Project

General

Profile

Collection API - Performance enhancements » History » Version 8

Radhika Chippada, 05/12/2015 02:24 AM

1 1 Radhika Chippada
h1. Collection API - Performance enhancements
2
3
h2. Problem description
4
5
Currently, we are experiencing severe performance issues when working with large collections in Arvados. Below are a few scenario descriptions.
6
7
h3. 1. Fetching a large collection
8
9
Fetching a collection with large manifest text from API server results in timeout errors. This is suspected to be either the root cause or contributing largely to the other issues listed below. Several issues are reported which are the side effects of this issue: #4953, #4943,  #5614, #5901, #5902 
10
11
h3. 2. Collection#show in workbench
12
13
Often times, we see timeout errors in workbench when showing a collection page with large manifest text. It may be mostly due to the above listed concern about fetching the large collections. #5902, #5908
14
15
h3. 3. Create a collection by combining
16
17
Creating new collections by combining other collections or several files from a collection almost always fail when one of more of the involved collections contain large manifest texts. A few issues about this: #4943,  #5614
18
19
h2. Proposed solutions 
20
21
Various operations dealing with these large manifest texts are certainly the cause of these performance issues. Sending and receiving the manifest text to and from the api server to clients, json encoding and decoding of these large manifest texts could be contributing to this performance issues. Reducing the amount of data and the number of times this data is exchanged can greatly help.
22
23
h3. 1. Fetching a large collection
24
25
* Compress the data transferred (We recently enabled gzip compression between API and workbench)
26
27
* Use efficient json encoding / decoding
28
** We are using Oj between API server and Workbench. Is there room for further improvement? (http://devblog.agworld.com.au/post/42586025923/the-performance-of-to-json-in-rails-sucks-and)
29
** Are we consistently using Oj in Ruby SDK? (Radhika: I need to do further research to answer this question)
30
31
* Send the data in smaller chunks (?)
32
** Is it possible for us to implement some form of “paging” strategy in sending the manifest text to the clients from the API server?
33
34
h3. 2. Collection#show in workbench
35
36 6 Radhika Chippada
h4. Observations
37
38 2 Radhika Chippada
Collection#show responses are profiled using rack-mini-profiler. When pointed the development environment to qr1hi api server, the following observations are made (based on about 20+ reloads of the page):
39
40
* On average it took about 70s for to show the collection qr1hi-4zz18-tcnxylwkxg0nfhi
41
42
* The most expensive operations (on average) are:
43
** collections/_show_source_summary  -- 30 seconds
44
** collections/show (api request to get collection) -- 15 sec
45
*** It took on average .2 sec to parse response (json)
46
** collections/_show_files  -- 15 sec
47
** applications/_projects_tree_menu -- 3 to 4 sec
48
*** For this collection, 6 requests were made to /groups each taking .2 to .5sec
49
50 1 Radhika Chippada
* It is also observed that the requests took an average of 120 seconds on May 08, probably when the server was much busier and hence cluster tuning is also called for.
51
52 6 Radhika Chippada
* {{collapse(Performance profile snapshot ...)
53 1 Radhika Chippada
54 7 Radhika Chippada
!{width:80%}https://arvados.org/attachments/download/595/perf-profile-qr1hi-4zz18-tcnxylwkxg0nfhi.png!
55 5 Radhika Chippada
}}
56 1 Radhika Chippada
57 6 Radhika Chippada
* {{collapse(Workbench console log ...)
