Low Latency Network Design
Low-latency network design has become a critical aspect of modern network architecture in today’s fast-paced digital world, where speed is of the essence. Low latency refers to the minimal delay or lag experienced during data transmission, ensuring quick and efficient communication between devices. This blog post explores the importance of low-latency network design, its benefits, and critical considerations for implementing such networks.
Latency, often measured in milliseconds (ms), is the time data travels from its source to its destination. In network communications, latency can significantly impact user experience, especially in applications that require real-time data, such as online gaming, video streaming, and financial transactions. High latency can lead to delays, buffering, and even disconnections, frustrating user experience.
Highlights: Low Latency Network Design
- A New Operational Model
We are; now all moving in the cloud direction. The requirement is for large data centers that are elastic and scalable. The result of these changes that are influenced by innovations and methodology in the server/application world is that the network industry is experiencing a new operational model. Provisioning must be quick, and designers look to automate network configuration more systematically and in a less error-prone programmatic way. Difficult to meet these new requirements with traditional data center designs.
- Changing Traffic Flow
Traffic flow has changed, and we have a lot of east-to-west traffic. Existing data center designs are designed to focus on north-to-south flows. East-to-west traffic requires changing the architecture from an aggregating-based model to a massive multipathing model. Referred to as Clos networks, leaf and spine designs allow for building huge networks with reasonably sized equipment enabling low latency network design.
Latency In Networking. |
|
Before you proceed, you may find the following post helpful:
- Baseline Engineering
- Dropped Packet Test
- SDN Data Center
- Azure ExpressRoute
- Zero Trust SASE
- Service Level Objectives (slos)
Forwarding Features | Control Features |
Network and Storage integration | Bridging without STP |
Multi pathing for Layer 2 and Layer 3 | Integration with server virtualization |
Low latency | Good MAC, ARP and L3 table size |
Optimal Layer 3 forwarding | Deep packet buffers |
Path isolation |
Back to basics with network testing.
Network Testing
A stable network results from careful design and testing. Although many vendors often perform exhaustive systems testing and provide this via 3rd party testing reports, they cannot reproduce every customer’s environment. So to determine your primary data center design, you must conduct your tests.
Effective testing is the best indicator of production readiness. On the other hand, ineffective testing may lead to a false sense of confidence, causing downtime. Therefore, you should adopt a structured approach to testing as the best way to discover and fix the defects in the least amount of time at the lowest possible cost.
- A key point: Lab Guide on RSVP.
In this example, we will have a look at RSVP. Resource reservation signals the network and requests a specific bandwidth and delay required for a flow. When the reservation is successful, each network component (primarily routers) will reserve the necessary bandwidth and delay.
- First, we need to enable RSVP on all interfaces: ip rsvp bandwidth 128 64
- Then, configure R1 to act like an RSVP host so it will send an RSVP send path message:
- Finally. Configure R4 to respond to this reservation:
What is low latency?
Low latency is the ability of a computing system or network to respond with minimal delay. Actual low latency metrics vary according to the use case. So, what is a low-latency network? A low-latency network has been designed and optimized to reduce latency as much as possible. However, a low-latency network can only improve latency caused by factors outside the network.
We first have to consider latency jitters when it deviates unpredictably from an average; in other words, it is low at one moment and high at the next. For some applications, this unpredictability is more problematic than high latency. We also have ultra-low latency measured in nanoseconds, while low latency is measured in milliseconds. Therefore, ultra-low latency delivers a response much faster, with fewer delays than low latency.
Importance of Low Latency Network Design:
1. Improved User Experience: Low latency networks ensure seamless and uninterrupted communication, enabling users to access and transmit data more efficiently. This is particularly crucial in latency-sensitive applications where any delay can be detrimental.
2. Competitive Advantage: In today’s competitive business landscape, organizations that deliver faster and more responsive services gain a significant edge. Low latency networks enable companies to provide real-time services, enhancing customer satisfaction and loyalty.
3. Support for Emerging Technologies: Low latency networks form the backbone for emerging technologies such as the Internet of Things (IoT), autonomous vehicles, augmented reality (AR), and virtual reality (VR). These technologies require rapid data exchange and response times, which can only be achieved through low-latency network design.
Low Latency Network Design:
Data Center Latency Requirements
- Latency requirements
Intra-data center traffic flows concern us more with latency than outbound traffic flow. High latency between servers degrades performance and results in the ability to send less traffic between two endpoints. Low latency allows you to use as much bandwidth as possible.
A low-lay network design known as Ultra-low latency ( ULL ) data center design is the race to zero. The goal is to design as fast as possible with the lowest end-to-end latency. Latency on an IP/Ethernet switched network can be as low as 50 ns.
High-frequency trading ( HFT ) environments push for this trend, where providing information from stock markets with minimal delay is imperative. HFT environments are different than most DC designs and don’t support virtualization. The Port count is low, and servers are designed in small domains.
Conceptually similar to how Layer 2 domains should be designed as small Layer 2 network pockets. Applications are grouped to match optimum traffic patterns where many-to-one conversations are reduced. This will reduce the need for buffering, increasing network performance. CX-1 cables are preferred over the more popular optical fiber.
