Building a Massively Scalable Cloud Service from the Grounds Up

Serving binaries for developers is not a trivial task. As opposed to software downloaded by end users, developer binaries are mostly consumed by software tools. As such, these binaries are exposed to massive request load, similar to a coordinated DDoS attack in the case of popular binaries. Add to that: the need to manage metadata; support for pricy REST queries; controlling storage quotas; collecting stats; calculating common repository indexes on demand; and across the globe distribution, and look, you’ve got yourself a pretty complicated system to run and manage. This talk will show you how Bintray, JFrog’s social binary distribution service, works, to allow any developer to serve and consume OSS binaries. We will speak about how the system is segmented to support massive loads across data centers with full redundancy and stateless vertical scaling; how Grails applications can scale and how we tie up different NoSQL technologies such as CouchDB, MongoDB, Elasticsearch and Redis; we’ll also see how Java technologies, such as Grizzly and Jersey can provide a lightweight alternatives to traditional web technologies; and demonstrate how no-interruption deployments are done to provide continuous distribution of binaries to developers worldwide.

Transcript

8. Some Numbers ___________ liftoff + 9 months
9. Some Numbers ___________ liftoff + 9 months Users 7K
10. Some Numbers ___________ liftoff + 9 months Users 7K Packages 90K
11. Some Numbers ___________ liftoff + 9 months Users 7K Packages 90K Requests 1.8B/Month
14. Requirements ___________
15. Requirements –Download binaries ___________
16. Requirements –Download binaries –Web application ___________
17. Requirements –Download binaries –Web application –REST API ___________
18. Requirements –Download binaries –Web application –REST API –Backend services ___________
26. Non-Func. Requirements _________________
27. Non-Func. Requirements Requirement RPS Availability _________________
28. Non-Func. Requirements Requirement RPS Availability Download 10K Always _________________
29. Non-Func. Requirements Requirement RPS Availability Download 10K Always Interaction 100 Almost always _________________
30. Non-Func. Requirements Requirement RPS Availability Download 10K Always Interaction 100 Almost always Services 1 Most of the time _________________
43. File A: 46b34 File B: a64ff7 /repo-z/package-y/file-x /repo-m/package-n/file-k /repo-w/package-t/file-f
44. File A: 46b34 File B: a64ff7
62. Requirements ___________
63. Requirements –Rapid Application Development ___________
64. Requirements –Rapid Application Development –Flexible schema ___________
65. Requirements –Rapid Application Development –Flexible schema –Java Back ground ___________
66. Requirements –Rapid Application Development –Flexible schema –Java Back ground –Stateless ___________
68. Why don’t you just use...? Framework Why not? ________________
69. Why don’t you just use...? Framework Why not? Angular.js Ember.js, ж.js Maturity ________________ -
70. Why don’t you just use...? Framework Why not? Angular.js Ember.js, ж.js Maturity Wicket State ________________ -
71. Why don’t you just use...? Framework Why not? Angular.js Ember.js, ж.js Maturity Wicket State JSF No comment ________________ -
72. Why don’t you just use...? Framework Why not? Angular.js Ember.js, ж.js Maturity Wicket State JSF No comment Non-java Not java bg ________________ -
88. Grails means Gorm!
99. Executive summary Framework Why not? ________________
100. Executive summary Framework Why not? Lucene/compass Only embedded, resource guzzler ________________
101. Executive summary Framework Why not? Lucene/compass Only embedded, resource guzzler solr Bad grail s integration ________________
102. Executive summary Framework Why not? Lucene/compass Only embedded, resource guzzler solr Bad grail s integration sphynx No incremental index ________________
129. SaaS for Download Server Component SaaS _________________
130. SaaS for Download Server Component SaaS blob storage SoftLayer datastore _________________
131. SaaS for Download Server Component SaaS blob storage SoftLayer datastore mapping Cloudant _________________
132. SaaS for Web and services Component SaaS _________________
133. SaaS for Web and services Component SaaS Model Mongohq _________________
134. SaaS for Web and services Component SaaS Model Mongohq Grail s N/A _________________
135. SaaS for Web and services Component SaaS Model Mongohq Grail s N/A ElasticSearch N/A _________________
136. SaaS for Web and services Component SaaS Model Mongohq Grail s N/A ElasticSearch N/A Redis N/A _________________
142. Virtualization __________
143. Virtualization __________ Pros
144. Virtualization __________ Pros – Cheap
145. Virtualization __________ Pros – Cheap – elasticity
146. Virtualization __________ Pros – Cheap – elasticity – Volatility
147. Virtualization __________ Pros – Cheap – elasticity – Volatility cons
148. Virtualization __________ Pros – Cheap – elasticity – Volatility cons – Overhead
149. Virtualization __________ Pros – Cheap – elasticity – Volatility cons – Overhead – Tenant, not owner
185. 1. Vagrant boots centos on virtualbox
186. 1. Vagrant boots centos on virtualbox 2. Chef install s all db and service rpms from private YUM repo
187. 1. Vagrant boots centos on virtualbox 2. Chef install s all db and service rpms from private YUM repo 3. Profit!
218. Conclusions time ______________
219. Conclusions time –Define Criticality ______________
220. Conclusions time –Define Criticality –Embrace the change ______________
221. Conclusions time –Define Criticality –Embrace the change –Plan for scale, but be real ______________
222. Conclusions time –Define Criticality –Embrace the change –Plan for scale, but be real –Backup everything! ______________