAE’s data center market, and the country’s data center sector is on track to roughly double in value by 2031, so the network underneath all that new capacity has to keep up.
Netmate IT Services is a certified Arista partner serving the UAE and GCC, working out of its Bur Dubai office with more than 20 certified network engineers who design and deploy Arista fabrics for data centers of every size. Arista is Netmate‘s trusted infrastructure partner for high-performance networking, and the company’s main focus stays on the UAE, with support extending into the wider GCC, Kenya, and Nepal for businesses running or planning a modern data center.
The older data centers weren’t always designed in a three-tiered fashion access, aggregation, and core and thus introduced hops and unpredictable latency as traffic flowed from one device to another. The leaf-spine model has been reduced to two layers by Arista. All the leaf switches are connected to all the spine switches, so as the fabric expands, users can reach any other server in the same number of hops. There are other instances where the predictability is useful, such as a bank with a trading platform or a hospital that is taking imaging data across racks.
Arista’s computers use a single operating system, EOS, regardless of whether it’s a switch in a small campus or an 800G spine switch. The single code base allows a network engineer to learn EOS once and manage the entire fabric without having to remember the different command sets across device families as the many competing products require. It is also based on an unmodified Linux kernel, and common Linux tools and scripts run directly on the switch.
CloudVision provides a network team with one dashboard to provision new switches, deploy configuration changes, and monitor telemetry throughout the fabric in real time. Zero-touch provisioning eliminates a lot of the manual setup work that typically takes a technician an afternoon to do by installing a new switch in the data center, racking it up, and configuring it. The other cool thing is that CloudVision also spins up configuration changes via a CI pipeline, testing and validating updates before they ever hit a live fabric, as opposed to shutting down a production rack.
This is how multiple virtual networks can coexist on a single physical fabric, separated from each other and independent of the need for multiple physical switches per tenant or department. A company with two or more client environments on common infrastructure, or even a bank that is separating its trading systems from its back office network, can operate both safely on the same physical switches.
Arista’s latest R4 series is the first to push 800-gigabit Ethernet to up to 25.6 terabits per second of total platform throughput. This sort of density is important for AI training clusters, where GPUs are required to transfer vast quantities of data amongst one another with no network becoming a bottleneck. For non-AI workloads, 800G spine switches provide a growing data center with years of expansion without needing another upgrade.
Each EOS switch pushes real-time state to Arista’s Network Data Lake (NetDL), which brings together telemetry, packet data, alerts, and even third-party system data in a single common model, rather than siloed logs in a variety of tools. That provides a network team with a single location to notice trends and identify issues early, before having to piece together reports from five distinct dashboards that are only showing what’s already happened.
Businesses turn to an Arista data center fabric when their existing network can’t keep pace with how much data moves between servers now, whether that’s a bank in Dubai’s DIFC processing trades in milliseconds, a cloud or hosting provider in Abu Dhabi running thousands of virtual machines across shared racks, a media company in Sharjah shuffling huge video files between storage and editing servers, or a research institution in Nairobi or Kathmandu building out its first real data center and wanting to start with an architecture that won’t need replacing in three years. What connects all of these is scale. A flat, high-speed fabric either handles tomorrow’s growth without a redesign, or it becomes the thing a business has to rip out and rebuild once traffic outgrows it.
Netmate IT designs and deploys Arista fabrics from its Bur Dubai office, backed by a team of more than 20 certified network engineers who handle everything from initial fabric design to EVPN VXLAN configuration and ongoing support. A leaf-spine migration is not a plug-and-play project. Getting the spine sizing wrong, or skipping proper EVPN design up front, tends to surface as a painful redesign eighteen months later once traffic has grown past what the original plan accounted for. Netmate’s engineers size the fabric against actual and projected traffic before a single switch gets ordered.
Netmate is Arista’s preferred infrastructure partner in the region, and supports the market in the UAE with secondary coverage in the GCC, Africa, and Nepal. It’s also a regionally focused solution with local hands-on support for installation, local stock and spares relationships, and support hours that follow the actual hours of the data center rather than a help desk operating in another time zone. Netmate can design, deploy, and support the relationship between the client and the fabric, whether it be a single rack spine deployment or a multi-site deployment across Abu Dhabi & Nairobi.
A traditional network typically consists of three layers: access, aggregation, and core, so traffic between two servers can require multiple hops across several switches based on their location. The leaf-spine fabric reduces that to two layers, so that any leaf switch is connected directly to any spine switch, regardless of which two switches are communicating, so the hop count stays the same and is predictable. This is why leaf-spine is the standard in any data center that needs to do a lot of east-west data exchange with servers.
A lot of migrations occur in stages, not one cutover. One way to do this is to add Arista leaf switches to new racks first and then bring these racks back into the core during a transition period, and then migrate the older racks over on an appropriate timeline that does not impact live production systems. The entire value of EVPN VXLAN overlays is achieved when the fabric is fully Arista, but with a phased rollout, the risk of one disruptive weekend migration is eliminated.
VXLAN provides virtual networks that are deployed atop the physical fabric, thereby enabling multiple, separate, and isolated networks to share the same switches. EVPN is a control plane that communicates the location of each device within the fabric to all the other switches in the fabric, thus allowing traffic to flow appropriately without having to broadcast all the way to the rest of the network to find its target. As a group, they enable multiple logically distinct environments to run safely on a single physical data center, which is significant for hosting providers and any enterprise with multiple business units or client environments sharing a common infrastructure.
It will be based on the traffic and not the size of the company. For years to come, 100G or 400G may be a good fit for any business using standard web and database workloads. However, a company with AI training, large-scale video processing, or big data analytics may find they are faced with bandwidth issues long before the headcount or revenue should indicate they are an enterprise-level business worthy of enterprise-level switching. The best answer is not based on company size, but rather on a company's east-west traffic volume.
Typically, a new switch must be set up with manual configuration, typically via a console cable, one switch at a time, before joining the network. The zero-touch provisioning process allows you to rack, cable, and power on a new switch and then connect to CloudVision and retrieve its entire configuration—without anyone's hands touching the console. This is what it does to a large deployment that used to take days to set up with dozens of switches; now it has a process that will run mostly on its own.
The segmentation, logging, and telemetry offered by Arista's platforms are designed to meet the typical requirements of regulated sectors in the UAE to demonstrate compliance with regulatory frameworks such as the UAE Information Assurance Regulation and focus on network visibility and access control. This does not mean that compliance is automatically granted with the architecture. It must be built and set up right for the specific regulatory requirements it is subject to, and that's where it's helpful to work with a partner who has a strong understanding of the technology and local compliance requirements.
This is one of the leading benefits of the design. As each leaf is linked to several spines, if one spine switch fails, the traffic continues to flow through the other spines, avoiding the failed spine. The fabric is designed from scratch to fail individual devices without losing connectivity, a significantly different failure mechanism than previous designs, where a failure of a core switch would completely affect the fabric behind it.
