Equipment / Dell Technologies / Network
The network determines the speed of everything else.
Servers and storage systems may have free resources and still be on hold.The network connects users, servers, storage systems, applications, and externally hosted services. With Dell switches PowerSwitch, we design an architecture scaled to your actual traffic—enterprise networking, datacenter, virtualization, remote sites, or accelerated computing.
Poorly designed architecture holds back what works well
In a “ datacenter, ” servers and storage may have unused resources but still experience delays. Virtualized and hyperconverged environments, as well as computing clusters, exchange a huge amount of data with one another: it is this internal data exchange that becomes saturated first, and it does not show up on any CPU performance graph.
Not all exchanges require the same thing
Users want high availability. Storage systems want low, consistent latency. Backups need high throughput at night. Artificial intelligence wants all three at the same time.
A single flow rate figure does not account for any of these four needs.
Redundancy is built into the design; it is not added on
Redundant a link after the fact often leaves a single point of failure in place: a power supply, a module, or a board. Resilience is determined in the design phase, before the order is placed.
A dual link on a single piece of equipment is still considered a single piece of equipment.
Each new piece of equipment adds another configuration
When set manually, ten switches end up having ten different configurations. The discrepancy isn't apparent as long as everything is working, but it becomes fully apparent the day a failure occurs.
Anything that isn't automated eventually drifts off course.
Four network levels, each with its own speed
The Dell PowerSwitch product line covers user connectivity, datacenter switching, the network core, and ultra-high-speed links. The fourth level is the one that gets cut from the quote, and it’s the only one left when the rest have been eliminated.
Connecting Users and Equipment
from copper to fiber, with power delivered over the cableWorkstations, phones, wireless access points, cameras, and everything else that plugs into an office. Depending on the model: copper or fiber connections, power over Ethernet, segmentation, redundancy, and centralized management.
This is where local security is determined: the decision to separate the cameras from the stations, and the stations from the guests, is made here.
Connect the servers and storage
from 1 gigabit to 400 gigabits per seconddatacenter switching platforms: servers, storage systems, and virtualized infrastructures. They can be placed at the top of the rack, in an intermediate position, or in an aggregation configuration.
It is this factor that determines the actual performance of a virtualized or hyperconverged platform, far more often than the number of cores.
The heart of the datacenter and its fastest connections
up to 800 gigabits per secondUltra-high-speed switches designed for the core of the datacenter and for aggregation. They are suited for highly distributed infrastructures, large-scale private environments, high-performance computing, and the internal networks required by artificial intelligence.
A model training cluster exchanges more data among its own nodes than with the rest of the company: it is this traffic that determines the capacity of this layer.
An administration path separate from the production environment
the line item that is removed from the estimateSwitches dedicated to managing servers, storage, and network equipment. They maintain access to the infrastructure for administration, diagnostics, and troubleshooting— even when the production network is down.
That's the line we're getting rid of because it's never used. It's used just once, and on that day it determines whether we'll fix it remotely or take the car.
Network modules in the modular chassis
shared cabling within the chassisNetwork modules that integrate connectivity directly into the modular PowerEdge MX chassis. Cabling is simplified, connections are shared, and both computing and networking scale through incremental additions within the same infrastructure.
The network switch only does half the work
You also choose a switch based on what runs on it and how you configure it. That’s what determines how much time your team will spend on it each week for the next five years.
Dell's Network Operating System
It provides the features expected of an enterprise and e datacenter network, and is designed to be automated, scheduled, and integrated with existing management tools.
An open distribution, to separate hardware from software
Built on SONiC, it is suitable for open environments involving multiple vendors. It enables the creation of standardized, automatable architectures without tying the operating system to the hardware on which it runs.
Deployment Automation
Some of the configuration, interconnection, and modifications occur automatically when new equipment is added to the infrastructure. This is what prevents ten switches from becoming ten special cases.
Describe what the network needs to do, not each switch
The network integrates with virtualization, orchestration, and automation platforms. Configurations are centralized, and connectivity is tailored to application needs rather than being managed on a per-device basis.
The Right Network for Your Data Transfers
The goal is not to deploy the fastest network, but to provide each use case with the bandwidth, resilience, and latency it actually needs. This requires looking at what’s actually being transmitted, not just what’s connected.
Record what's circulating
Connected devices, data volumes transferred, peak times, internal data transfers between servers and storage, and data sent externally.
This reading almost always shows that the saturation point is not where we expected it to be.
Designing the Architecture, Including Redundancy
Levels, flow rates, connections, number of units, and the backup path for each one. Redundancy is included in the initial design, not in a second phase.
The separate administration network is also included there, and it remains there.
Choose how the network will be managed
SmartFabric OS10 or Dell Enterprise SONiC, automation, policy-based management, integration with existing tools.
This decision will determine how much time your team spends on this network each week. We’re making this decision together with you right now, because it depends as much on your in-house expertise as it does on the equipment.
Switch over without interruption
Phased switchover, testing of each fallback path under real-world conditions, and delivery of documentation that can be used by a team other than ours.
An emergency route that has never been tested is not an emergency route.
The scope is defined before we begin. This includes what we’ll handle, what you’ll retain, the option to reverse course, and a clearly stated limit. The same engineers will support you from the site survey through commissioning.
What Complements Your Network
A switch never comes alone. Here's where to find the rest.
The servers connected to this network
The four product lines— PowerEdge —and the chain that connects what you measure to what you buy.
View servers StorageStorage platforms that depend on this latency
Block, file, and object; six platforms; and what those reduction rates actually mean.
View Storage Dell TechnologiesThe entire Dell product line, from computers to datacenter
Hyperconvergence, data protection, and preparing hardware before delivery.
View the entire collectionWhen decisions are made on-site, the network determines the latency. Factories, warehouses, and remote sites that runartificial intelligence right next to the machine—or that maintain a local component in a hybrid architecture—rely on these switches.
Let's talk about your network
Tell us what traffic is flowing, and we'll tell you where the network is overloaded
What you connect with today, what resonates most, and the moments when you feel it. With these three elements, the first meeting is already productive.
What we offer is a meeting with the engineers who will do the work. The scope of work and sizing are part of the project. The preliminary discussion, however, costs nothing and is often enough to determine whether the project is feasible.
Renens, Sion, Châtel-Saint-Denis
Microsoft Gold Partner and a Gold Partner Dell Technologies. A single point of contact for sizing, ordering, commissioning, and support.
Renens VD +41 21 806 37 15
Sion VS +41 27 552 00 22
Châtel-Saint-Denis FR +41 26 322 59 05