Tag: cloud

Datacenter Energy

Bapat is also right to point to electricity as a weak link in the system. The centralization of computing can bring dramatic increases in overall energy efficiency, compared to the fragmented, subscale private data centers we have today. The megacenters tend to operate at much higher levels of capacity utilization than the private centers can achieve, and their operators have the skill and wherewithal to install cutting-edge power-management and cooling technologies. At the same time, however, the centralization of computing assets concentrates energy demand, putting new strains on the aging electric grid. Efficiency increases, but vulnerability does, too.

politicians will soon be forced to wake up to the importance of data centers as national assets. maybe data centers can lead the necessary changes to the energy grid?

Kernel Debugging at scale

This paper will discuss the difficulties and methods involved in debugging the Linux kernel on huge clusters. Intermittent errors that occur once every few years are hard to debug and become a real problem when running across 1000s of machines simultaneously. The more we scale clusters, the more reliability becomes critical. Many of the normal debugging luxuries like a serial console or physical access are unavailable. Instead, we need a new strategy for addressing thorny intermittent race conditions. This paper presents the case for a new set of tools that are critical to solve these problems and also very useful in a broader context. It then presents the design for one such tool created from a hybrid of a Google internal tool and the open source LTTng project. Real world case studies are included.

how to deal with rare error conditions that are hard to reproduce

Trailer Park Showdown

Sun, which has its Blackbox containerized data center out on tour, is suddenly facing some tough competition in the burgeoning trailer park computing market. Rackable Systems is rolling out a copycat product called Concentro that may just outdo the original. Besides sporting a most excellent name – if Flash Gordon had a computer, it would be called Concentro – Rackable’s portable data center comes in a 12m shipping container, making Sun’s 6m model look downright wimpy, and it can be packed with 9600 processing cores or 3.5 petabytes of storage. Best of all is the interior. Check it out:

another “cluster in a container”. see also, microsoft is stealing a page from the sun playbook.

Cloud Operations

Her belief is that there’s going to be a tipping point in Web 2.0 where the operational environment will be a key differentiator. I mentioned the idea that Web 2.0 has been summed up as “Fail Fast, Scale Fast,” and she completely agreed. When it hit its growth inflection point, MySpace was adding 1m users every 4 days — not at all an easy feat. As these massive apps become the norm, unless you can play in a game where services can be highly stable, geodistributed, etc., you won’t be in the game. And that’s where she came to the idea that being a developer “on someone’s platform” may ultimately mean running your app in their data center. Why did Fedex win in package delivery? They locked up the best locations with access to airports, warehousing, etc. so they had the best network. A similar thing will happen with packet delivery.

where ms gloats about their datacenters

Terracotta

The Terracotta Server provides powerful distributed in-memory data management capabilities for Terracotta products (such as Ehcache) and is the backbone for Terracotta clusters. A Terracotta Server Array can vary from a basic 2-node tandem to a multi-node array (Terracotta Server Array (TSA)) providing configurable scale, high performance, and deep failover coverage.

low level java clustering. i love that more and more infrastructure projects slap together a quick video to explain where they fit in

AWS savings

Depending on my comfort level with EC2’s reliability, we may not even need a failover server as we have now. But just assuming we were comparing apples to apples and we assumed that Mass Events Labs needs 2 servers, the total annual cost is $1752. Today, our annual cost for 2 servers is $8400.

about time, too. there are far too many small, incompetent hosters out there