Scaled ARP Proxy for Data Center Broadcast Domain Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In massive data centers, the complexity of networked virtual machines (VMs) migrating between access segments leads to a significant scaling of layer-2 broadcast domains, resulting in inefficient ARP and ND protocol operations due to the need for broadcast messages to be sent across multiple segments, which increases network traffic and cache size.

Innovation Solution

Implementing network devices with address resolution processors at the edge of access segments to selectively transmit address resolution requests and update caches, allowing layer-2 broadcast domain reduction while maintaining unchanged layer-3 and layer-2 network addresses for VMs, even after migration, by using a scaled ARP (SARP) proxy and cache management to correlate layer-2 and layer-3 addresses.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If VMs are allowed to migrate between access segments while maintaining unchanged layer-2 and layer-3 network addresses, then VM mobility and network continuity are improved, but the layer-2 broadcast domain scales up significantly

Engineering Contradiction:
ImproveVM mobilityVSAvoidbroadcast domain
Core Design Contradiction:
Adaptability or versatilityVSArea of stationary object

Solution Approach 1:

The patent divides the large broadcast domain into multiple smaller broadcast domains by introducing border devices at access segment boundaries. Each border device maintains an address resolution cache that maps layer-3 addresses to layer-2 addresses across segments. This segmentation allows VMs to migrate between segments while keeping their addresses unchanged, but limits broadcast messages to only necessary segments, thus reducing overall broadcast domain size.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Border devices act as intermediaries between access segments. When a VM migrates to a different access segment, the border device at the new segment's boundary intercepts address resolution requests and provides the correct layer-2 address mapping without requiring full broadcast propagation across all segments. This intermediary function enables seamless VM migration while containing broadcast traffic.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If ARP and ND broadcast requests are sent across multiple access segments to resolve addresses, then address resolution completeness is improved, but network traffic and protocol overhead increase

Engineering Contradiction:
Improveaddress resolution completenessVSAvoidnetwork traffic
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

Border devices pre-populate address resolution caches by intercepting and processing address resolution requests before they propagate across multiple segments. When a border device receives an address resolution request, it stores the mapping information in its cache, enabling future requests to be resolved locally without broad broadcasting. This preliminary action ensures complete address resolution while minimizing subsequent network traffic.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements localized address resolution at each access segment boundary rather than uniform broadcasting across the entire network. Each border device independently manages its own address resolution cache and makes local decisions about whether to forward or satisfy address resolution requests. This local quality approach resolves addresses using only the minimum necessary network traffic while maintaining completeness.

Inventive Principle:
Principle #3Local quality

3Measurement precision

If the address resolution cache stores mappings for all VMs across all segments, then address resolution accuracy is improved, but cache size and memory requirements scale significantly

Engineering Contradiction:
Improveaddress resolution accuracyVSAvoidcache size
Core Design Contradiction:
Measurement precisionVSVolume of stationary object

Solution Approach 1:

The address resolution cache is segmented and distributed across multiple border devices rather than centralized in a single location. Each border device maintains a cache that primarily stores mappings for VMs in its local access segment and nearby segments. This segmentation reduces the cache size at each individual device while collectively maintaining accurate address resolution across the entire network through coordinated cache management.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentUS9992041B2Scaling address resolution for massive data centers
Publication Date: 2018.06.05 MARVELL ISRAEL (M L S L) LTD
  • US9992041B2 patent drawing
  • US9992041B2 patent drawing
  • US9992041B2 patent drawing

AI summary

There is provided a network device disposed at an interface between an access segment and an interconnecting layer of a data center. The network device includes an address cache and an address resolution processor configured to receive an address request addressed to virtual machines in a transmission domain of the network device. The address request requesting a layer 2 address of a target virtual machine in the data center, and specifying a layer 3 address of the target virtual machine. In response to receiving a reply, the network device updates the address cache to include an entry specifying the layer 2 address of an edge device of an access segment which has the target virtual machine having a respective layer 3 address corresponding to the specified layer 3 address.