OpenSearch

The multi-region shopping mall platform uses Amazon OpenSearch Service 2.11 to provide product search, autocomplete, and log analysis features. It uses the nori analyzer for Korean language search and enables UltraWarm tier for cost optimization.

Note

Unlike Aurora, DocumentDB, and ElastiCache, OpenSearch does not support global clusters. Independent domains are deployed in each region, and data synchronization is handled at the application level.

Architecture

Domain Specifications

Item	us-east-1	us-west-2
Domain Name	`production-os-use1`	`production-os-usw2`
Engine Version	OpenSearch 2.11	OpenSearch 2.11
Master Nodes	r6g.large.search x 3	r6g.large.search x 3
Data Nodes	r6g.xlarge.search x 6	r6g.xlarge.search x 6
Storage	500GB gp3 / node	500GB gp3 / node
UltraWarm	ultrawarm1.medium.search x 2	ultrawarm1.medium.search x 2
Availability Zones	3	3

Domain Name Limitation

OpenSearch domain names must be 28 characters or less. Therefore, region names are abbreviated:

us-east-1 -> use1
us-west-2 -> usw2

Connection Endpoints

Region	Endpoint
us-east-1	`https://vpc-production-os-use1-xxxxxxxxxxxxxxxxxxxxxxxxxxxx.us-east-1.es.amazonaws.com`
us-west-2	`https://vpc-production-os-usw2-yyyyyyyyyyyyyyyyyyyyyyyyyyyy.us-west-2.es.amazonaws.com`

Terraform Configuration

locals {
  # OpenSearch domain name must be <= 28 chars
  short_region = replace(replace(var.region, "us-east-", "use"), "us-west-", "usw")
  domain_name  = "${var.environment}-os-${local.short_region}"
}

resource "aws_opensearch_domain" "this" {
  domain_name    = local.domain_name
  engine_version = "OpenSearch_2.11"

  cluster_config {
    dedicated_master_enabled = true
    dedicated_master_type    = var.master_instance_type   # r6g.large.search
    dedicated_master_count   = var.master_instance_count  # 3

    instance_type  = var.data_instance_type   # r6g.xlarge.search
    instance_count = var.data_instance_count  # 6

    zone_awareness_enabled = true

    zone_awareness_config {
      availability_zone_count = 3
    }

    warm_enabled = var.enable_ultrawarm  # true
    warm_type    = var.enable_ultrawarm ? var.warm_instance_type : null
    warm_count   = var.enable_ultrawarm ? var.warm_count : null
  }

  ebs_options {
    ebs_enabled = true
    volume_type = "gp3"
    volume_size = var.ebs_volume_size  # 500
    iops        = 3000
    throughput  = 250
  }

  vpc_options {
    subnet_ids         = var.data_subnet_ids
    security_group_ids = [var.security_group_id]
  }

  encrypt_at_rest {
    enabled = true
  }

  node_to_node_encryption {
    enabled = true
  }

  domain_endpoint_options {
    enforce_https       = true
    tls_security_policy = "Policy-Min-TLS-1-2-PFS-2023-10"
  }

  advanced_security_options {
    enabled                        = true
    internal_user_database_enabled = true

    master_user_options {
      master_user_name     = "admin"
      master_user_password = var.master_user_password
    }
  }

  log_publishing_options {
    cloudwatch_log_group_arn = aws_cloudwatch_log_group.index_slow_logs.arn
    log_type                 = "INDEX_SLOW_LOGS"
  }

  log_publishing_options {
    cloudwatch_log_group_arn = aws_cloudwatch_log_group.search_slow_logs.arn
    log_type                 = "SEARCH_SLOW_LOGS"
  }

  log_publishing_options {
    cloudwatch_log_group_arn = aws_cloudwatch_log_group.es_application_logs.arn
    log_type                 = "ES_APPLICATION_LOGS"
  }
}

Korean Analyzer (Nori)

Analyzer Settings

{
  "settings": {
    "analysis": {
      "analyzer": {
        "korean_analyzer": {
          "type": "custom",
          "tokenizer": "nori_tokenizer",
          "filter": [
            "nori_readingform",
            "lowercase",
            "nori_part_of_speech"
          ]
        },
        "korean_search_analyzer": {
          "type": "custom",
          "tokenizer": "nori_tokenizer",
          "filter": [
            "nori_readingform",
            "lowercase",
            "nori_part_of_speech",
            "edge_ngram_filter"
          ]
        }
      },
      "tokenizer": {
        "nori_tokenizer": {
          "type": "nori_tokenizer",
          "decompound_mode": "mixed",
          "discard_punctuation": "true"
        }
      },
      "filter": {
        "nori_part_of_speech": {
          "type": "nori_part_of_speech",
          "stoptags": [
            "E", "IC", "J", "MAG", "MAJ",
            "MM", "SP", "SSC", "SSO", "SC",
            "SE", "XPN", "XSA", "XSN", "XSV",
            "UNA", "NA", "VSV"
          ]
        },
        "edge_ngram_filter": {
          "type": "edge_ngram",
          "min_gram": 1,
          "max_gram": 20
        }
      }
    }
  }
}

