[
https://issues.apache.org/jira/browse/CHUKWA-747?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Eric Yang updated CHUKWA-747:
-----------------------------
Description:
It is somewhat a waste to use HDFS file to host widget information because HDFS
files are designed to be big data blocks, and namenode RAM is too expensive to
host meta data that is only a few bytes. Therefore, it would be more efficient
to store widget information on HBase.
The current widget information looks like this:
{code}
{
"id":"graph_explorer",
"title":"Graph Explorer",
"version":"0.1",
"categories":"Developer,Utilities",
"url":"iframe/jsp/graph_explorer.jsp",
"description":"Graph explorer for visualize data on Hbase.",
"refresh":"0",
"parameters":[
{
"name":"width",
"type":"select",
"value":"300",
"label":"Width",
"options":[
{
"label":"300",
"value":"300"
}
]
}
]
}
{code}
However, it is possible to reduce the required information to only:
{code}
{
"title" : "System Load Average",
"url" : "/hicc/chart/draw/1"
}
{code}
Title is already the unique identifier of a subject at hand. URL is necessary
to load an iframe. The rest like version, description, id, category. They are
useful to organize the data, if a tree structure make sense to organize the
data. However, in BigData environment, the tree structure is too limited to
help user to find the information. Hence, we promote usage of search box to
find the widget base on title. Parameters were used to customize the look and
feel of the widget. Those parameters are stored in server side to improve
efficiency of client and server communication.
We need the following REST API:
1. GET /hicc/v2/widget/list?limit=1000&offset=0 - List all widgets with
pagination
2. GET /hicc/v2/widget/search/[query] - find widget based on title matching
3. GET /hicc/v2/widget/view/[title] - get a single widget
4. POST /hicc/v2/widget/create - create a new widget
5. PUT /hicc/v2/widget/update/[title] - update a widget
6. DELETE /hicc/v2/widget/delete/[title] - delete a widget
was:
It is somewhat a waste to use HDFS file to host widget information because HDFS
files are designed to be big data blocks, and namenode RAM is too expensive to
host meta data that is only a few bytes. Therefore, it would be more efficient
to store widget information on HBase.
The current widget information looks like this:
{code}
{
"id":"graph_explorer",
"title":"Graph Explorer",
"version":"0.1",
"categories":"Developer,Utilities",
"url":"iframe/jsp/graph_explorer.jsp",
"description":"Graph explorer for visualize data on Hbase.",
"refresh":"0",
"parameters":[
{
"name":"width",
"type":"select",
"value":"300",
"label":"Width",
"options":[
{
"label":"300",
"value":"300"
}
]
}
]
}
{code}
However, it is possible to reduce the required information to only:
{code}
{
"title" : "System Load Average",
"url" : "/hicc/chart/draw/1"
}
{code}
Title is already the unique identifier of a subject at hand. URL is necessary
to load an iframe. The rest like version, description, id, category. They are
useful to organize the data, if a tree structure make sense to organize the
data. However, in BigData environment, the tree structure is too limited to
help user to find the information. Hence, we promote usage of search box to
find the widget base on title. Parameters were used to customize the look and
feel of the widget. Those parameters are stored in server side to improve
efficiency of client and server communication.
We need the following REST API:
1. GET /hicc/v2/widget/list?limit=1000&offset=0 - List all widgets with
pagination
2. GET /hicc/v2/widget/search/[query] - find widget based on title matching
3. GET /hicc/v2/widget/view/[title] - get a single widget
4. POST /hicc/v2/widget/create - create a new widget
5. PUT /hicc/v2/widget/update/[title] - update a widget
> Update Widget storage to HBase
> ------------------------------
>
> Key: CHUKWA-747
> URL: https://issues.apache.org/jira/browse/CHUKWA-747
> Project: Chukwa
> Issue Type: Sub-task
> Components: User Interface
> Reporter: Eric Yang
> Assignee: Eric Yang
>
> It is somewhat a waste to use HDFS file to host widget information because
> HDFS files are designed to be big data blocks, and namenode RAM is too
> expensive to host meta data that is only a few bytes. Therefore, it would be
> more efficient to store widget information on HBase.
> The current widget information looks like this:
> {code}
> {
> "id":"graph_explorer",
> "title":"Graph Explorer",
> "version":"0.1",
> "categories":"Developer,Utilities",
> "url":"iframe/jsp/graph_explorer.jsp",
> "description":"Graph explorer for visualize data on Hbase.",
> "refresh":"0",
> "parameters":[
> {
> "name":"width",
> "type":"select",
> "value":"300",
> "label":"Width",
> "options":[
> {
> "label":"300",
> "value":"300"
> }
> ]
> }
> ]
> }
> {code}
> However, it is possible to reduce the required information to only:
> {code}
> {
> "title" : "System Load Average",
> "url" : "/hicc/chart/draw/1"
> }
> {code}
> Title is already the unique identifier of a subject at hand. URL is
> necessary to load an iframe. The rest like version, description, id,
> category. They are useful to organize the data, if a tree structure make
> sense to organize the data. However, in BigData environment, the tree
> structure is too limited to help user to find the information. Hence, we
> promote usage of search box to find the widget base on title. Parameters
> were used to customize the look and feel of the widget. Those parameters are
> stored in server side to improve efficiency of client and server
> communication.
> We need the following REST API:
> 1. GET /hicc/v2/widget/list?limit=1000&offset=0 - List all widgets with
> pagination
> 2. GET /hicc/v2/widget/search/[query] - find widget based on title matching
> 3. GET /hicc/v2/widget/view/[title] - get a single widget
> 4. POST /hicc/v2/widget/create - create a new widget
> 5. PUT /hicc/v2/widget/update/[title] - update a widget
> 6. DELETE /hicc/v2/widget/delete/[title] - delete a widget
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
