12 responses

  1. Mike Dolan
    December 5, 2008

    Interesting thoughts. Let me pose a couple questions to build on the discussion.

    If you take your example of MySQL… where does the telemetry service end? At MySQL, the operating system, the hardware? The challenge may be that one OSS vendor may not have enough reach to help with the stack. Your example is limited to MySQL questions so perhaps they can do targeted offerings for users with MySQL databases.

    Many large enterprises already adopt this telemetry model – they pay for remote monitoring services (for entire server/app stack) and they often compare the uptime of the individual servers to their entire datacenter (potentially with tens of thousands of servers).

    I’m not sure I’m convinced one way or the other, but just some thoughts that you may want to consider.

    I actually still do consider RH to fit your model. Consider that RH knows what software is downloaded from RHN on every server out there because RHN is a centralized service. They know that data and can customize their product in the future releases (slightly different than your monitoring service). So while they are not offering the monitoring service, they apply the same principles to guide the future direction of which packages go into their product.

    It almost sounds like a community opportunity… a centralized service could easily aggregate telemetry data across more than just MySQL – a single package installed in RHEL/SLES/Ubuntu could do it vs having an agent service for every open source app.

  2. Gordon Haff
    December 5, 2008

    To the degree that cloud computing tends to aggregate/centralize computing to a greater degree than today, I would imagine that this would make it easier in some respects to also aggregate the telemetry associated with that computing–subject of course to appropriate security and anonimization measures.

  3. sogrady
    December 5, 2008

    @Mike Dolan:

    “If you take your example of MySQL… where does the telemetry service end? At MySQL, the operating system, the hardware? The challenge may be that one OSS vendor may not have enough reach to help with the stack. Your example is limited to MySQL questions so perhaps they can do targeted offerings for users with MySQL databases.”

    given that i view this as a commercial opp first and foremost, i think it largely depends on the vendor. from a customer standpoint, would it present certain advantages to monitor the stack top to bottom?

    probably. but i also think they’d take what they could get, and if it was mysql and mysql only, it would still be worth investing in.

    as for the reach, most of the platforms i’d start with – infrastructure plays – have sufficient volume to generate useful telemetry, i’d think. much would depend on customer buy-in to the idea, but there’s an awful lot of open source around.

    “Many large enterprises already adopt this telemetry model – they pay for remote monitoring services (for entire server/app stack) and they often compare the uptime of the individual servers to their entire datacenter (potentially with tens of thousands of servers).”

    right, but i’m talking about transcending the boundaries of individual corps, which is – frankly – a leap that many of them will struggle with.

    until they see the benefits, i think.

    “I actually still do consider RH to fit your model. Consider that RH knows what software is downloaded from RHN on every server out there because RHN is a centralized service. They know that data and can customize their product in the future releases (slightly different than your monitoring service). So while they are not offering the monitoring service, they apply the same principles to guide the future direction of which packages go into their product.”

    i think RHAT is indeed well positioned to build and deliver the kind of service i’m talking about, but i don’t believe the current incarnation of RHN is precisely what i’m talking about. certainly it’s generating useful telemetry that can help guide RHAT decisions.

    but at present – as best i’m aware – that telemetry is not being resold to customers.

    “It almost sounds like a community opportunity… a centralized service could easily aggregate telemetry data across more than just MySQL – a single package installed in RHEL/SLES/Ubuntu could do it vs having an agent service for every open source app.”

    a single agent would indeed be ideal, and Canonical may have a start on this with Landscape. but ultimately, for commercial reasons, i believe it likely that vendors will pursue their own agendas b/c ownership of the dataset is imperative – that’s the value.

    as for a community effort, that might be possible, but i see several issues:

    1. cost and complexity
    2. enterprises tend to trust other enterprises
    3. management and ownership of telemetry
    4. security of the dataset
    etc.

  4. Cote’
    December 12, 2008

    Spiceworks is one of the only companies I see doing the beginning stages of this. If you dig around in their community forums, you can see they do reports like “online backup services our 500,000+ installs use.” They have the advantage of being able to pool together data from all the micro data-center inventorying they do, and because they give their stuff away for free, to a market that doesn’t freak out about data protection, they can start to benefit from telemetry faster. But, like I say, when it comes to Collaborative IT Management like this, there’s not an incredible amount going in practice, though many vendors have told me they’re working on it for the past 6-12 months.

  5. Brian de Haaff
    December 12, 2008

    Data is obviously valuable if it leads to more informed decisions. And it’s true that companies who collect massive, difficult-to-collect, heterogeneous sets of data become market leaders. See your Google example, look to Wikipedia, Facebook, or even go back in time and track Bloomberg’s rapid raise. Yet, not every wildly successful business (open source or proprietary) follows this model or would really benefit from telemetry or what we at Paglo call “social solving.” https://community.paglo.com/blog_topic/index/53–social-solving-the-power-of-community

    For example, tens (maybe hundreds) of millions of people use Microsoft Word every day, but would we all be significantly better off if Microsoft provided more info back on what the “community” was doing? Maybe auto grammar and spell checking would be better, but I am not sure how valuable the “insights” would really be. (Side note: What’s actually interesting is what is in the individual Word documents, so maybe you could argue that users could be incented to contribute some of their writings to a massive index for the communities benefit).

    Now, with that said, clearly SaaS offerings (and those that collect massive data sets that are relevant for a large community) have a real advantage where telemetry does make sense. And I think that if you look closely, you will find innovation all over the place today. As an example, our open source tool RogueScanner, for network device discovery and classification, depends on the network effect. The value an individual user receives is driven by a unique Collaborative Device Classification system which automatically looks up and identifies the device type and device’s identity in real time. RogueScanner collects information from devices on the user’s network and sends this information securely to a hosted server that uses the collected evidence and a massive database to make classification decisions. The central server benefits from being able to look up the evidence that is gathered from devices against all of devices that have ever been identified.

Leave a Reply

 

 

 

Back to top
mobile desktop