The tangent
There is a little used term that I love, "bespoke". It isn't something you would use everyday and very few people would ever use it in a year or possibly their whole lives. It refers to hand tailored suits, hand made shoes, really expensive shotguns - costly and exclusive items. But it means even more than "handmade". It is the utmost in custom, where every single item in a design is measured, discussed and "spoken for": "bespoke". I don't actually have anything that qualifies as "bespoke", but I love the concept.
Here's the deal: We have a legacy of bespoke systems in the Information Technology industry.
I was once an IT architect and spent huge sums of money creating bespoke systems. Every element of application and systems hardware and software was measured and discussed and "spoken for". We gathered our experts and solved problems uniquely. I lived a trait common among engineers where we optimize, overbuild, and make highly modified solutions specifically fit to the problem being solved. This is expensive to build and hard to manage.
For a contrast, look down at the clothes you're wearing right now. You aren't wearing anything bespoke, there's very little chance you're even wearing anything that was even tailored at the cuff or hem. Even if you are wearing expensive clothes I bet they aren't bespoke. If this was a time before the 1800s, every item you have on would have been made by hand, down to the fabric and thread. While the finery of the rich was drastically different than the rags of the poor, it was all handmade for an individual.
Back then this wasn't bespoke so much as it was the only way to solve the problem of clothing. You needed clothes, somebody had to make them specifically for you.
how an industry can change your daily existence
This ended in the 1800s. In 1813 Francis Lowell created the power loom. Creating fabric on a hand loom is tedious and slow, the power loom could replace dozens of workers and output material far faster with less labor. The power loom was set to change an industry but the industry didn't know it needed a change, taking over a decade before it reached large scale use and replaced hand loomed fabric. Even then, the power loom solved the problem of creating fabrics more effectively but ready-to-wear clothing didn't really catch on until the invention and adoption of two other important technologies, the sewing machine and the graded paper pattern. The sewing machine provided speed of construction but even it wasn't enough to shift to mass produced clothes.
The graded paper pattern was the final ingredient. It was smallest of the three in terms of leaps in technology. It wasn't some new machine. It was the embodiment of an idea. That idea was a way to create consistently sized clothes that matched each other and didn't require an expert to design and build each piece. The pattern - how to cut and where to sew pieces of fabric - made quality clothing achievable by a factory trained worker, no tailor needed.
The combination of those three items enabled a full paradigm shift from handmade to a mass produced garment industry.
It took the creation of this industrial ecosystem to enable ready-to-wear clothes. It was one of the first really large positive impacts of industry in society. We went from owning and wearing few clothes to owning a lot more. The concept of "fashion" was adopted in lower classes of society when it was only a notion for the very rich before. Modes of dress changed. It revolutionized what before was a static way of living for centuries. The realm of hand tailored clothes switched to a privileged and exclusive reality. The masses wore ready-to-wear. The clothes were cheaper, the quality improved, poor people might actually have a change of clothing, things got better.
Blowing stuff up
Jump forward about one hundred years and think about the first computers. They were hand built of expensive materials by highly trained experts to solve very specific problems. (Generally the first problems took the form of either cracking encrypted information or getting better at blowing something up.)The computer was the same as the application at first, it did one thing. General purpose, programmable computers followed. The computer then started solving problems of collating and manipulating data for government census. The next decades saw business buying computers to solve accounting, HR and other needs. We've been through mainframes, minicomputers, engineering workstations, client-server models, virtualization. What we still do a lot of is spoken for components and specifically designed systems and infrastructures.
But, just like clothes serve similar purposes on every body and an industry was created be to deliver consistency, performance, and better price point: business computing problems aren't that unique. Why do businesses still choose bespoke systems solutions?
You may have responded "cloud" in some way: "we are generic", "the cloud is ready-to-wear", "our industry has moved beyond this". And you would be kind of right, but we've only moved a few applications to the cloud and we are often doing fundamentally new stuff there. We haven't moved away from enterprise compute, networks and SANs. We don't completely trust the cloud with everything and we never will. We are still building highly complex enterprise environments.
What's missing? Where is our power loom and sewing machine and graded patterns? How do we get to a place where we have IT that is ready-to-wear?
We are there. Just like the power loom and sewing machine, the technologies arrived but the industry is taking time to catch up. There was a final missing piece and it showed up a few short years ago.
Bring it on home
Virtualization replaced custom compute hardware with immortal virtual machines. Software defined networking has pulled network smarts down into the systems. We have orchestration, rapid deployment, we can move virtual machines while in production, and we've had all of these things for a while. The last custom and bespoke hardware that required special networks and big budget design and acquisition was storage. Storage needed to live in refrigerator sized cabinets and were separated from the amazing changes happening in compute. Storage is the anchor and SANs are the chain.
To shed this anchor we had to provide resiliency and performance in the same systems that virtualized everything else. Flash was the enabling bit. Instead of hundreds or thousands of drives in storage arrays, you can run enterprise levels of IOPs at even lower latency than spinning media- with just a few SSDs. Smart software that takes advantage of all of this can put you in the the ready-to-wear consumer space for your IT infrastructure.
Here is the big question though: Now that you have all of this, what have you gained? Why does a shift from bespoke to read-to-wear matter in information technology.
The answer is scale and more specifically web-scale.
Google, Amazon, Microsoft Azure and a few other companies have applications running at web-scale and infrastructures to support them. That being said: Enterprise applications aren't, in general, web-scale. You don't see distributed computing concepts and technologies in applications that were originally created 20 or more years ago. These apps won't take advantage of MapReduce, Cassandra, Zookeeper and all other components of the modern distributed computing world. Old enterprise applications don't do web-scale. But If you provide a web-scale infrastructure for them to live on, you get Webscale benefits while running your old apps.
Many of the advantages of web-scale aren't in the massive parallelization of jobs running on large data sets. That is the sexy side that makes it into the industry press. The resilient infrastructure underneath it all doesn't get talked about as much but that's the foundation for it all. The ability to add in more compute and storage and to scale horizontally is the other part. What hyper-converged systems like Nutanix has built, is simply provide the last piece of the technology puzzle that makes IT go from bespoke to ready-to-wear.
Nutanix software presents to hypervisor hosts in a way that allows UVMs to use what appears to be traditional storage. In reality, the storage controller is another VM on the same hypervisor and this controller VM (CVM) interacts with other CVMs on other hosts to provide a web-scale storage infrastructure to all of the UVMs in the hypervisor cluster. You can add nodes as needed and use hypervisor techniques to spread load and move VMs without disruption. Add in some orchestration and you are creating the material, using the patterns and constructing an IT environment like it is an IT factory - not a bespoke tailor shop.
Keep bespoke for insanely expensive clothes, shoes and shotguns. Your IT should be ready-to-wear.
No comments:
Post a Comment