During the past few months I have met frequently with visitors to EMC's Executive Briefing Center. The goal of these meeting was usually to discuss the current state of the industry and where it is heading next. I often kick off these sessions by drawing a basic plumbing diagram that highlights the historically close relationship between applications and disk storage.
This diagram is a great starting point to discuss the concept of application nearness. In days gone by the application was compiled to run directly on a CPU, while the application data was stored on a physically proximate disk (on the other end of a SCSI cable, for example). ALL of the application data was placed on one physical spindle.
From this starting point we launch into a discussion about how the evolution of application workloads drove innovation into high-tech infrastructures. This inevitably leads to a discussion about the current state of high-tech data centers in 2013. There are two primary points of emphasis:
- The application distance has grown considerably.
- The geographic distribution of application data has grown considerably.
The diagrams below help to highlight these two points.
The Application Axis
The application (symbolized by the "pointer" icon at the top of the stack) is often generating data (e.g. wellness data for health care) from mobile devices outside of the enterprise.
This data has to traverse multiple networks and multiple logical and physical layers down the stack to the storage media, including:
- Multiple carriers or cloud provider environments
- A virtualization stack (e.g. VMware)
- A physical server
- Highly available hardware logic on the physical server (e.g. Host Bus Adapters or HBAs in a fiber channel environment, network adapters for IP networks)
- Multi-pathing or routing software (e.g. PowerPath to facilitate high availability and also enable high performance).
- A storage network in between the server and physical devices which usually contains highly available network switches.
- A variety of protocols supported by the storage devices, including raw block I/O, file, object, and SQL.
- The physical disk drives themselves, in the form of spinning media and/or solid state disk.
This complex application stack facilitates additional questions and conversations with customers about which pain points are the most prevalent for them, including:
How do you secure the application data end-to-end?
How do you guarantee performance (e.g. latency), especially with the traversal across so many networks?
How is configuration management handled?
How does backup and recovery across this infrastructure occur?
The Data Axis
In addition to the complexity of the application stack, the "distribution" or "span" of the storage media has transformed from being located on one physical disk drive (the left hand side of the X-axis pictured below) to being stored in multiple geographic locations.
Applications used to position all of their application data on one physical disk. Since that time the span of application data has spread considerably. Technologies such as RAID distributed the data across multiple devices, and then other technologies such as snap copy and remote mirroring distributed data even further. Each level of expansion spread the data out more and more, to the point where application data is now spread amongst clouds.
These two axes are a great discussion point with customers. The area between the axes form a "plane" or "canvas" for discussion about the future direction of the industry.
One obvious area of discussion is application performance. An I/O has to traverse multiple logical and physical layers in order to arrive at the storage media. This won't do for applications with sensitive response time requirements.
This leads to an obvious discussion about a flash strategy, but the truth is that there are other solutions that can help "collapse" this stack and improve I/O performance.
I will spend some time in my next post covering these solutions.
Steve
EMC Fellow
Comments
You can follow this conversation by subscribing to the comment feed for this post.