For this week’s blog posting I’m going to reflect on operating a data center as well as some of the current trends impacting data centers. A couple of blog postings ago I wrote about how SMBs are anchored down by maintaining their own data centers, stunting progress with the core business and other areas of IT such as SOA implementations and analytics, so it seems worth exploring data centers a bit deeper.
For my first reflection, I wanted to find a real-world example of a business moving entirely to the cloud, ideally an SMB. In my research, I came across this article from last year on how Netflix finished moving its entire business to Amazon’s cloud platform. Netflix is far from an SMB, but the article was too compelling for me to ignore.
The second paragraph immediately answers the important question of “why?”. In 2008 Netflix experienced a serious outage with their data center, which is probably the primary impetus behind most cloud migrations, especially for SMBs where, I believe, proper investments in reliability are lagging. Many businesses will examine the costs and benefits and cautiously contemplate a move to the cloud, but as soon as there is an outage, that cautious contemplation quickly becomes an irrepressible demand, which speeds matters up.
The article, unfortunately, doesn’t provide any cost comparisons between the Netflix operated data centers and the cost of operating in the Amazon cloud, but direct, measurable costs are probably not going to be the deciding factor. It’s going to be the indirect costs associated with outages that will be the deciding factor, along with the realization that running a data center is simply not a core capability of most businesses. Focusing on the core and offloading or outsourcing most of the rest is a common and sound strategy these days.
The only piece of their home-grown infrastructure that Netflix did not move to Amazon’s cloud was their content delivery network, which essentially consists of all the video caches around the world, allowing their content to be close to the customer. There are plenty of edge computing companies, including Amazon, that could handle this for Netflix, so one can only assume that Netflix considers their content delivery network a core capability and competitive advantage - something they do better than others - while the rest of the IT infrastructure was not.
It’s ironic to note that as I was writing Reflection 3 (below), Netflix had a serious worldwide outage for at least one hour. It would be interesting to know if it was Amazon’s doing or something self-inflicted by Netflix. Regardless, I don’t think Netflix will be reversing direction any time soon.
In looking for articles on data centers and moving to the cloud, this article, while not providing insights on the value of moving to the cloud, was still too interesting to ignore. I’ve always marveled at Google’s ability to do what they do, especially given the performance problems of the small, custom-built application that manages the back-office operations of my current employer.
What amazed me was the revelation that Google now designs their own hardware to address their unique performance and load requirements rather than use commercially available options, which Google deemed unsuitable for their needs.
At first, I was curious why Google hasn’t commercialized their networking/hardware innovations, but I quickly concluded that they must consider such innovations a competitive advantage and, as such, choose to keep their technology private.
More importantly, I wondered how networking giants like Cisco could be out-innovated in their core area of expertise by a company who’s core is not in networking and hardware. Is it because commercializing hardware and software for a mass-market requires compromises in order to achieve certain cost and utility goals? I’m going to assume Cisco knows how to do what Google did and probably figured it out before Google did. Cisco must have concluded that the capabilities were too specialized and/or too expensive for the commercial market - if I were to give them the benefit of the doubt.
Conversely, it could be that Cisco is withholding innovations that would disrupt their own market segment. That’s not unheard of, albeit very risky, as I have to believe there are numerous other potential disruptors, that unlike Google, would be happy to commercialize such innovations.
Reflection 3: Why Hyperconverged Infrastructure is so Hot
One of Google’s innovations (from the previous reflection) was to build software-based switches using cheap hardware. This sounded very similar to the recent trend in data centers called hyperconvergence, which is the third article I want to reflect on.
The hyperconvergence trend simply takes the software-based virtualization paradigm for compute environments and adds storage to the mix. This allows data centers to simplify and further extend the cost savings and flexibility that virtualized compute environments have long delivered.
The article compares and contrasts the hyperconvergence trend with its convergence predecessor. Both trends try to commonize on hardware, however, the convergence trend still saddles a data center with separate storage and compute devices, which in turn, requires an emphasis on expensive networking hardware. By contrast, hyperconvergence melds storage and compute platforms together, managing them from the same software-driven virtualized environment, in turn, deemphasizing the network and the need for expensive networking hardware.
It is interesting how strong a resemblance the hyperconnectivity trend bears to Google’s data center innovations from Reflection 2. In Reflection 2, I pondered why Cisco (or others like them) hadn’t beat Google to the punch with such software-driven data center innovations. One of my speculations was that it might be too disruptive to their lucrative market. I find it interesting that hyperconvergence, with its emphasis on virtualized storage and compute environments and the de-emphasis on networking, seems to support that speculation. There should be no need for the networking giants to worry though as I’m sure the Internet holds plenty of opportunities for growth.
Great post! I really tend to agree with your thoughts on the Netflix article especially and found it very intuitive. We have studied the benefits from moving to the cloud and some have even gone through an actual migration at their jobs, but it's interesting to see the reasons for other companies moving to the cloud as well. Great perspective.
ReplyDeleteHi Scott,
ReplyDeleteVery interesting post on Netflix's migration to the cloud. That must have been a huge decision to move completely to the cloud. Every migration to the Cloud has to be well justified by a business driver like cost, technology, etc. This is where the role of EA gets to be challenging to be architecting migrations to the Cloud and as per their Cloud strategies. One of the commonly forgotten component in a Cloud strategy is not accounting for an exit strategy. There have been organizations that have got burnt, since exit strategies are not easy to implement, in case there is a need. A great post. Thanks for sharing it.
Veena.