Outages Bring Cloud Monitoring in Focus!

April 23, 2011

The outage of Amazon's cloud service impacted many businesses hosted in their cloud data center

Over the last week, Amazon’s cloud service had a serious outage that caused many popular web businesses to go offline for several hours and resulted in significant loss of business.

All of a sudden, many in the press (and users as well!) are beginning to realize that applications hosted in the cloud are actually hosted on servers in data centers and are hence, prone to same kind of problems as servers in their own data center. Just because you have not purchased a server or have to provision it, provide power/space, etc., does not mean that the server is failure-proof. As this article indicates, failures can happen due to any number of reasons – a hardware failure, a network outage, an application coding error, etc. Even a configuration error inadvertently made by an administrator can cause catastrophic failures.

Many have gone over-board, predicting the end of cloud computing! If you look at the service contract from these cloud service providers, they have not guaranteed that the infrastructure will be 100% failure-proof. With cloud computing as with everything else, you get what you pay for. Not every business that used Amazon suffered during this outage. The outage was limited to the Amazon east coast (Northern Virginia) data center and for enterprises that had paid for Amazon Web Services’ redundant cloud architecture, it was business as usual. Netflix, the popular movie rental site, was one such.

"You get what you pay for" applies to monitoring tools as well!

Outages like the one Amazon had bring cloud monitoring tools into focus. The saying “You get what you pay for” applies to monitoring tools as well. If you are looking to be alerted once a problem happens, a simple low-cost up/down monitoring tool suffices. On the other hand, if you are looking to be like Netflix and be proactive, want to detect problems before they become revenue impacting, you need a monitoring tool that can alert you to abnormal situations in advance, well before users notice the problem. More sophisticated cloud monitoring tools can also help you rapidly triage where the root-cause of a problem lies – i.e., is it in the cloud data center? is it in your application? is it in the infrastructure services (DNS, Active Directory, etc.)?

Monitoring tools provide insurance cover for your infrastructure. Like Netflix would have assessed the cost of redundancy vs. the benefit from having their business up during the outage, you should assess the return on investment from a monitoring tool.

There are several ways to assess the ROI from a monitoring tool::

  1. By the number of times the monitoring tool can help you avert a problem by proactively alerting you and enabling you to take action before users notice the issue;
  2. By the time the monitoring tool saves by helping you to pinpoint the root-cause of a problem;
  3. By the amount of time that the monitoring tool saves for your key IT personnel by allowing your first level support teams to handle user complaints;
  4. By the savings that the tool provides by enabling you to optimize your infrastructure and to get more out of your existing investment, without having to buy additional hardware or to use additional cloud services.

Related links:

The top requirements for a Cloud-Ready Monitoring SolutionClick here >>>

Agility – Not Cost – is Driving Private Cloud Computing

March 13, 2011

According to Gartner, agility (not cost) is the most important driving factor motivating private cloud computing deployments. If you have used a public cloud service provider and experienced a cloud-based service first hand, this is not surprising. Having the ability to get a server you need in minutes with the applications you need pre-installed is a huge change from the traditional way of ordering a server and waiting for days to have the server OS and applications to be installed.

The term “cloud computing” has been one of the most over used terms in the last year. To understand what cloud computing can deliver, let me take a simple analogy. Remember the days when you had to physically visit a bank to get to know your bank balance or to transfer money between accounts? Internet and mobile banking has made it significantly easier for us to do the same today. The result – great convenience for you the consumer. At the same time, it also introduces operational efficiency for the bank – the bank can have fewer tellers. Cloud computing with its self-service interface promises to do the same. Users will benefit from the increased agility and the convenience to create the servers they need. The IT operations team will also benefit from having to deal with fewer routine enquiries.

The challenge with private clouds is getting the IT team's buy-in.

The challenge though is that not many IT operations teams may believe so! According to Gartner, early adopters of private cloud computing indicated that technology is one of the least important challenges in building private cloud solutions. Much tougher issues are the cultural, process, business model and political issues! This is not surprising because IT administrators are likely to believe that role is being undermined by allowing users to self provision what they need. Gartner’s recommendation is therefore to focus first on the fundamental concept of turning IT into a service provider that delivers through a service catalog, leveraging automation and probably being paid by usage.

Management Technologies will Play a Central Role in Fulfilling the Promise of Cloud Computing and Virtualization Technologies

December 18, 2010

2011 is almost here and it promises to be an exciting and challenging year!  Here are my top 10 predictions in the monitoring and management space for 2011.

Virtualization and cloud computing have garnered a lot of attention recently. While virtualization has been successfully used for server applications, its usage for desktops is still in its early stages. Cloud computing is being tested for different enterprise applications, but has yet to gain complete acceptance in the enterprise. 2011 will be the year that these technologies become mainstream.

A key factor determining the success of these technologies will be the total cost of ownership (TCO). The lower the TCO, the greater the chance of adoption. By proactively alerting administrators to problems, pointing to bottleneck areas and suggesting means of optimizing the infrastructure, management technologies will play a central role in ensuring that these technologies are successful. With this in mind, I make the following predictions for 2011:

1. Virtualization will go mainstream in production environments. Very few organizations will not have at least one virtualized server hosting VMs. Enterprises will focus on getting the maximum out of their existing investments and will look to increase the VM density – i.e., the number of VMs for each physical server. In order to do so, administrators will need to understand the workload on each VM and which workloads are complementary (e.g., memory intensive vs. CPU intensive), so IT can use a mix and match of VMs with different workloads to maximize usage of the physical servers. Management tools will provide the metrics that will form the basis for such optimizations.

