Massive Migration Project

5 posts / 0 new
Last post
puravida's picture
puravida
Jedi Warrior
Offline
Joined: 09/01/2007
Visit puravida's Website

Due to extreme growth and use of our service, we knew that a migration to a more robust and scalable solution would be necessary. We knew that solution, however, would be quite expensive to implement. So we have been saving as much as possible and we have been investing and working in the background for the past couple of YEARS to get to this point.

So far, we have invested approximately $150,000 in research, development, and services to push forward with this major migration. We expect the final cost to approach $250,000 by the time we're done. Those numbers REALLY scared us a few years back (since we never took any venture capital), but to our pleasant surprise, our loyal customers supported us to the point of being able to afford this evolution for the service. So, a BIG thank you to all of you!

There are three(3) phases to this project:

  1. Capture Generator Migration (Completed: Details)
  2. Storage Array Migration
  3. Primary Server Cluster Migration

Capture Generator Migration
The first phase was already completed as of 08/02/2012. This was a huge milestone that has exceeded our expectations in performance and automation. This migration came along with an overhaul to the generator technology, and that allowed us to address quite a few lingering issues with captures. Our capture failure rate has been reduced to less than 0.0001% errors.

Storage Array Migration
The second phase will significantly reduce operational costs and will provide the flexibility to automatically scale growth without further intervention. The savings will offset the added expenses of automation required for the first phase. The new storage platform is comparable in performance but is unsurpassed in scalability and reliability. This migration also paves the way to eliminating the main retention policy, which means that -once we can afford it- we can begin to build a massive pre-captured cache of website screenshots. Eventually, this will allow for the closest to "real-time" that is possible, since we will strive to capture all pages on the Internet (quite a lofty goal, eh!?).

Primary Server Cluster Migration
The third phase will further reduce costs and allow for nearly infinite scalability for capture generators and delivery of screenshots. This is the final step required in order to support the world's growing need for fully automated web page screenshots, including heavy users such as Google, Yahoo (Ask & Axis), Alexa, Bing, et cetera; should they ever decide to outsource for more features, better support, and reduced operational expenses.

PLANNED DOWNTIME

Phase I - Capture Generator Migration
We had zero downtime for phase I, but we did experience extended capture delays for about 24 hours, due to a roll-out glitch. Ultimately, that's not too bad, considering the scope of the project phase and also that some competitors take 24 hours to generate a capture under normal circumstances.

Phase II - Storage Array Migration
We do not expect any service downtime, but we will definitely experience capture and delivery delays. The reason for the delays is that we will be wiping out our entire cache and basically starting over. The first reason for this action is to clean up lingering issues and correct those for much more efficient delivery. The second reason is to avoid a 60-90 day migration of the current cache, which would increase costs by another $15,000 or so. That savings will make a huge difference, since we are quickly draining our savings to invest in these improvements.

Phase III - Primary Server Cluster Migration
We expect a minimal amount of downtime, which will be announced beforehand, due to a required IP address change. Typically, propagation could cause up to 24 hours of downtime, but we will keep both, old and new, services operational during the transition. That "should" avoid any downtime, but in the case that downtime is unavoidable, this should keep it to less than an hour that will occur during off-peak times.

END RESULT
Now, all of these growing pains and huge expenses wouldn't make any sense unless we all got something out of it, right? So, here's why we have been working so diligently:

  • Evolve the company to handle the entire world's screenshot needs
  • Provide maximum scalability to keep up as screenshot demands grow
  • Fully automate the entire process so that we can focus on improvements and sales
  • Lower operational costs so that we can afford to scale and to reduce prices

We expect to complete this project, all phases, by the end of this year. After a few months of analysis on the new operational expenses and maintenance and everything involved; we will be able to determine whether or not we can lower costs at that time.

We are hopeful that we can reduce the price of new requests, bandwidth, and dedicated generators (once officially re-launched for Enterprise accounts). Any reduction in rates will be applied retroactively for all users, so it will severely reduce our income. We are OK with that, as long as we are certain that the reduced revenue will still cover all expenses. Currently, our revenue does not cover salaries that we need to really grow the company, but we are banking on lower prices and extreme growth to help boost revenue enough to bring us on full-time. That will be good for everyone, because we will finally have the time to focus on upgrades, support, new PRO features, and custom requests.

The pricing reductions will also coincide with the launch of "Enterprise" level accounts. These accounts will offer quite a few new features, such as multiple "Access Key ID" tracking, more reporting, secondary test accounts, and access to dedicated generators. There will be a sizable monthly upgrade cost associated with the "Enterprise" level accounts and will be mandatory for customers over a certain usage level. However, the pricing of the "Enterprise" level account upgrade will be offset by reduced prices, still resulting in some savings.

We greatly appreciate everyone's support thus far! Thank you!

DelCKS
Offline
Joined: 02/25/2011
Visit DelCKS's Website

Heya PV,

Great to hear the good news.

Do hope there will be a couple of other (cheaper) upgrades for the mini-users. I only use < 200 images (4% of the monthly limit) and have to pay the same price as those who max their 5000 requests limit.

It would be nice if I could pay less...

D.

puravida's picture
puravida
Jedi Warrior
Offline
Joined: 09/01/2007
Visit puravida's Website

Hello,

While we are not currently considering to add any new levels (which always has potential to confuse users), we are working diligently to lower costs for bandwidth, new requests, and dedicated generators. However, it is requiring a significant investment in time and money ($250,000 approximately).

We expect to be able to announce a major pricing reduction (will apply retroactively) by mid-2013.

puravida's picture
puravida
Jedi Warrior
Offline
Joined: 09/01/2007
Visit puravida's Website

We are entering the final stages of the migration project.

So now, we are in DNS duplication mode and will be migrating DNS to a geo-distributed DNS platform. There should be no interruption in service (unless we make a typo!), so there should be no customer impact. However, we are announcing it because things can go wrong.

We are being very careful and will make a follow-up when the DNS migration has completed.

Since there is little potential impact from this change, it is not a separate post that will email everyone on the list. However, the final step will be announced, ahead of time, just in case there is end-user impact.

*****
The next phase is final testing and then to make a DNS zone records update. That will result in 1-5 minutes of downtime, or more if something is not working quite right and forces us to revert to the old system. The downtime could be doubled plus the time it takes us to discover the issue. However, as always, we plan to do a very thorough set of tests before we make the switchover. We will keep everyone on the maintenance notification list informed.

puravida's picture
puravida
Jedi Warrior
Offline
Joined: 09/01/2007
Visit puravida's Website

Now, I'm being told that we should plan for 30-60 minutes of actual, complete downtime. The reason is that we have to shut down the system for new requests (cached images may still deliver, but that is subject to change) while we migrate to the new servers and deploy into the new storage array. Then, there will need to be some final, LIVE testing before we turn it loose.

Therefore, we will most likely announce a 2-hour window of downtime for maintenance and migration. We are still not sure if we can be ready for this weekend or the next, but we will make an announcement as soon as we know.

ShrinkTheWeb® (About STW) is another innovation by Neosys Consulting
Contact Us | PagePix Benefits | Learn More | STW Forums | Our Partners | Privacy Policy | Terms of Use

Announcing Javvy, the best crypto exchange and wallet solution (coming soon!)

©2018 ShrinkTheWeb. All rights reserved. ShrinkTheWeb is a registered trademark of ShrinkTheWeb.