Voice of the DBA

Category: Editorial

  • Making New Infrastructure Investments

    server room build
    Will we do away with much of this data center build out with cloud services available?

    I was at a cloud conference recently, learning about the different ways in which people are starting to use cloud services. To be clear, I know that the cloud is nothing new, but it does have the potential to change the way we make decisions about infrastructure, and it absolutely will change the way we invest in IT if we choose to use cloud services.

    I had the chance to talk with three companies that were using the cloud. One was a startup that had foregone all investment in IT infrastructure. They had laptops for employees and one local server to keep copies of development code, but the rest of their “IT stuff” was in the cloud. As a new company, with no existing hardware, this makes sense. Their ability to offload all the sysadmin work meant that their developers didn’t have to worry about IT services; they just wrote code.

    The second company was a small company, one with some servers, but a growing company. They decided to move their email and document management to the cloud (in Office 365) rather than invest in bigger servers to handle the load. They maintained a local Sharepoint server, a development server, SQL databases, etc., but their decision to use cloud services was based on the desire to limit their future investment in computer hardware, and potentially additional staff.

    The last company surprised me. They were an established, medium sized company, and they used a few services in the cloud, like Salesforce, but they also moved some of their customer applications to the cloud. Why? They had a fairly unpredictable spike in traffic multiple times a month. They could have invested in hardware to handle the spikes, but they would have had that hardware idle quite often. Instead they maintain a small set of cloud server, growing them inside of a few hours when spikes occur, and trimming them down quickly when they are no longer needed.

    The cloud isn’t for every application. However there are probably more applications than you might expect that fit it well. As with any hyped technology, it pays to learn about the pros, cons, and potential resource costs to incorporate it and then making a case with your management for or against it’s inclusion in your IT tool belt.

    Steve Jones

    The Voice of the DBA Podcasts

    We publish three versions of the podcast each day for you to enjoy.

  • Use Tools

    Fortunately building tools for IT is easier than this.

    As someone that works with Information Technology, I usually work on applications designed to make work easier. For example, we have accounting systems that ease the handling of debits and credits, they detect mistakes in data entry, and in general require less people to handle the accounting needs of many firms. Do we have less accountants?

    I suppose we might have less, but it seems that the finance departments, handling AP and AR in many companies is as large as I remember from my younger days. Perhaps the department is smaller than it might otherwise be, but it’s usually not small.

    In IT, we have all kinds of tools available to us that can help with our jobs. What’s more, we can create new tools as needed to do our jobs. Powershell or Perl scripting, Codeplex projects, and more are available to most of us to help us better manage systems. In SQL Server, we even have a built in framework, Policy Based Management, that helps us prevent changes or problematic configurations. If you find a third party tool that can help, you can make a case for its purchase, using the time savings translated to actual costs. One of those tools might really make your job more enjoyable..

    In theory, we ought to be able to manage many more systems per person than we used to. Does that many that we need less people? Perhaps, but it seems we grow systems fast enough that we still need to hire more staff in many environments.

    I so often see people working in IT fail to take advantage of all the tools we have to automate much of their jobs. They often tell me if they automate too much of their jobs, they won’t be needed and may get let go. Personally I think that’s an excuse not to exercise your skills, challenge your mind, and get rid of tedious work There’s no shortage of work to be done in most companies. If that’s true, why not use your tools to get rid of the tedious work and spend your time on something more interesting?

    Steve Jones

    The Voice of the DBA Podcasts

    We publish three versions of the podcast each day for you to enjoy.

  • The TCO of the Cloud

    Cloud computing is growing, and it makes sense for some applications.

    I know that many of you don’t want to put your databases in the cloud. There are definitely concerns for many applications that might have identity, financial, medical, or other data. However there are many other types of applications that might make sense for cloud computing, and I am sure that many of you will encounter pressure to move to the cloud at some point in the future.

    When you do, how do you determine if the cloud makes sense? Amazon has a nice whitepaper for NoSQL databases, which examines a scenario of varying traffic for a system whose usage peaks and then declines. I don’t know how many of us have applications that follow this pattern of traffic, but I think there are plenty of our systems that are over-provisioned and have hardware sitting idle over time. Virtualization helps recover some efficiency from hardware, but the study was interesting to me in showing that the majority of the costs were in administration, redundancy, and support costs (power, space, cooling).

    Whether or not you feel the assumptions made in the whitepaper are valid, there are some interesting ideas here that show where scale can be a huge benefit. Redundancy is very expensive for single systems, but when it is spread out over the large number of systems a cloud provider maintains, it’s a minor cost. The reduction in administration costs can also be a significant savings, especially when you consider how hard it can be for a company to scale up, or down, it’s staff to meet its needs.

    I still prefer the idea of a private cloud inside a company where database services are provided, not database instances. I hope that SQL Server moves in this direction, giving us the chance to choose to manage data and consume a database service, leaving the management of the underlying instances to a specialized group that prefers managing the hardware and platform.

    Steve Jones

    The Voice of the DBA Podcasts

    We publish three versions of the podcast each day for you to enjoy.

  • Continuous Releases

    An actual graph of the features complete, but not released.

    One of the interesting things that I find with my iOS app is that they are updated fairly regularly. I have 20-30 apps, and I would guess that 3-4 of them are updated on any given week. Some of those are bug fixes, some of those are enhancements, but they are regularly changing. I don’t update them all every week, but I try not to wait too long between updates.

    Plenty of other software works like this. Windows Update provides fixed to various software on my desktop, I get a list of patches I can download when VideoStudio starts, and SQL Prompt lets me know about updates when I start SSMS. Those are all channels that allow a company to easily deliver patches to customers that improve the software. Usually these are patches, but sometimes there are improvements, like Experimental Features. In SQL Server, we have usually frowned on feature enhancements in patches, but is that what we really want?

    I noticed an interesting metric being tracked when I was in the UK. Development teams were looking at the time between when the completion of a feature (development and testing) and its release. It’s an interesting metric as there are substantial delays, which can be problematic. Users don’t get the chance to test the feature and give feedback. Developers move on to something new and may have trouble going back to enhance or patch the feature later. If there are helpful new features, and turn them on or off, wouldn’t you want to see them as soon as you can?

    Continuous Integration (CI) is a technique to help smooth the process of merging code from multiple developers, and the next step in smoothing the software life cycle is continuous release of software to get your changes in the hands of customers quicker. It’s not suited for everyone, testing needs to be more thorough, and there needs to be a smooth process that allows customers to pick and choose which updates they apply. However this can be a way to get feedback from customers, and even decide on future priorities as customers provide feedback.

    Steve Jones

    The Voice of the DBA Podcasts

    We publish three versions of the podcast each day for you to enjoy.