2. Multiple virtualization platforms in an organization will become a reality. Over the last year, different vendors have come up with virtualization platforms that offer lower cost alternatives to the market leader, VMware. Expect enterprises to use a mix of virtualization technologies; the most critical applications being hosted on virtualization platforms with the best reliability and scalability, while less critical applications may be hosted on lower-cost platforms. Enterprises will look for management tools that can support all of these virtualization platforms from a single console.

3. Enterprises will realize that they cannot effectively manage virtual environments as silos. As key applications move to virtual infrastructures, enterprises will realize that mis-configuration or problems in the virtual infrastructure can also affect the performance of business services running throughout the infrastructure. Further, because virtual machines share the common resources of the physical server, a single malfunctioning virtual machine (or application) can impact the performance seen by all the other virtual machines (and the applications running on them). If virtualization is managed as an independent silo, enterprise service desks will have no visibility into issues in the virtual infrastructure and, as a result, could end up spending endless hours troubleshooting a problem that was caused at the virtualization tier. Enterprise service desks will need management systems that can correlate the performance of business services with that of the virtual infrastructure and help them quickly translate a service performance problem into an actionable event at the operational layer.

4. Virtual desktop deployments will finally happen. VDI deployments in 2010 have mostly been proof of concepts; relatively few large scale production VDI deployments of VDI have occurred. Many VDI deployments run into performance problems, so IT ends up throwing more hardware at the problem, which in turn makes the entire project prohibitively expensive. Lack of visibility into VDI is also a result of organizations trying to use the same tools they have used for server virtualization management for VDI. In 2011, enterprises will realize that desktop virtualization is very different from server virtualization, and that management tools for VDI need to be tailored to the unique challenges that a virtual desktop infrastructure poses. Having the right management solution in place will also provide VDI administrators visibility into every tier of the infrastructure, thereby allowing them to determine why a performance slowdown is happening and how they can re-engineer the infrastructure for optimal performance.

5. Traditional server-based computing will get more attention as organizations realize that VDI has specific use cases and will not be a fit for others. For some time now, enterprise architects have been advocating the use of virtual desktops for almost every remote access requirement. As they focus on cost implications of VDI, enterprise architects will begin to evaluate which requirements really need the flexibility and security advantages that VDI offers over traditional server-based computing. As a result, we expect server-based computing deployments to have a resurgence. For managing these diverse remote access technologies, enterprises will look for solutions that can handle both VDI and server-based computing environments equally well and offer consistent metrics and reporting across these different environments.

6. Cloud computing will gain momentum. Agility will be a key reason why enterprises will look at cloud technologies. With cloud computing, enterprise users will have access to systems on-demand, rather than have to wait for weeks or months for enterprise IT teams to procure, install and deliver the systems. Initially, as with virtualization, less critical applications including testing, training and other scratch-and-build environments will move to the public cloud. Internal IT teams in enterprises will continue work on public clouds and ultimately a hybrid cloud model will evolve in the enterprise. Monitoring and management technologies will need to evolve to manage business services that span one or more cloud providers, where the service owner will not have complete visibility into the cloud infrastructure that their service is using.

7. Enterprises will move towards greater automation. For all the talk about automation, very few production environments make extensive use of this powerful functionality. For cloud providers, automation will be a must as they seek to make their environments agile. Dynamic provisioning, automated load balancing and on-demand power on/power off of VMs based on user workloads will all start to happen in the data center.

8. Do more with less will continue to be the paradigm driving IT operations. Administrators will look for tools that can save them at least a few hours of toil each day through proactive monitoring, accurate root-cause diagnosis and pinpointing of bottleneck areas. Cost will be an important criterion for tool selection and, as hardware becomes cheaper, management tool vendors will be forced away from pricing per CPU, core, socket or per application managed.

9. Enterprises will continue to look to consolidate monitoring tools. Enterprises have already begun to realize that having specialized tools for each and every need is wasteful spending and actually disruptive. Every new tool and introduced carries a cost and adds requirements for operator training, tool certification, validation, etc. In 2011, we expect enterprises to look for multi-faceted tools that can cover needs in multiple areas. Tools that can span the physical and virtual worlds, that can offer both active and passive monitoring capabilities, support both performance and configuration management will be in high demand. Consolidation of monitoring tools will result in tangible operational savings and actually work better than a larger number of dedicated element managers.

10. ROI will be the driver for any IT initiative. In the monitoring space, tools will be measured not by the number of metrics they collect but by how well they help solve real-world problems. IT staff will look for solutions that excel at proactively monitoring and issuing alerts in advance of before a problem happens, and how they can help customers be more productive and efficient (e.g., by reducing the time an expert has to spend on a trouble call).

Reposted from VMBlog – http://vmblog.com/archive/2010/12/09/eg-innovations-management-technologies-will-play-a-central-role-in-fulfilling-the-promise-of-cloud-computing-and-virtualization-technologies.aspx


Get every new post delivered to your Inbox.

Join 327 other followers

%d bloggers like this: