A recent MIT Technology Review reveals how IBM’s strategy to compete with the likes of Amazon.com and Salesforce.com is to employ a tailored Cloud, designed to work for specific types of tasks, rather than offering generic storage and processing that can be used for whatever a customer need.

Read the rest of this entry »

One of the main concerns for companies when pondering whether to use Cloud technologies, is the potential vendor lock-in problem. After all, they are giving up the control of the infrastructure of their IT systems, which are essential for them. They must evaluate very carefully the chances that they are not able in the future to move those systems to other clouds, or even reinstall them back in the corporate infrastructure. Read the rest of this entry »

Although this blog does not usually cover low level features, today we will break this “rule” in order to illustrate the complicated features that one may be dealing with and that the Cloud itself is hiding for its users.

In order to avoid confusing the hardware TLB about the tasks ID (tasks belonging to different virtual machines (VMs) can actually have the same ID), flushing the TLB every time a VM request is dealt with by the hypervisor is a simple, yet highly inefficient way of solving this issue.

TLB tagging (i.e. adding a virtul machine-specific tag, called the address sspane identifier to the TLB) is a commonly employed technique. This way, each machine has a unique ID, known to the hypervisor and the TLB hardware only. This ID is invisible for the OS, thus eliminating the need to modify the guest and avoiding the performance degradation derived from flushing the TLB.

Although tagging the TLB allows a system to support several VMs running without flushing the TLB every now and then, it does not solve the mapping from that Guest physical memory to the actual physical memory plugged into the system (the host physical memory). Often, this is done with a technique called Shadow Page Tables. While the TLB translated virtual memory to Guest physical memory, the Shadow Page Tables translate Guest physical memory to actual host physical memor. This is a complex issue that is often dealt with froma hardware approach, i.e., building hadrware Virtualized Page Tables (translators from virtual memory to guest physical memory to host physical memory)  achieving early native performance and response times.

However, on the other side of the same coin, these hardware specific solutions often imply different approaches by different vendors, which increases the underlying heterogeneity and complicates the building of general software on top of the custom hardware.

Green IT is one of the top topics in the ICT area nowadays. It is related both with the need of companies to avoid expendings (power is an important part of datacenters’ bill), and with the increasing concern about environmental problems.

“The Road to Greener IT Pastures” (May 2009 issue of IEEE Computer Magazine) marks how, as the summer nears, the need for efficient cooling solutions in datacenter is even more pressing. This article is an overview of the most inmediate “green practiques” that can be/are applied by companies to reduce their energy consumption (specially from cooling needs):

• Applying metrics to measure network consumption and energy efficiency, like the ones defined by the Green  Grid and the SPEC consortium.
• Updating equipment to get more efficient, less power consuming hardware (from CPUs to disks arrays).
• Looking for management techniques to improve datacenters efficiency, and enforcing green policies in the organization (even automate them, applying remote control such as Wake-on-LAN…).
• Analyzing how to use virtualization techniques for servers consolidation. Despite the clear advantages of virtualization, its adoption it is not as straightforward as hype pretends. Organizations must carefully analize the (lack of) tools for the management of virtual machines and the skills they require, along with the impact of virtualization on the performance of applications.
• Thinking on the long term: using green techniques undoubtedly will pay off, but it must be done carefully and following a  green IT plan. A proper adoption strategy of green techniques will lessen the required initial investments on time and money, and will avoid increasing the complexity of the infrastructure management tasks.

Yet, we must remark these are only the most general action points for any organization to move to more energy efficient IT infrastructures. Each organization must evaluate their own needs and possibilities, and be ready to apply the new solutions for green IT management that will appear during the next years.

Some months ago we discussed here the implications of Open Source Software in a Cloud business model. Today we bring a new chapter of this unavoidable marriage.

We are in the days when even Microsoft admits that user’s expertise to improve software deserves its credits. The Open Source War seems to be almost won and, today, the future is envisioned by containing a mix of proprietary and open-source software.  A recent article in The Economist claims that this fact reduces the likelihood of vendor lock-in.

This very same article also claims that Cloud computing is a new potential trap to lead us into vendor lock-in, since customers risk losing control once again, in particular over their data, as they migrate into the cloud. Moving from one service provider to another could be even more difficult than switching between software packages in the old days. The authors cleverly highlight the need of moving “your MySpace profile to Facebook without manually retyping everything”.

Standardization efforts are a key issue to prevent this potential vendor lock in and Telefónica is actively playing a leading role by co-leading flagship projects such as RESERVOIR, contributing to the Open Cloud Manifesto, and standardization initiatives in international standardization bodies like OGF’s OCCI .

However, the Cloud is still evolving and the standardization process must keep pace with the advance of the technology itself. Time will tell where the Cloud finally gets and if any of the standardization efforts accomplishes the mission.