58 5 Radhika Chippada
59 2 Radhika Chippada
<pre>
60
Started GET "/collections/qr1hi-4zz18-tcnxylwkxg0nfhi" for 127.0.0.1 at 2015-05-11 14:47:03 -0400
61
Processing by CollectionsController#show as HTML
62
  Parameters: {"id"=>"qr1hi-4zz18-tcnxylwkxg0nfhi"}
63
API client: 0.0007654 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/users/current    
64
API client: 0.313339245 API transaction
65
API client: 0.000289537 Parse response
66
API client: 9.8434e-05 Prepare request https://qr1hi.arvadosapi.com/discovery/v1/apis/arvados/v1/rest    
67
API client: 0.250356943 API transaction
68
API client: 0.005898489 Parse response
69
API client: 0.000356541 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/collections/qr1hi-4zz18-tcnxylwkxg0nfhi    
70
API client: 21.405180053 API transaction
71
API client: 0.170310714 Parse response
72
API client: 0.000316374 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/jobs  {"output":"55152d2b989c6b174e298dba10ae3ff7+57708684"}  
73
API client: 0.221916151 API transaction
74
API client: 0.000178293 Parse response
75
API client: 0.000356427 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/jobs  {"log":"55152d2b989c6b174e298dba10ae3ff7+57708684"}  
76
API client: 0.078525257 API transaction
77
API client: 0.00017414 Parse response
78
API client: 0.000424393 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/links  {"head_uuid":"qr1hi-4zz18-tcnxylwkxg0nfhi","link_class":"name"}  modified_at DESC
79
API client: 0.059534807 API transaction
80
API client: 0.000152869 Parse response
81
API client: 0.000302943 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups  {"uuid":[]}  
82
API client: 0.06070466 API transaction
83
API client: 0.000145646 Parse response
84
API client: 0.00029954 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/links  {"head_uuid":"qr1hi-4zz18-tcnxylwkxg0nfhi","link_class":"permission","name":"can_read"}  modified_at DESC
85
API client: 0.062907368 API transaction
86
API client: 0.000149375 Parse response
87
API client: 0.00028907 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/logs  {"object_uuid":"qr1hi-4zz18-tcnxylwkxg0nfhi"}  created_at DESC
88
API client: 0.079528074 API transaction
89
API client: 0.00016842 Parse response
90
API client: 0.000523936 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/links  {"head_uuid":"qr1hi-4zz18-tcnxylwkxg0nfhi","tail_uuid":"qr1hi-tpzed-ktpvhqu89qoib9f","link_class":"resources","name":"wants"}  
91
API client: 0.062874638 API transaction
92
API client: 0.000175978 Parse response
93
API client: 0.000377932 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/api_client_authorizations   [["scopes","=",["GET /arvados/v1/collections/qr1hi-4zz18-tcnxylwkxg0nfhi","GET /arvados/v1/collections/qr1hi-4zz18-tcnxylwkxg0nfhi/","GET /arvados/v1/keep_services/accessible"]]] 
94
API client: 0.158867905 API transaction
95
  Rendered application/_show_autoselect_text.html.erb (0.9ms)
96
  Rendered application/_show_autoselect_text.html.erb (0.2ms)
97
  Rendered collections/_show_source_summary.html.erb (26534.7ms)
98
  Rendered collections/_sharing_button.html.erb (1.1ms)
99
API client: 0.000270093 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups/qr1hi-j7d0g-ixtfnxccvzm7ui4    
100
API client: 0.227171335 API transaction
101
API client: 0.000226508 Parse response
102
  Rendered application/_title_and_buttons.html.erb (239.8ms)
103
  Rendered collections/_show_files.html.erb (14531.8ms)
104
  Rendered application/_loading_modal.html.erb (1.5ms)
105
  Rendered application/_content.html.erb (14545.1ms)
106
  Rendered application/show.html.erb (14790.7ms)
107
  Rendered collections/show.html.erb within layouts/application (41352.7ms)
108
API client: 0.000296285 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/authorized_keys  {"authorized_user_uuid":"qr1hi-tpzed-ktpvhqu89qoib9f"}  
109
API client: 0.151415184 API transaction
110
API client: 0.000207343 Parse response
111
API client: 0.000238512 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/collections  {"created_by":"qr1hi-tpzed-ktpvhqu89qoib9f"}  
112
API client: 0.330716574 API transaction
113
API client: 0.000160901 Parse response
114
API client: 0.000376137 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/pipeline_instances  {"created_by":"qr1hi-tpzed-ktpvhqu89qoib9f"}  
115
API client: 0.263794223 API transaction
116
API client: 0.000841214 Parse response