Data Center Latency Requirements
Oversubscription
The optimum low-latency network design should consider and predict the possibility of congestion at critical network points. An example of unacceptable oversubscription would be a ToR switch with 20 Gbps traffic from servers but only 10 Gbps uplink. This will result in packet drops and poor application performance.
Previous data center designs were 3-tier aggregation model-based ( developed by Cisco ). Now, we are going for 2-tier models. The main design point for this model is the number of ports on the core; more ports on the core result in more extensive networks. Similar design questions would be a) how much routing and b) how much bridging will I implement c) where do I insert my network services modules?
We are now designing networks with lots of tiers – Clos Network. The concept comes from voice networks from around 1953, previously built voice switches with crossbar design. Clos designs give optimum any-to-any connectivity. Requires low latency and non-blocking components. Every element should be non-blocking. Multipath technologies deliver a linear increase in oversubscription with each device failure and are better than architectures that degrade during failures.
Lossless transport
Data Center Bridging ( DCB ) offers standards for flow control and queuing. Even if your data center does not use (the Internet Small Computer System Interface) ISCSI, TCP elephant flows benefit from lossless transport, improving data center performance. However, research has shown that many TCP flows are below 100Mbps.
The remaining small percentage are elephant flows but consume 80% of all traffic inside the data center. Due to their size and how TCP operates, when an elephant flows experience packet drops, they will slow down, affecting network performance.
Distributed resource scheduling
VMmobiliy is a VMware tool used for distributed resource scheduling. Load from hypervisors is automatically spread to other underutilized VMs. Other use cases in cloud environments where DC requires dynamic workload placement, and you don’t know where the VM will be in advance.
If you want to retain sessions, keep them in the same subnet. Layer 3 VMotion is too slow as routing protocol convergence will always take a few seconds. In theory, you could optimize timers for routing protocol fast convergence, but in practice, Interior Gateway Protocols ( IGP ) give you eventual consistency.
VMmobiliy
Data Centers require bridging at layer 2 to retain the IP addresses for VMobility. TCP stack currently has no separation between “who” and “where” you are, i.e., IP address represents both functions. Future implementation with Locator/ID Separation Protocol ( LISP ) divides these two roles, but bridging for VMobility is required until fully implemented.
Spanning Tree Protocol ( STP )
Spanning Tree reduces bandwidth by 50%, and massive multipathing technologies allow you to scale without losing 50% of the link bandwidth. Data centers want to move VMs without distributing traffic flow. VMware has VMotion. Microsoft Hyper-V has Live migration.
Network convergence
Layer 3 network requires many events to complete before it reaches a fully converged state. Layer 2 when the first broadcast is sent, every switch knows precisely where that switch has moved. There are no mechanisms with Layer 3 to do something similar. Layer 2 networks result in a large broadcast domain.
You may also experience large sub-optimal flows as the Layer 3 next hop will stay the same when you move the VM. Optimum Layer 3 forwarding – what Juniper is doing with Q fabric. Every Layer 3 switch has the same IP address; they can all serve as the next hop—resulting in optimum traffic flow.
Data center network design: Deep packet buffers
We have more DC traffic and elephant flows from distributed databases. Traffic is now becoming very bursty. We also have a lot of microburst traffic. The bursts are so short that they don’t register as high link utilization but are big enough to overflow packet buffers and cause drops. This type of behavior with TCP causes TCP slow start. A slow start with elephant flows is problematic for networks.
Key Considerations for Low Latency Network Design:
1. Network Infrastructure: To achieve low latency, network designers must optimize the infrastructure by reducing bottlenecks, eliminating single points of failure, and ensuring sufficient bandwidth capacity.
2. Proximity: Locating servers and data centers closer to end-users can significantly reduce latency. Data can travel faster by minimizing the physical distance, resulting in lower latency.
3. Traffic Prioritization: Prioritizing latency-sensitive traffic within the network can help ensure that critical data packets are given higher priority, reducing the overall latency.
4. Quality of Service (QoS): Implementing QoS mechanisms allows network administrators to allocate resources based on application requirements. By prioritizing latency-sensitive applications, low latency can be maintained.
5. Optimization Techniques: Various optimization techniques, such as caching, compression, and load balancing, can further reduce latency by minimizing the volume of data transmitted and distributing the workload efficiently.
Conclusion:
Low-latency network design is vital in delivering fast and responsive services in an increasingly connected world. By minimizing delays in data transmission, low latency networks enhance user experience, provide a competitive advantage, and support the seamless operation of emerging technologies.
Network designers must consider critical factors such as network infrastructure, proximity, traffic prioritization, QoS, and optimization techniques to achieve low latency and meet the ever-growing demands of today’s digital landscape.
- DMVPN - May 20, 2023
- Computer Networking - April 7, 2023
- eBOOK – SASE Capabilities - April 6, 2023
[…] Within the DC, every point is connected together and you can assume unconstrained capacity. A common data centre design is a leaf and spine architecture, where all nodes have equidistant endpoints. This is not the case in the WAN. The WAN has […]