Hacker News new | ask | show | jobs
by buahahaha 2922 days ago
Just retrain them.

They're going to have tangential skills to the new platform.

Get your network team to learn all about VPCs and networks and security groups and peering and VPNs, train them to be your in-house experts to deal with changes and expansions and to help trouble-shoot issues.

Chances are you're still going to have systems, your systems engineers can be repurposed to better tie in your monitoring tools, your security tools, etc. You're going to need authentication and identity management, and probably have internal people handling it.

If they're motivated to keep their jobs, all it should cost you is a $5-20K/head in training.

That's peanuts compared to the turmoil of everyone thinking they're going to lose their jobs, or the costs of hiring new people w/ pre-existing cloud skills. It is cheaper than the stop-gap of paying outside consultants till you can bring in the interior staff. Also, these people _know_ your existing systems. They know your business logic, they know what needs to be up and they know what can't break on Thanksgiving weekend. That is valuable knowledge that is a hidden cost in rehiring the work force.

You're going to find new challenges in the new system. If your employees are willing to transition, then transition them. It is your best bet.

If you have a subset of employees that can't/refuse to transition, put them on performance improvement plans. No rumor mill - if you're on a PiP you need to shape up or get out; if you're not on a PiP keep doing well, your job is safe.

3 comments

I can't agree more about retraining these guys. Hiring cloud engineers with good skills is extremely difficult
But is OP really going to have the same cost savings if they retain everyone? I am assuming labour is a huge part in that figure.
The ideal situation is to be in a growing company, and to move employees from shrinking areas into growing areas.

For example, if the company needs 8 fewer people to install and set up servers, but 10 more people to work on robust automated code deployments to support business growth, and both jobs need Unix skills.

Hence, the setting-up-servers department budget shows cost savings, the overall technology budget grows by 2 employees instead of by 10, and business growth means overall profit increases.

The cost saving from going cloud in a large part comes from utilization. 1 box that you can keep busy at 100% by clever workload timing is cheaper than 10 boxes that are idle 90% of the time that you just need to spike on. Or capacity that you can give back when you don’t need it. Rack space is expensive, power and cooling are expensive.

The job of the people is to translate the business requirements into stuff actually happening, and that doesn’t go away with cloud. If anything you need to do more of it (someone has to take care of auto-scaling, a thing that wasn’t possible in your legacy DC, for example). The worst thing you can do is sack the people then realize you need consultants at 5x the cost to do anything now...

This claim of massive savings is ludicrous for an f50 in the first place. It’s doubtful their cost is primarily maintaining and admining fleets of x64 servers, which is what the “cloud” is optimized for. These companies generally have massive costs from legacy mainframe, iseries, mvs etc systems. The ops creds are suspect to say the least.
Your comment would be more palatable if you gave the OP benefit of the doubt. It seems reasonable to assume that someone who is in the position to affect such change, and is willing to ask HN for advice, likely has actual cost savings estimates.

Perhaps saying something like, "Are you sure the cost savings are really there, and you won't face unexpected hidden costs in the future? I have experience with X,Y,Z, and we found that while the cloud is optimized for x64 deployment, we had hidden costs A, B, C, crop up after 9 months."

As it is, it's only added noise and made you look bad.

Your comment would be more palatable if you gave the parent poster the benefit of the doubt.

Perhaps using http://www.paulgraham.com/disagree.html and avoid dismissing the content because you don't like the way they said it (DH2, responding to tone).

He mentioned 90% of their datacenter operations. I assume the remaining 10% would be the assorted legacy every large company has of SPARC, PA-RISC, POWER, i and zSeries accretes over time.

Those will remain where they are and, eventually be ported or replaced (or kept alive - IBM mainframes have been legacy-friendly for over 50 years now) by newer systems as they are phased out.