By: Andrew Shieh (@shandrew)
Edited By: Alfonso Cabrera (@alfonso__c)
Your December may be full of thoughts and planning around the future of your computing infrastructure. If so, try to take a vacation! Still thinking about your infrastructure? If you’re still operating your own hardware, you’ll need to evaluate how the cloud fits into your technology strategy. Understanding your choices requires a quick look at the recent history of the systems readily available.
A Short and Incomplete History of Cloud Computing
1995: This web thing is cool, and I can pay a web host to handle my Under-Construction.gif! But I need more on the backend, and don’t want to share resources with 254 other users. I’m going to keep running my own.
2000: I survived Y2K! This VMware product looks neat, but what in the world would I do with it? Running multiple “servers” on my servers seems like more work for me.
2006: So Amazon, the book retailer, is trying to sell me storage, by the byte? Seems odd, but I like the idea better than my current storage strategy of buying new NFS servers every time we’re nearing capacity. It typically endswith a mess of different server types and disk configurations.
2007: Hm, Amazon’s offering virtual servers too. They seem pretty weak compared to what we are running ourselves, and are a bit pricey, and unreliable. Maybe we can try them out for some temporary usage, or try using them for peak loads, but our data center’s servers are faster, cheaper and more reliable. We know how to run them.
2010: That AWS ball has really been rolling! They’ve lowered their prices, added load balancing, autoscaling, and private networking. It’s beginning to look more like my nice, stable data center. We could try them out, Netflix seems to like it, but we’re quite happy with our blazingly fast SSDs, and AWS has nothing like that.
2012: Google and Microsoft are getting into this game too? Maybe we should take this more seriously. And OpenStack makes it look like I can run the same kind of things on my servers. Wait, AWS has SSDs now? I just spent a month rotating SSDs out of all of our RAIDs to fix their #%! firmware bugs. Hmmm.
2014 : Virtual GPUs, 100 more things
2016 : 10x of 2014
The pace of development of cloud computing, especially in AWS, resembles that of a paperclip manufacturing AI. The concept of having your computing infrastructure completely hosted by a third party has gone from controversial to “boring technology”. Even the virtualization technology that we rarely surface has improved at an astonishing rate. All of these advancements feed back into the development machine, making the technologies cheaper and better.
Did I Miss the Boat?
You saw allthis cool stuff go by, but you’re still running your own data center. It’s worked for you, and you don’t have the staff and the time to migrate to something else. But your management is pushing you to have better peak load capacity, lower costs, and greater agility. You finally decide to reply to one of the hundreds of emails you have in your Vendor inbox, on Hybrid Clouds for Easy Cloud Migration. They promise you that you can keep running in your data center while you move to the cloud at your own pace. Maybe you can even keep the same infrastructure running in both.
In some short time window in our cloud past, this may have made sense. Using the cloud for peak loads made sense; using the cloud for things where the servers were disposable made sense. But while you were busy running your own metal, the cloud grew cheaper and the uptimes got way better than yours. The benefits of moving everything to the cloud grew, far past the break even point. Every day you’re not there is a larger opportunity lost.
A consultant/vendor-supported hybrid cloud solution doesn’t get you on the boat fast enough. They’re looking for your dollars every month in perpetuity, while your goal should be around migrating to 100% cloud. Supporting two technologies and infrastructures adds complexity to your architecture. It also adds cost and stretches your staff thin. The trap happens when supporting the hybrid model sucks your resources to the point where you cannot complete your migration to the cloud–the last bits are the hardest.
To avoid this trap, you’ll need to carefully plan your management of your datacenter and cloud systems. Developing their configurations in parallel rather than trying to plaster an awkward unifying layer on top will pay off in the long run. Follow best practice recommendations of your cloud provider. Join the communities of people using the same provider; your systems are inevitably much more similar, and working together is even more valuable.
Make 100%-cloud your priority, and have a quiet, pager-free SysAdvent.
No comments:
Post a Comment