Al Hoang

Data Center Knowledge links to a story on data center rollouts in Vietnam:

Vietnamese technology company Quang Trung Software City (QTSC) and a U.S. company called DOT Vietnam have announced plans to build a $10 million data center in Ho Chi Minh City.

Some things that I’ve been wondering about with this data center are.

• Cooling and humidity (Vietnam is a tropical country after all)
• What companies will be using that data center once it is built.
• What will the network capacity outbound & inbound from this data center will be
• Price ranges for renting out a rack

I guess I’ll never know since this type of news will probably never get a follow-up article. Perhaps this is a good impetus to have a visit and see for myself what it will be like.

Just for contrast on budgets, M$ is plugging $500M USD to build a data center in Iowa.

Seems that Joyent experienced some downtime (8 days) due to some bugs in OpenSolaris that got triggered during an upgrade. (Read here and here)

One rather well known service was also affected by this outage as they are hosted by Joyent.

I guess this does not reflect very well on running OpenSolaris in a production environment. Especially, if downtime cannot be tolerated. I guess even with competent Solaris folk on board at Joyent this type of stuff can happen.

The stories point out that the bug that affected them was over a year old in the OpenSolaris bug database and seems to point some blame at Joyent for not reacting sooner to fixing the problem. However, the reality is that upgrading systems is never a button click like it is to upgrade some piece of software on most people’s Windows machines. In a non-trivial production system, there are usually multiple intertwining dependencies that make it difficult to change the system on the fly. Usually one tries their best to isolate the interdependencies of different pieces so that it IS possible to swap out pieces without affecting other pieces. But that is the ideal to work towards.

This is one reason why having windows of maintenance as well as trying to practice failure fire drills in some manner can help in trying to catch these issues before they happen. However, the trick is getting time to perform windows of maintenance and being able to duplicate enough of the system in a test environment to be able to run a fire drill. Sounds easy on paper, very difficult in practice.