Managing Infrastructure as an API

Smart system admins script. They push the work for common, well-known tasks to cheaper, junior resources. They take 140 steps to 4 steps or 4 questions to answer in a deployment script or wizard. Some customers are looking for 1 step for a complex deployment of 100s of steps. Example Real Use Cases Description Refactored Steps Add Server to Available Resources Bring a new server online, classify as web server, pre-stage containers 1) Cable HW 2) Power On HW 3) Classify 4) Discover -> Provision 30 Containers Deploy Platinum Quality of Service Application – 2 Tier Deployment Deploy JVM Container, deploys and register application across multiple tiers (network zones) and systems 1) Select Deployment Pattern 2) Select QOS 3) Deploy Service – WAIT (network config to external group) 4) Activate Service 200712181652 This model works when you have 10-100 systems. The typical system admin handles approximately 50, and 100 in well-known shops. Wizards will simply take too long in environments that need to build quickly and change quickly across 10s or 100s of nodes. Think of 1 server = NN smaller servers of nested complexity. In a year, NNN or NNNN smaller servers (ROCK and beyond) in a physical chassis. Now, there are two big areas that effect this process: virtualization (increasing levels of abstraction and inter-dependency) and demand (resulting growth and changes.) The impact of virtualization undoubtably helps in many numbers of ways. Most servers aren't utilized anywhere near where they can be. Period. But as mentioned before, tools to manage 1000s of instances, much less in nested complexity that virtualization brings to the table, don't exist. We now have virtualization sprawl whose effects will be multiplied because of the levels of abstractions. Another big change – demand -- the way programmers release code changes, new apps, and the business requirements for change have out-paced most management tools (like scripts) over the last few years. It is not uncommon for a business unit to roll out 100s of changes a year, when they used to make 1 or 2. Because of this, IT infrastructure planning now looks more like urban planning rather than simply building design. Data centers are “zoned” for the right conditions. Some areas are mixed use, others reserved for specific tasks. IT companies a few years ago promised resource pools and great grids of services. The fact is this has been far more evolutionary than revolutionary. The journey continues quickly in some areas, more slowly in others.
Comments:

Post a Comment:
Comments are closed for this entry.
About

Thoughts from Jason Carolan -- Distinguished Engineer @ Sun and Global Systems Engineering Director - http://twitter.com/jtcarolan - http://archmatters.wordpress.com

Search

Categories
Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today