117
API client: 0.000266079 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups   [["group_class","=","project"]] name
118
API client: 0.968549864 API transaction
119
API client: 0.001104326 Parse response
120
API client: 0.001218433 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups   [["group_class","=","project"]] name
121
API client: 1.214942211 API transaction
122
API client: 0.001293295 Parse response
123
API client: 0.001586803 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups   [["group_class","=","project"]] name
124
API client: 0.737031104 API transaction
125
API client: 0.000921665 Parse response
126
API client: 0.000487645 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups   [["group_class","=","project"]] name
127
API client: 0.848679769 API transaction
128
API client: 0.002339424 Parse response
129
API client: 0.000264704 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups   [["group_class","=","project"]] name
130
API client: 0.681490624 API transaction
131
API client: 0.000958536 Parse response
132
API client: 0.000484713 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups   [["group_class","=","project"]] name
133
API client: 0.412487956 API transaction
134
API client: 0.000900607 Parse response
135
  Rendered application/_projects_tree_menu.html.erb (5474.8ms)
136
API client: 0.00045925 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups/qr1hi-tpzed-rv6f0l8lbvdzkog    
137
API client: 0.078490258 API transaction
138
  Rendered application/_browser_unsupported.html (0.6ms)
139
  Rendered getting_started/_getting_started_popup.html.erb (2.7ms)
140 1 Radhika Chippada
  Rendered layouts/body.html.erb (6343.2ms)
141
Completed 200 OK in 70876ms (Views: 47804.5ms | ActiveRecord: 0.0ms)
142 2 Radhika Chippada
</pre>
143 1 Radhika Chippada
}}
144
145 6 Radhika Chippada
h4. Proposed enhancements
146 1 Radhika Chippada
147 6 Radhika Chippada
* API: Add files_count and files_size to collection data model
148
** Rather than computing it for each page display, we should consider adding these into the data model and update them when manifest_text changes
149
150
* Implement paging / scrolling in the collection#show page(?). Get “pages” of collection and display them as needed.
151
** This will address the next two big ticket items (the time taken in getting the collection json from API and _show_files
152
** This might also be inevitable for even larger collection than the one used in this profiling exercise
153
154 1 Radhika Chippada
* Avoid making multiple calls to the API server for the same data by caching or preloading data (See #5908)
155 6 Radhika Chippada
** Clicking on the Advanced tab resulted in making one more call to the API server to get the collection (which as seen above takes an average of 15 seconds or more)
156
** Cache the collection and other objects in workbench and avoid making unnecessary calls to the API server (while in the same page context)
157 1 Radhika Chippada
158
* Show less information in the collection page (such as not linking images that are going to 404)? (See #5908)
159 6 Radhika Chippada
160
* Add methods in the API server (?) to get my_projects and shared_project trees in one call and hence eliminating the average 3 seconds or so lag for "each" page display
161 1 Radhika Chippada
162
h3. 3. Create a collection by combining
163
164 8 Radhika Chippada
h4. Observations
165
Creating a new collection by combining collections is profiled by combining qr1hi-4zz18-ms5x87xf1389ldv, qr1hi-4zz18-0q225z4ktr432mg, qr1hi-4zz18-i5o4ba4mmxub69b from the project qr1hi-j7d0g-3d06b1jtiwrizqm (#4943)
166
167
* It took about 110 seconds to generate the combined manifest text, save new collection making an API server, get API server response for save
168
** The server sent the new collection, including manifest_text, after save
169
170
* It took an additional 70 seconds to "show" the new collection
171
** The workbench made yet another GET /collections/<uuid> request (about 20 seconds per the log), even though the server just sent it after saving
172
** All the other delays as listed in collection#show section above are part of this lag
173
174
175
* {{collapse(Performance profile snapshot ...)
176
177
!{width:80%}https://arvados.org/attachments/download/595/perf-profile-qr1hi-4zz18-tcnxylwkxg0nfhi.png!
178
}}
179
180
* {{collapse(Workbench log ...)
