Skip to content

Commit

Permalink
Fixed broken links.
Browse files Browse the repository at this point in the history
  • Loading branch information
PahudPlus committed Jul 15, 2020
1 parent 5887765 commit 98b7035
Show file tree
Hide file tree
Showing 6 changed files with 19 additions and 30 deletions.
18 changes: 2 additions & 16 deletions site/en/menuStructure/en.json
Original file line number Diff line number Diff line change
Expand Up @@ -74,22 +74,8 @@
"label3": "",
"order": 2
},
{
"id": "storage_concept.md",
"title": "Storage concepts",
"label1": "milvus_basics",
"label2": "",
"label3": "",
"order": 3
},
{
"id": "write_ahead_log.md",
"title": "Write Ahead Log",
"label1": "milvus_basics",
"label2": "",
"label3": "",
"order": 4
},


{
"id": "terms.md",
"title": "Milvus Terms",
Expand Down
5 changes: 3 additions & 2 deletions site/en/milvus_basics/write_ahead_log.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ id: write_ahead_log.md

# Write Ahead Log

![wal_structure](wal_workflow.jpg)
![wal_structure](../../../assets/wal/wal_workflow.jpg)

Write ahead log records the insertion and deletion requests into the log file, and then the background thread writes it to the system. Once the requests are successfully written to the log file, the server returns success. This function enhances data reliability and reduce client blocking.

Expand All @@ -17,10 +17,11 @@ Write ahead log guarantees the atomicity of modification requests. All requests
The buffer size of the write ahead log is determined by the `wal.buffer_size`. To ensure the write performance of the write ahead log, it is recommended to set the buffer size to at least twice the size of the data imported in a single batch.

<div class="alert info">
For how to set <code>wal.buffer_size</code>, see <a href="configuration.md">Milvus configuration</a>.
For how to set <code>wal.buffer_size</code>, see <a href="milvus_config.md">Milvus configuration</a>.
</div>



## Delete old log files

Milvus automatically deletes log files that have been applied to the system.
2 changes: 1 addition & 1 deletion site/zh-CN/quick_start/install_milvus/cpu_milvus_docker.md
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ id: cpu_milvus_docker.md
| Docker | 19.03 或以上 |

<div class="alert note">
请确保可用内存大于你在 <b>milvus.yaml</b> 文件中设置的 <code>cache.insert_buffer_size</code> 和 <code>cache.cache_size</code> 之和。
请确保可用内存大于你在 <b>server_config.yaml</b> 文件中设置的 <code>cache.insert_buffer_size</code> 和 <code>cache.cache_size</code> 之和。
</div>

## 确认 Docker 状态
Expand Down
16 changes: 8 additions & 8 deletions site/zh-CN/reference/metric.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ Milvus 基于不同的距离计算方式比较向量间的距离。选择合适

欧氏距离的计算公式为:

![euclidean](euclidean_metric.png)
![euclidean](../../../assets/euclidean_metric.png)

其中 **a** = (a1, a2,..., an) 和 **b** = (b1, b2,..., bn) 是 n 维欧氏空间中的两个点。

Expand All @@ -31,7 +31,7 @@ Milvus 基于不同的距离计算方式比较向量间的距离。选择合适

内积距离的计算公式为:

![ip](ip_metric.png)
![ip](../../../assets/ip_metric.png)

假设有 A 和 B 两条向量,则 `||A||``||B||` 分别代表 A 和 B 归一化后的值。cosθ 代表 A 与 B 之间的余弦夹角。

Expand All @@ -43,27 +43,27 @@ Milvus 基于不同的距离计算方式比较向量间的距离。选择合适

杰卡德相似系数计算数据集之间的相似度,计算方式为:数据集交集的个数和并集个数的比值。计算公式可以表示为:

![Jaccard similarity coefficient](jaccard_coeff.png)
![Jaccard similarity coefficient](../../../assets/jaccard_coeff.png)

杰卡德距离是用来衡量两个数据集差异性的一种指标,被定义为 1 减去杰卡德相似系数。对于二值变量,杰卡德距离等价于谷本系数。

![Jaccard distance](jaccard_dist.png)
![Jaccard distance](../../../assets/jaccard_dist.png)

杰卡德距离适合字符串相似性度量。

### 谷本距离

对于二值变量,谷本距离公式可表示为:

![tanimoto distance](tanimoto_dist.png)
![tanimoto distance](../../../assets/tanimoto_dist.png)

在 Milvus 中,谷本距离仅支持二值变量。

值域从 0 到正无穷。

对于二值变量,谷本系数等价于杰卡德距离:

![tanimoto coefficient](tanimoto_coeff.png)
![tanimoto coefficient](../../../assets/tanimoto_coeff.png)

对于二值变量,谷本系数值域为 0 到+1(+1 的相似度最高)

Expand All @@ -83,7 +83,7 @@ Milvus 基于不同的距离计算方式比较向量间的距离。选择合适

超结构的公式可表示为:

![superstructure](superstructure.png)
![superstructure](../../../assets/superstructure.png)

其中

Expand All @@ -98,7 +98,7 @@ Milvus 基于不同的距离计算方式比较向量间的距离。选择合适

子结构的公式可表示为:

![substructure](substructure.png)
![substructure](../../../assets/substructure.png)

其中

Expand Down
2 changes: 1 addition & 1 deletion site/zh-CN/reference/monitor.md
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,7 @@ Milvus 会收集监控数据并将其推送到 Pushgateway。同时,Prometheus

Grafana 是一个开源的时序数据分析及可视化平台。Milvus 使用 Grafana 来展示各项性能指标:

![prometheus.png](prometheus.png)
![prometheus.png](../../../assets/prometheus.png)

### 需要报警的事件

Expand Down
6 changes: 4 additions & 2 deletions site/zh-CN/reference/storage_operation.md
Original file line number Diff line number Diff line change
Expand Up @@ -103,7 +103,7 @@ Milvus 使用布隆过滤器(bloom filter)来快速判断一个实体 ID 是
* 否则,根据每个分段的布隆过滤器判断该实体所处的分段,然后更新该分段的 **delete_docs** 以及 **bloom_filter** 文件。

## 数据段整理

查询一个分段时,Milvus 会将该分段的实体数据以及 **delete_docs** 文件读入内存。虽然被删除的实体不参与计算,但它们也会被读入内存。所以,一个分段中被删除的实体越多,浪费的内存资源和磁盘空间越多。为了减少此类不必要的资源消耗,Milvus 提供了数据段整理(compact)的操作,流程如下:

1. 客户端调用 `compact` 接口。
Expand All @@ -112,9 +112,11 @@ Milvus 使用布隆过滤器(bloom filter)来快速判断一个实体 ID 是
<div class="alert note">
<code>compact<code> 操作会忽略被删除向量占比小于 10% 的分段。
</div>



## 数据读取


1. 客户端调用 `get_entity_by_id` 接口读取原始实体数据。
2. 服务端接收到请求后,通过布隆过滤器找到实体所在的段,返回该实体 ID 对应的数据。

Expand Down

0 comments on commit 98b7035

Please sign in to comment.