Stressed? How time gets done more carving, smooth
We all know these fantastic people. , Those who seem so much to do to get away much and also appear to be able to do it again and again
Love 'em or hate' em -. You stop, never surprise us mortals. it closely though watching, you can see it at least two kinds. Just for fun, let us call the Preparer finisher and the flat-out types.
The Preparer finisher are organized systematically and take care to ensure that they do not take on another task until the previous one is on its way, safely handled and the preparations are ready for the next to will. The Flat-out types are focused on more and more tasks in recording, multi-tasking increasingly until they have reached their limits and then to keep their stamina with this load, as high as possible, since they then progressively through the work grind.
both types-Bothin observed and admired, the workplace and on the personal front, it always fascinated me, trying to work out which approach is generally best, or, alternatively, under what circumstances is an approach better than the other?
Somewhat surprisingly give the latest innovations in the XenServer hypervisor platform an insight into this question. Even better, it gives us a tunable model for seeing exactly which method works best. (That is, how much preparation and completion, you need to mix with the flat-out periods, and what are the consequences for you and others who work and live with you tasks more one way than the others do).
Let's start with a "real-life" scenario.
They come into work on a Monday morning. It's a great week. You and your team have a number of tasks in a project or an organization kickoff. So you can try an experiment - and start the half of the team with the more control, preparer finisher approach. You begin the tasks parcelling, make sure that each person is in progress and the time and bandwidth, the lists and schedules and to keep abreast of how actually do the task themselves. The other half of the team, take the flat-out approach. They say they cut through the work, as soon as they can, and give each of them as quickly as many tasks as possible. This means it with as much load as they can handle at one time, and perhaps more, once. Completely tasks already on the stack
to give a fictional scenario? Far from it. In fact, this is equivalent to what we call a storm in the computer industry. This can be a boot storm, in the login start all users of a system, or it could be, when users start spinning-up or a system, hammering when a factory or storage for work or as different time zones open to use clouds like Amazon, Rackspace or SoftLayer. So go computers through this type of scenario every day -. And sometimes several times a day
So, to place your bet, let's see who will win
with the XenServer Dundee project, our. Engineering Team have the underlying Linux platform on CentOS 7 updated and cgroups exploited tax level daemons in a separate CPU control group placing. What is this piece of techno-babble means that it the Preparer Finisher enables us approach to modeling. It allows the system to carve the time and processing bandwidth to stay organized, and both to finish a task and prepare for the next. This is less can flat-out than previous XenServer versions, in which the cgroup technology was not available
a look at the results of a computer-equivalent of the split team challenge look described above -. A. Boot Storm
The following graph shows the total amount of time to complete (on the vertical axis) all of 125 separate tasks (on the horizontal axis shown). Red dots indicate the periods of Preparer finisher approach cgroups. The green dots show the flat-out approach.
The difference in the two approaches is quite striking.
The Preparer Finisher (cgroup) approach completes all tasks in less than 60% of the time the flat-out approaches. This is because - with many virtual machines (VMs) to go flat-out much input-output do - the system does not find itself in a position to keep the control plane in motion (dealing with the preparation and completion aspects of the tasks ), as it is overloaded by the data plane are actually doing the work themselves.
Therefore, the results are very clear. To be a high achiever, prepare the time carving and to go to complete work flat-out to be preferred.
There are some other interesting information that we can glean from this model. This refers to how the system behaves when it comes to work under this kind of stress. Like their human counterparts -., If hard work, computer and people react differently when prompted "a just a bit"
As a person, I can be a bit unpredictable, as long as I'm there to do respond, something new when asked - if I have a lot of other things that have. Similarly - when I started on yet another object in addition to the ones I'm doing already, then it takes me longer to finish it - as if I was not multi-tasking to such a high degree. I'm only human, after all ... ..
and behold, systems like XenServer are exactly like humans in this regard. You can see this from the following graph, taken (on the vertical axis) the amount of time, shows complete each of 125 tasks (on the horizontal axis shown). Re - show the red dots to indicate the times for the Preparer finisher approach cgroups and the green dots the flat-out approach
re -. If the results are strongly dependent on the new XenServer Dundee Project. They show that by using Preparer finisher as cgroups approaches then we get two more useful advantages:
The time to each task is taken to do in green about the same - no matter how many tasks requested be
.There are a lot less volatility (the red lines in both graphs smooth) of task as the system object is loaded.
that makes sense. Simply put, cgroups give that reliable, high performance, less jagged working approach. The type of individual and systems everyone likes to be around and work with. What this means for XenServer end users is that this protection of the control plane processes bring not only improved boot storm times, but also increases the reaction for cloud management tasks remotely and increased stability of the host under intensive I / O load of VMs to a greater VM densities lead should
also - with large computing pools for large system users reduce the total cost of ownership (TCO) and enable .. Therefore, pay attention to speed advantages with Cloud Platform and OpenStack running on XenServer and also for the large number of XenDesktop users on XenServer (now in the majority) is on
So -. Here not much ambiguity. If you want to be a consistent high flyer, then the flat-out approach does not get you where you want to be - and that you Preparer finisher must approaches such cgroups in the new XenServer Dundee project.
A technical preview of Dundee is here from citrix.com and regularly updated alpha and beta versions are also available here from xenserver.org.