181
<pre>
182
Started POST "/combine_selected?action_data=%7B%22current_project_uuid%22%3A%22qr1hi-j7d0g-3d06b1jtiwrizqm%22%7D" for 127.0.0.1 at 2015-05-11 22:14:55 -0400
183
Processing by ActionsController#combine_selected_files_into_collection as HTML
184
  Parameters: {"authenticity_token"=>"U0PnyOLtNoz1Bet+ZUcC2+/1DTqN9Imzwhsk9VFGvro=", "selection"=>["qr1hi-4zz18-ms5x87xf1389ldv", "qr1hi-4zz18-0q225z4ktr432mg", "qr1hi-4zz18-i5o4ba4mmxub69b"], "action_data"=>"{\"current_project_uuid\":\"qr1hi-j7d0g-3d06b1jtiwrizqm\"}"}
185
API client: 0.000323633 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/users/current    
186
API client: 0.219897489 API transaction
187
API client: 0.000180886 Parse response
188
API client: 0.000744984 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/collections  {"uuid":["qr1hi-4zz18-ms5x87xf1389ldv","qr1hi-4zz18-0q225z4ktr432mg","qr1hi-4zz18-i5o4ba4mmxub69b"]}  
189
API client: 16.46763541 API transaction
190
API client: 0.174250085 Parse response
191
API client: 0.000405262 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/groups/qr1hi-j7d0g-3d06b1jtiwrizqm    
192
API client: 0.242986425 API transaction
193
API client: 0.000161982 Parse response
194
API client: 1.334179882 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/collections    
195
API client: 62.347996094 API transaction
196
API client: 0.129231689 Parse response
197
API client: 0.000315959 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/links    
198
API client: 0.247526392 API transaction
199
API client: 0.000215447 Parse response
200
API client: 0.000176321 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/links    
201
API client: 0.09771474 API transaction
202
API client: 0.00015503 Parse response
203
API client: 0.000181164 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/links    
204
API client: 0.088184468 API transaction
205
API client: 0.00018323 Parse response
206
Redirected to https://localhost:3031/collections/qr1hi-4zz18-cdwp2zwz1nj3uvh
207
Completed 302 Found in 109373ms (ActiveRecord: 0.0ms)
208
209
210
Started GET "/collections/qr1hi-4zz18-cdwp2zwz1nj3uvh" for 127.0.0.1 at 2015-05-11 22:16:44 -0400
211
Processing by CollectionsController#show as HTML
212
  Parameters: {"id"=>"qr1hi-4zz18-cdwp2zwz1nj3uvh"}
213
API client: 0.000233106 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/users/current    
214
API client: 0.069534939 API transaction
215
API client: 0.000270891 Parse response
216
API client: 0.00174422 Prepare request https://qr1hi.arvadosapi.com/arvados/v1/collections/qr1hi-4zz18-cdwp2zwz1nj3uvh    
217
API client: 17.157845279 API transaction
218
API client: 0.13215165 Parse response
219
...
220
</pre>
221
}}
222
223
224
225
226 1 Radhika Chippada
* Offer an API server method that accepts the selections array (and optionally owner_uuid and name) and performs the creation of the new collection in the backend. Doing so can help as follows:
227
** When combining entire collections: We can completely eliminate the need to fetch the manifest text for the collections in workbench. Also, workbench would no longer need to work through the combining logic and generate the manifest text for the new collection to be created. No need to do JSON decode and encode the manifest text. Lastly, it would not need to send this manifest text to the API server on the wire. Instead, the API server can do all these steps on the server and create the new collection and send the generated collection uuid to workbench (which will then reduce the performance issue down to collection#show issue; yay)
228
** When combining selected files from within a collection: Here also, we can see significant performance improvements by eliminating need to generate the combined manifest text and sending it on wire.
229
230
h3. 4. Implement caching using a framework such as Memcache
231
232
* One of the issues listed above (#5901) is around being able to access collection in multiple threads in parallel. Also, #5908 highlights several API requests being repeated within one page display. In fact, we have this issue in several areas of workbench implementation.
233
234
* By implementing caching, we will be able to reduce the need to make round trip API requests to fetch these objects. Instead, we can improve performance by fetching these objects from the shared cache. 
235
236
* Question: Not sure how caching would work if / when we cache these huge collections.