Disabling Qos rules (within the network) normalized critical performance issue with large data upload

Products

• ISG Proxy
• ProxySG Software - SGOS

Issue/Introduction

Quality of Service (QoS) refers to a broad set of networking technologies and techniques designed to guarantee predictable levels of network performance. QoS is critical in networks to ensure that certain traffic receives higher priority, especially in situations where network capacity is limited or becomes congested. The primary aspects of network performance that QoS aims to manage include bandwidth (data transfer rate), latency (delay in data transmission), jitter (variation in delay), and packet loss.

In the context of QoS, various types of network traffic are prioritized based on their importance or sensitivity to delay. For example, voice over IP (VoIP) and video conferencing data might be given higher priority over regular web traffic or file downloads, as they are more sensitive to latency and jitter.

In this particular scenario, no visible bandwidth limitation is found on the Edge SWG (ProxySG).

Cause

This isn't about bandwidth limitation.

The improvement in data upload speeds after disabling QoS, despite no apparent bandwidth limitations on your proxy appliance, can be explained by several factors related to how QoS policies can affect network traffic:

• Implicit Prioritization: Even if there's no explicit bandwidth limitation set by the proxy appliance, QoS policies might still prioritize certain types of traffic over others. This could inadvertently de-prioritize your upload traffic. When QoS is disabled, this implicit prioritization is removed, potentially allowing for more equitable bandwidth distribution and faster uploads.

• Complex Traffic Shaping: QoS involves more than just bandwidth limits. It encompasses various forms of traffic shaping and management. These might include packet scheduling, where packets are queued and sent based on priority, or traffic classification, which may inadvertently affect your uploads. Disabling QoS simplifies traffic handling, possibly eliminating these unintended impacts.

• Reduced Overhead: QoS mechanisms themselves can introduce processing overhead. Each packet might be inspected and classified, which takes time and processing power. Disabling QoS removes this overhead, allowing data packets to be processed and forwarded more quickly.

• Burst Traffic Handling: QoS policies often handle burst traffic in specific ways to prevent network congestion. This could inadvertently slow down uploads if the policy is not optimized for the type of data being uploaded. Without QoS, the network might handle bursty upload traffic more efficiently.

• TCP Window Size and Flow Control: QoS policies can influence TCP window sizing and flow control mechanisms, which are crucial for determining how much data can be sent before requiring an acknowledgment. In some cases, QoS settings might cause these windows to be smaller or the flow to be more tightly controlled, slowing down uploads. Disabling QoS might result in more favorable TCP flow control conditions for uploads. 

• Interaction with Other Network Devices: QoS settings on one device (like your proxy appliance) can interact in complex ways with QoS or traffic management policies on other network devices (like routers, switches, or firewalls). This interaction can sometimes produce unexpected results, such as slowing down uploads.

• Resource Allocation: QoS often involves dynamic allocation of network resources based on current conditions and policies. If the QoS algorithm isn't optimized for your specific network environment or traffic patterns, it might allocate resources in a way that doesn't favor upload traffic. Disabling QoS might lead to a more favorable allocation for uploads.

In conclusion, while QoS is designed to optimize network performance, its complexity means that it can sometimes have unintended consequences, especially in mixed-traffic environments or when the QoS policies are not perfectly tailored to the specific needs of the network. Disabling QoS can, in some cases, lead to simpler and more efficient traffic handling, thereby improving certain aspects of network performance such as upload speeds.

Resolution

 

Resolution:

To implement Quality of Service (QoS) in a way that does not negatively impact data uploads, especially in environments where TCP ZeroWindow issues are observed, you need to carefully design and configure your QoS policies. The goal is to ensure efficient network traffic management while avoiding unintended bottlenecks. Here are some guidelines:

• Understand Traffic Patterns: Analyze the types of traffic on your network. Understanding the bandwidth needs and characteristics of different traffic types (e.g., uploads, downloads, real-time applications like VoIP) is crucial for effective QoS configuration.

• Prioritize Traffic Wisely: Prioritize traffic based on its importance and sensitivity to delay. For instance, real-time applications might need higher priority, but this shouldn't come at the expense of completely starving other traffic types like uploads. Ensure there's a balance.

• Dynamic Bandwidth Allocation: Implement dynamic bandwidth allocation strategies that can adjust to varying network conditions. Instead of rigidly capping certain types of traffic, use algorithms that can provide more bandwidth to uploads when the network is underutilized.

• Buffer Management: Configure buffer sizes and queue management policies appropriately. Proper buffer sizing can prevent overflow and underflow, reducing the occurrence of TCP ZeroWindow conditions. Techniques like Tail Drop, Random Early Detection (RED), or Weighted Random Early Detection (WRED) can be used for managing buffers efficiently.

• Monitor and Adjust: Continuously monitor network performance. Use network monitoring tools to track the effectiveness of your QoS policies. Be prepared to adjust QoS settings based on real-time feedback and changing network conditions.

• TCP Optimization: On networks with high latency or packet loss (like WANs), consider using TCP optimization techniques. These can include TCP window size adjustments or advanced TCP variants designed for specific network environments.

• Avoid Overly Aggressive QoS: Be cautious with overly aggressive QoS policies that severely limit certain types of traffic. If upload traffic is essential, ensure that its QoS policy doesn't reduce its bandwidth to a trickle.

• Test Before Deployment: Before rolling out new QoS policies network-wide, test them in a controlled environment or a limited segment of your network. This helps identify potential issues before they impact the entire network.

• User Feedback: Pay attention to user feedback. If users start experiencing problems after QoS policies are implemented, it may be a sign that the policies need to be revisited.

• Expert Consultation: If possible, consult with network experts, especially for complex environments. They can provide insights into best practices and help tailor QoS settings to your specific network needs.

Remember, the goal of QoS is not just to prioritize certain traffic types but to ensure the overall efficient and fair use of network resources. Balancing different types of traffic and continuously monitoring and adjusting your QoS policies are key to a successful implementation.