Index Mapping (products)

{
  "mappings": {
    "properties": {
      "productId": {
        "type": "keyword"
      },
      "name": {
        "type": "text",
        "analyzer": "korean_analyzer",
        "search_analyzer": "korean_search_analyzer",
        "fields": {
          "keyword": {
            "type": "keyword"
          },
          "suggest": {
            "type": "completion",
            "analyzer": "korean_analyzer"
          }
        }
      },
      "description": {
        "type": "text",
        "analyzer": "korean_analyzer"
      },
      "category": {
        "type": "keyword"
      },
      "categoryPath": {
        "type": "keyword"
      },
      "brand": {
        "type": "keyword"
      },
      "price": {
        "type": "scaled_float",
        "scaling_factor": 100
      },
      "discountPercent": {
        "type": "integer"
      },
      "rating": {
        "type": "float"
      },
      "reviewCount": {
        "type": "integer"
      },
      "attributes": {
        "type": "nested",
        "properties": {
          "name": { "type": "keyword" },
          "value": { "type": "keyword" }
        }
      },
      "tags": {
        "type": "keyword"
      },
      "createdAt": {
        "type": "date"
      },
      "updatedAt": {
        "type": "date"
      },
      "popularity": {
        "type": "rank_feature"
      }
    }
  }
}

Search Query Examples

Basic Product Search

{
  "query": {
    "bool": {
      "must": [
        {
          "multi_match": {
            "query": "삼성 갤럭시",
            "fields": ["name^3", "description", "brand^2"],
            "type": "best_fields",
            "fuzziness": "AUTO"
          }
        }
      ],
      "filter": [
        { "term": { "category": "smartphones" } },
        { "range": { "price": { "gte": 500000, "lte": 2000000 } } }
      ]
    }
  },
  "sort": [
    { "_score": "desc" },
    { "popularity": "desc" }
  ],
  "highlight": {
    "fields": {
      "name": {},
      "description": {}
    }
  }
}

Autocomplete (Suggest)

{
  "suggest": {
    "product-suggest": {
      "prefix": "갤럭시",
      "completion": {
        "field": "name.suggest",
        "size": 10,
        "skip_duplicates": true,
        "fuzzy": {
          "fuzziness": 1
        }
      }
    }
  }
}

Aggregations

{
  "query": {
    "match": { "name": "스마트폰" }
  },
  "aggs": {
    "brands": {
      "terms": { "field": "brand", "size": 20 }
    },
    "price_ranges": {
      "range": {
        "field": "price",
        "ranges": [
          { "to": 500000, "key": "Under 500K KRW" },
          { "from": 500000, "to": 1000000, "key": "500K-1M KRW" },
          { "from": 1000000, "to": 1500000, "key": "1M-1.5M KRW" },
          { "from": 1500000, "key": "Over 1.5M KRW" }
        ]
      }
    },
    "avg_rating": {
      "avg": { "field": "rating" }
    }
  }
}

UltraWarm Tier

Index State Management (ISM) Policy

{
  "policy": {
    "description": "Product index lifecycle policy",
    "default_state": "hot",
    "states": [
      {
        "name": "hot",
        "actions": [
          {
            "rollover": {
              "min_size": "50gb",
              "min_index_age": "7d"
            }
          }
        ],
        "transitions": [
          {
            "state_name": "warm",
            "conditions": {
              "min_index_age": "30d"
            }
          }
        ]
      },
      {
        "name": "warm",
        "actions": [
          {
            "warm_migration": {},
            "replica_count": {
              "number_of_replicas": 1
            }
          }
        ],
        "transitions": [
          {
            "state_name": "delete",
            "conditions": {
              "min_index_age": "365d"
            }
          }
        ]
      },
      {
        "name": "delete",
        "actions": [
          {
            "delete": {}
          }
        ]
      }
    ]
  }
}

Cross-Region Data Synchronization

Since OpenSearch does not support global clusters, data is synchronized to both regions through DocumentDB Change Stream.

Monitoring

Key Metrics

Metric	Description	Alarm Threshold
ClusterStatus.red	Cluster status	= 1
FreeStorageSpace	Remaining storage	< 20GB
JVMMemoryPressure	JVM memory pressure	> 85%
CPUUtilization	CPU usage	> 80%
SearchLatency	Search latency	> 200ms
IndexingLatency	Indexing latency	> 100ms

Next Steps

MSK Kafka - Event streaming
CloudFront & Edge - CDN configuration

Architecture​

Domain Specifications​

Connection Endpoints​

Terraform Configuration​

Korean Analyzer (Nori)​

Analyzer Settings​

Index Mapping (products)​

Search Query Examples​

Basic Product Search​

Autocomplete (Suggest)​

Aggregations​

UltraWarm Tier​

Index State Management (ISM) Policy​

Cross-Region Data Synchronization​

Monitoring​

Key Metrics​

Next Steps​