99
edits
Changes
Start
{{UBC toc}}
This page describes a very promising way of beancounters optimisation.
== Current accounting model ==
Basically allocation of any kind of resource looks like this:
<pre>
struct some_resource *get_the_resource(int amount)
{
struct some_resource *ret;
ret = find_or_allocate_the_resource(amount);
return ret;
}
</pre>
We change this behaviour to work like this:
<pre>
struct some_resource *get_the_resource(int amount)
{
struct some_resource *ret;
if (charge_beancounter(amount) < 0)
return NULL;
ret = find_or_allocate_the_resource(amount);
if (ret != NULL)
return ret;
uncharge_beancounter(amount);
return NULL;
}
</pre>
The <code>charge_beancounter()</code> call is responsible for checking whether the user is allowed to consume desired amount of the resource, i.e. if the resource consumption level is lower than the limit set.
Obviously, this change slows down the original code, as charge_beancounter() takes some slow operations like taking locks. We have an idea of how to optimize this behavior.
== On-demand accounting basics ==
The main idea sonds like this:
{{out|If the cunsumption level of any resource can be easily upper estimated with some value, and this setimation is lower than the limit, then we do not need to know the exact consumption level and allow the resource allocation without additional checks}}
Apparently, when the estimation exceeds the limit, we must switch to the slower mode, that will give us more presice value of the consumption level and (probably) allocate another portion of the resource.
== Examples ==
Let's look at some examples of how this will work.
=== The user memory ===
Currently we account for the [[physpages]] resource. That is -- the number of physical pages consumed by the processes. The accounting hooks are placed inside the page faults and hurt the performance. The accounting looks like this:
<pre>
struct page *get_new_page(struct mm_struct *mm)
{
struct page *pg;
if (charge_beancounter(1) < 0)
return NULL;
pg = alloc_new_page(mm);
if (pg != NULL)
return pg;
uncharge_beancounter(1);
return NULL;
}
</pre>
However, we have a good estimation of the RSS size -- that is the lenghts of mappings of the processes. Since the physical pages can only be allocated within these mappgins the RSS value can never exceed the sum of theis lenghs. The accounting will then look like this:
<pre>
struct vm_area_struct *get_new_mapping(struct mm_struct *mm,
unsigned long pages)
{
if (!mm->fast_accounting)
goto allocate;
if (charge_beancounter(pages) == 0)
goto allocate;
mm->fast_accounting = 0;
recalculate_the_rss(mm);
allocate:
expand_mapping(mm);
}
struct page *get_new_page(struct mm_struct *mm)
{
if (mm->fast_accounting)
goto fast_path;
if (charge_beancounter(1) < 0)
return NULL;
fast_path:
pg = alloc_new_page(mm);
if (pg != NULL)
return pg;
if (!mm->fast_accounting)
uncharge_beancounter(1);
return NULL;
}
</pre>
We do not call the slow <code>charge_beancounter()</code> function in the page fault (<code>get_new_page()</code>). Instead we account for the upper estimation in <code>get_new_mapping()</code> call that happens rarely and thus increase the performance.
Note, that the <code>recalculate_the_rss()</code> is called to calculate the exact RSS value on the beancounter.
This page describes a very promising way of beancounters optimisation.
== Current accounting model ==
Basically allocation of any kind of resource looks like this:
<pre>
struct some_resource *get_the_resource(int amount)
{
struct some_resource *ret;
ret = find_or_allocate_the_resource(amount);
return ret;
}
</pre>
We change this behaviour to work like this:
<pre>
struct some_resource *get_the_resource(int amount)
{
struct some_resource *ret;
if (charge_beancounter(amount) < 0)
return NULL;
ret = find_or_allocate_the_resource(amount);
if (ret != NULL)
return ret;
uncharge_beancounter(amount);
return NULL;
}
</pre>
The <code>charge_beancounter()</code> call is responsible for checking whether the user is allowed to consume desired amount of the resource, i.e. if the resource consumption level is lower than the limit set.
Obviously, this change slows down the original code, as charge_beancounter() takes some slow operations like taking locks. We have an idea of how to optimize this behavior.
== On-demand accounting basics ==
The main idea sonds like this:
{{out|If the cunsumption level of any resource can be easily upper estimated with some value, and this setimation is lower than the limit, then we do not need to know the exact consumption level and allow the resource allocation without additional checks}}
Apparently, when the estimation exceeds the limit, we must switch to the slower mode, that will give us more presice value of the consumption level and (probably) allocate another portion of the resource.
== Examples ==
Let's look at some examples of how this will work.
=== The user memory ===
Currently we account for the [[physpages]] resource. That is -- the number of physical pages consumed by the processes. The accounting hooks are placed inside the page faults and hurt the performance. The accounting looks like this:
<pre>
struct page *get_new_page(struct mm_struct *mm)
{
struct page *pg;
if (charge_beancounter(1) < 0)
return NULL;
pg = alloc_new_page(mm);
if (pg != NULL)
return pg;
uncharge_beancounter(1);
return NULL;
}
</pre>
However, we have a good estimation of the RSS size -- that is the lenghts of mappings of the processes. Since the physical pages can only be allocated within these mappgins the RSS value can never exceed the sum of theis lenghs. The accounting will then look like this:
<pre>
struct vm_area_struct *get_new_mapping(struct mm_struct *mm,
unsigned long pages)
{
if (!mm->fast_accounting)
goto allocate;
if (charge_beancounter(pages) == 0)
goto allocate;
mm->fast_accounting = 0;
recalculate_the_rss(mm);
allocate:
expand_mapping(mm);
}
struct page *get_new_page(struct mm_struct *mm)
{
if (mm->fast_accounting)
goto fast_path;
if (charge_beancounter(1) < 0)
return NULL;
fast_path:
pg = alloc_new_page(mm);
if (pg != NULL)
return pg;
if (!mm->fast_accounting)
uncharge_beancounter(1);
return NULL;
}
</pre>
We do not call the slow <code>charge_beancounter()</code> function in the page fault (<code>get_new_page()</code>). Instead we account for the upper estimation in <code>get_new_mapping()</code> call that happens rarely and thus increase the performance.
Note, that the <code>recalculate_the_rss()</code> is called to calculate the exact RSS value on the beancounter.
