Former Netflix Architect Adrian Cockcraft and Speaking In Tech host Greg Knierieman had an interesting exchange on Twitter. Adrian compared Netflix’ $100M investment in House of Cards to Zynga’s $100 outlay in a data center.
.@adrianco If you are going to compare Netflix & Zynga tech investments, then compare Netflix IT investments and AWS costs: apples/apples
— Greg Knieriemen (@Knieriemen) May 10, 2015
Knierieman explained that this was an apple to oranges comparison. On his podcast, Knierieman expanded that a more appropriate comparison would be Netflix’ AWS expenses to Znyga’s data center investment. However, the conversation reminded me of the barrier for enterprises migrating to cloud-based services.
Some of the chatter from the OpenStack Summit has been that the technology is ready, but culture is the barrier to adoption of both OpenStack and private cloud in general. I partially agree with this statement. I still believe OpenStack has a little ways to go to appeal to traditional enterprises. I question if the skillset of enterprise technology will ever align with the technical direction of OpenStack – a topic for another podcast or blog post.
— Ali Amagasu (@aliamagasu) May 19, 2015
Startup vs. Large Enterprises
Startups are at an advantage when making the selecting cloud over the on-premises infrastructure. Startups have no legacy applications or operations to consider. Developers also come in knowing there’s no safety net at the infrastructure level. Developers are forced to treat infrastructure as cattle vs. pets. Developers are forced to design redundancy and scalability into the actual application. Enterprises considering moving to the cloud have a major challenge in the form of focus on infrastructure service delivery.
The modern enterprise infrastructure is extremely redundant and offers built-in high availability to the majority of enterprise applications. Application developers can take advantage of infrastructure redundancy without much effort. Services include OS based high availability (HA), storage replication for DR and infrastructure orchestration. A developer can write the typical enterprise application based on the LAMP or .NET stack and request a level of service from the infrastructure team. For example, an application developer can create an analytics application that identifies the buying habits of their customers. The need for application disaster recovery services may. For most applications, the team only needs to work with the server administrators to ensure the OS images (pets) are replicated via storage to the DR site.
Contrast this with building a similar application on AWS. A developer needs to understand AWS architecture and how to build applications that run atop of infrastructure that could disappear at any time. Of course, the advantage is being able to build highly elastic applications. A question I don’t often see from the pro-cloud crowd (say that 5 times fast) is how many of these use cases exist in a non-tech enterprise? Are there enough of these use cases to change the nature of internal enterprise application develop and infrastructure delivery?
Wikibon had an interesting comparison of AWS and EMC Hybrid Cloud built on top of VMware vSphere. The conclusion of the comparison is that it’s the horses for courses argument. Enterprises with a large investment in redundant infrastructure may benefit from leveraging a hybrid cloud built on a “legacy” infrastructure technology such as VMware or even Hyper-V. The advantage is that the existing infrastructure services can be leveraged for traditional enterprise applications while providing a platform for elastic cloud-based applications.
Wikibon notes that startups may better spend their money on public cloud solutions such as AWS, Azure and Google Compute. I tend to agree with this approach. I don’t believe it’s reasonable for a large enterprise to pivot to a cloud delivery model for infrastructure services.