Cloudian at cassandra conference in tokyo

  • 3,135 views
Uploaded on

Presentation at Cassandra Conference in Tokyo on October 5, 2011

Presentation at Cassandra Conference in Tokyo on October 5, 2011

More in: Technology
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
3,135
On Slideshare
0
From Embeds
0
Number of Embeds
1

Actions

Shares
Downloads
0
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. Cassandra Conference in Tokyo Cloudian: Storage Layer October 5, 2011 Gary Ogasawara Gemini Mobile Technologies (c) Copyright and Confidential, Gemini Mobile2011/10/5 1 Technologies, Inc. & KK, 2011, All rights reserved.
  • 2. Outline1. What is Cloudian™2. Cassandra in Cloudian3. Cassandra for Object Store4. Cassandra for Reports5. Cassandra for Account Info6. HyperStore™7. HyperStore Performance(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 2
  • 3. Cloudian = S3 Cloud Storage as Packaged Software S3 = Simple Storage Service (S3) by Amazon Web Services (AWS)(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 3
  • 4. Cloudian Features1. Full AWS S3 API Compatibility, including error codes2. Multi-datacenter, peer-to-peer architecture. No single point of failure.3. Multi-tenant: QoS controls, billing, reporting by each User and each Group4. Public and Private Clouds. Elastic Capacity: small start and scale-out as needed5. System, Group, and User management by Management Console or REST API6. Easy to Use Packaged Software, backed by 24x7 carrier grade support team with 10 years experience in Japan(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 4
  • 5. Cloudian customers Launched beta service on August 1, 2011(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved.
  • 6. Cloudian customers Launched “Nifty Cloud Storage” service from September 29, 2011(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved.
  • 7. Cloudian customersIn 2 months since Cloudian commercial release in July 2011 Over 30 companies in the world have been evaluating Cloudian as trial(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved.
  • 8. Cloudian Use Cases1. Public Cloud Storage Service 2. Enterprise Private Cloud/NASUse Case: S3 API compatible service Use Case: Secure, Low cost, Reliable, Easy to Scale Enterprise StorageBy: Data Centers By: Appliance Vendors Cloud Service Providers System Integrators Telecom Carriers Data CentersSize: Peta-Byte to Exa-Byte Cloud Service Providers 10s to 100s of nodes Telecom CarriersTarget: File Share, Archive, Backup Size: Tera-Byte to Peta-Byte Social Games, SNS 3 to 10s of nodes Multi-device Applications Target: NAS/SAN replacement Media Backup, Streaming NAS/SAN Backing Store Enterprise NAS backup Any size Enterprise … … many others … … Government, Medical, Financials … … many others … … (c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 8
  • 9. Cassandra in CloudianWhy Cassandra? Scalable. - Scale out by adding nodes to running system. - Distributed (P2P architecture), no single point of failure Reliable. - Resilient to network or hardware failures. - Multi-datacenter replication - Tuneable data consistency level. Fast. - Write path especially fast.(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 9
  • 10. Cloudian™: Logical Architecture Admin Credentials Server DB HTTPS Login HTTP QoSAccount profile HTTPS Servlets S3 Server/ Security keys Servlets DB Reports HTTP Management ConsoleData Explorer HTTP Data UserData DB Servers (Cassandra) WEB UI HTTP or AccountInfo HTTPS DB (S3) (Cassandra) Reports DB Applications (Cassandra) (c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 10
  • 11. Network Design: Minimum Redundant Configuration Servlets Cred/QOS DBBrowser HTTP/HTTPS HTTP/S Sticky Cassandrarequests sessions Serverfor UI LBApplication HTTP/HTTPSrequestsfor S3 Servlets Cred/QOS DB HTTP/S Cassandra Server (c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 11
  • 12. Network Design: Multi-Datacenter Example 2 datacenters / 4 nodes per datacenterServlets Servlets Servlets ServletsHTTP/S HTTP/S HTTP/S HTTP/S Cassandra Cassandra Cassandra CassandraServer Server Server Server Cred/QOS Cred/QOS Cred/QOS Cred/QOS DB (M) DB (S) DB (M) DB (S)HTTP/S HTTP/S HTTP/S HTTP/S Cassandra Cassandra Cassandra CassandraServer Server Server Server DC1 DC2 Users routed to their “home” DC using S3 bucketname. “Home” DC holds primary copy of user data. Storage objects, Reports, User profiles replicated across DCs by Cassandra. Credentials/QOS DB (KV DB) has local DC slave and remote DC slave. (c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 12
  • 13. Cassandra for Object Store Blob storage Object metadata Large object support - Multi-part. Amazon S3 multi-part API. Parts uploaded independently and in any order. Single parts can be re-transmitted. After all parts are uploaded, then presented as a single object. - Chunking. Break single objects into smaller chunks (e.g., 10 MB) when storing. - HTTP Range header. For downloads, HEAD request used to get object size, then download using start byte and length. - HyperStore™. Store objects on multiple storage options.(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 13
  • 14. Cassandra for Account InfoDATA MODEL User - ID, name, contact info, etc. Group - ID, name, contact info, etc. Rating Plan Security CredentialsNOTES “Static” data. Fixed number of columns. Secondary index in User CF on groupID. Allows query to get all userIDs for a specified groupID. Could be put in a Relational DB like MySQL, but no need to add another component.(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 14
  • 15. Cassandra for ReportsDATA MODEL “Raw” column family - User, Group, System - Transaction type (HTTP GET, PUT, DELETE) - Object path - Size “Rollup” column families. - RollupHour. Summarizes data for each hour using Raw data. - RollupDay. Summarizes data for each day using RollupHour data. - RollupMonth. Summarizes data for each month using RollupDay data.NOTES High write rate. Low read rate. Automatic deletion using Cassandra TTL (time-to-live).(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 15
  • 16. HyperStore™ Admin CredentialsHyperStore: Management policies tailored for different object types. S3 QoS Object metadata is still stored in Cassandra Data Store Use Cassandra’s distributed systems (Alternate) methods for data partitioning, replication, HyperStore™ node health detection. Manager Data Store Fork Cassandra source for customizations. (Cassandra) Cloudian™Benefits: Server Accounting (Cassandra) Better performance More capacity per node Reporting (Cassandra) Higher disk utilization Storage layer flexibility (c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 16
  • 17. HyperStore: Hybrid Storage Example Storage 1 Storage 2 optimal U X Optimal solution is to choose the storage method that minimizes latency. Generally, you want to maximize/minimize U, a performance metric, based on random variables X using a mixture of N storage layers. In a simple case, U : average latency X = {object size} N = {cassandra, ext4 fs}.(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 17
  • 18. HyperStore: Initial Results on Latency 50 40ms 30 >30% faster PUT-Cass 20 PUT-HS 10 0 KB 0.5 5 50 500 60 50 40 >400% fasterms 30 GET-Cass 20 GET-HS 10 0 KB 0.5 5 50 500 (c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 18
  • 19. HyperStore: Next Steps Disk space usage: Improve disk utilization percentage by HyperStore. Failure cases: Need extensive testing in cluster environment. Integrated operations: Interaction between nodetool and “hsstool”. Seeking alpha testers Contact bigdata@geminimobile.com Twitter: @geminimobile(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 19
  • 20. Request Free Cloudian Trial Japanese: http://geminimobile.jp/big-data-solutions/cloudian_eval_agr.php5 English: http://geminimobile.com/cloud-solutions/cloudian_eval_agr.php(c) Copyright and Confidential, Gemini Mobile Technologies, Inc. & KK, 2011, All rights reserved. 20