AWS Project

Questions and steps for moving ITSB into cloud.

As of Jan-2015 all our production and development servers are co-located at InterNAP Seattle data center. Several of these machines, in particular our F5 Firewall/Load Balancers are nearly 15 years old and getting less reliable. Cost just for the hosting is running over $1K/month. Moving our web and database servers to the cloud is likely to cut this cost by half or more, especially considering the cost of replacing some of the ageing hardware.

-- Laurel Glidewell - 2014-10-20

ITSB In the Cloud

CloudMigrationForITSB

AWS Project Worksheet

https://docs.google.com/spreadsheets/d/13UwxEp7oMosR9Yp83k5XSyz3D7Dl4LCS85jqZGRf288/edit?usp=sharing

This spreadsheet has some figures for cost analysis; see below for details.

Project Dependencies

High Priority Projects to complete before migration (dev hours = 60)

(GOOGLE SHEET with ROADMAP ITEMS) https://docs.google.com/a/realtimeperformance.com/spreadsheets/d/1fr3NvgcenYH9_7LvSllD78Nv0_hszAElnylH0r1cHFE/edit?usp=sharing

  • Platform/technologies Upgrade:
    • tomcat,
    • log4j logging framework,
    • upgrade "struts" controller framework (Prod Roadmap #150)
      itsb_Struts2-MigrationPlanning.ods
    • change hibernate database pool to something more reliable (Prod Roadmap #151)
SW and HW Migration and Deployment Planning Items:
  • Design the deployment architecture/network & virtual machines for all our services on AWS
    • ITSB application servers (production, staging, dev, trial)
    • Database server(s)
    • Other services:
      • nagios
      • limesurvey (do we need this?)
      • apache + shibboleth (for SSO service provider)
      • itsb4me
      • corp site(s)
      • what else?
    • firewall, router design & plan: routing rules for all URL's we manage
  • Backup planning: applications, database, other application data
  • Reduancy/failover planning
  • Security planning and review
  • Database design/deployment changes (consolidate Prod01 & Prod02, what about staging, dev, trial?)
  • Shutdown & removal of the hardware at InterNap; disposition?
Network Migration Strategy + Planning
  • What services will migrate to AWS and when?
  • How to manage the "zero hour" rollover when we point DNS for "www.itskillbuilderassessment.com" to the new AWS site(s)?

Questions

  1. What are the inputs to a decision?
  2. What are our technical hurdles? see CloudMigrationForITSB for listing and discussion of the tech pre-work and planning
  3. How to address potential Customer push-back on cloud hosting. Boeing and Johnson&Johnson (at least) have expressed strong bias against using any application that runs in the cloud?
  4. When can we move off InterNap? What are cancellation consequences
  5. (Our Internap contract expires on 2/2016. I know we have to provide 60 days notice if we dont plan to renew. Im not sure what is involved if we try to get out of it beforehand. We may need Dans legal advisor to look into it. Below is a link to their standard agreement wording. http://www.internap.com/internap-master-services-agreement/
  6. Cost considerations for IBM-DB2 -- looks like we can continue to use our IBM WorkGroup server license on an AWS EC2 instance, only incurring the license cost of ~100 per year.

Cost Estimates

Current Status

We have one AWS server running Nagios production monitoring application.

Notes from 10/12 call Next steps: Scott: Build out development plan & timeline details and send to team. Basing on on est 50 dev hrs. Targeting a 11/15 deliverable date for upgrades/ updates. Stephen: draw up recommended deployment structure (servers, apps, etc) and proof with Scott before next team mtg. research traffic mgmt/ networking plan. use virtual f5 or EngineX, Apache, other... Dump export of all routing rules from f5 and get quote from Steve L to build in EngineX or other Research best location for Nagois outside network (Linode, Digital Ocean, other) finish Code repository update

NOTES Recommend starting with 2 servers setup (1 front end, 1 db). Snapshots and cloning for additional servers. Open questions: Do we have a complete inventory? We will start with Stephen drawing end solution and look at what's missing. How can we simplify? What are our redundancy requirements. Combine NIX and Hercules to run all non application? Run dev apps on Production server? AWS load balancer doesn't do rule-based traffic mgmt - more of a fail safe. Could use virtual F5 but expensive. Find another tool for proxy-ing. NgineX, Apache. Steve L has setup NgineX routing for RTP. Check if he wants to consult on this with us. Struts upgrade will touch a lot of code - impacts the ways the back end receives and acts on messages. Need regression testing added in release plan.

-- Laurel Glidewell - 2015-10-26

Next steps: Scott Code library updates - goal 11/15 Stephen Complete TomCat and DB2 server setup - goal 10/30 Code repository ready (BitBucket?) Start side server setup, networking setup - goal 11/30 Confirm dates with Steve L for completing NgineX rules. Research restore options from curret Prod to new single DB (table export?) and script out Laurel Client facing material/ contracts referencing hosting details Client communications (SSO clients and others) Setup mtg with Stephen to break down timeline for migration Notes: Moving towards October 30 goal of having TomCat server online and DB2 server. We will keep v9.7, instead of moving to 10.5. We may look at that upgrade later this year. 9.7 supported until April 2019 so no rush. NgineX will handle all traffic/ routing rules instead of using f5. Still use Elastic for SSO InterNap is providing quote for Cloud services. Unlikely we will use them unless they offer well over and above AWS - given prior history. Code Repository - BitBucket free up to 5 repositories. Scott and Stephen wil meet offline to discuss options for using this tool. Confirmed we are going with 1 DB 2 schemas. Currently we retore off PROD backup. For migration - we will likely use export of tables (like for Trial). Stephen will confirm approach, scrip and test. We will need to proof this several times before cut-over weekend. Plan to run Dev and Sandbox dbs on 1 front end server. Redundancy - Configure additional TC server but leave offline - for recovery or scale needs. Keep exact copy of config setup (hot standby) Client communication, contracts. this migration will be invisible to most clients (except SSO client). Laurel will work with Karen and Amy on updating any client facing materials that reference our hosting details. Add Prudential and possible C&J Energy SSO changes to migration plan/ schedule.

-- Laurel Glidewell - 2015-11-10

Topic revision: r11 - 2015-11-10 - LaurelGlidewell
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright © 2008-2024 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback