perf&sizing image updates

f5devcentral · Jan 24, 2024 · 6c476c7 · 6c476c7
1 parent e9f2daa
commit 6c476c7
Show file tree

Hide file tree

Showing 4 changed files with 2 additions and 2 deletions.
diff --git a/docs/images/rseries_performance_and_sizing/image12d.png b/docs/images/rseries_performance_and_sizing/image12d.png
diff --git a/docs/images/rseries_performance_and_sizing/image15d.png b/docs/images/rseries_performance_and_sizing/image15d.png
diff --git a/docs/images/rseries_performance_and_sizing/image19d.png b/docs/images/rseries_performance_and_sizing/image19d.png
diff --git a/docs/rseries_performance_and_sizing.rst b/docs/rseries_performance_and_sizing.rst
@@ -133,7 +133,7 @@ To see how this translates into real performance, it is good to look at a Layer7
   :align: center
   :scale: 90%
 
-Because each appliance has a different number of CPUs, a common sizing exercise is to look at the per vCPU performance by using the formulas above to come up with a per vCPU metric. In the graph below, it is done for Layer7 RPS (Inf-Inf), but you could use the same math for any metric. The graph below is not derived from a per vCPU test, it is taking a published appliance metric and dividing it by the number of vCPUs (minus the platform vCPUs) to come up with a per vCPU metric. As mentioned above, using the rSeries metric which is (minus the platform CPUs) is the most realistic. As noted below migrating from an i10600 to an r10600 will have better per vCPU performance. This is also the case when migrating from an i11600 or i11800 to an r10900. There is one case where the per vCPU performance is lower. When going from an i10800 to an r10800 the per vCPU metrics are lower on rSeries even though the aggregate performance is higher for the entire appliance. This is due to the speed of the processors, but since there are more processors, the aggregate performance is higher.
+Because each appliance has a different number of CPUs, a common sizing exercise is to look at the per vCPU performance by using the formulas above to come up with a per vCPU metric. In the graph below, it is done for Layer7 RPS (Inf-Inf), but you could use the same math for any metric. The graph below is not derived from a per vCPU test, it is taking a published appliance metric and dividing it by the number of vCPUs (minus the platform vCPUs) to come up with a per vCPU metric. As mentioned above, using the rSeries metric which is (minus the platform CPUs) is the most realistic. As noted below migrating from an i10600 to an r10600 will have better per vCPU performance. This is also the case when migrating from an i11600 or i11800 to an r10900. There is one case where the per vCPU performance is lower. When going from an i10800 to an r10800 the per vCPU metrics are lower on rSeries **even though the aggregate performance is higher for the entire appliance**. This is due to the speed of the processors, but since there are more processors, the aggregate performance is higher.
 
 .. image:: images/rseries_performance_and_sizing/image12d.png
   :align: center
@@ -204,7 +204,7 @@ To see how this translates into real performance, it is good to look at a Layer7
   :align: center
   :scale: 90%
 
-Because each appliance has a different number of CPUs, a common sizing exercise is to look at the per vCPU performance by using the formulas above to come up with a per vCPU metric. In the graph below it is done for Layer7 RPS (Inf-Inf) but you could use the same math for any metric. Note the graph below is not derived from a per vCPU test, it is taking a published appliance metric and dividing it by the number of vCPUs (minus the platform vCPUs) to come up with a per vCPU metric. As mentioned above using the rSeries metric which is (minus the platform CPUs) is the most realistic. As you will note below, migrating from an i5600 to an r5600 will have better per VCPU performance. This is also the case when migrating from an i7600 to an i5900. There are two cases where the per vCPU performance is lower. When going from an i5800 to an r5800 or when going from and i7800 to an r5900 the per vCPU metrics are lower on iSeries. This is due to the speed of the processors.
+Because each appliance has a different number of CPUs, a common sizing exercise is to look at the per vCPU performance by using the formulas above to come up with a per vCPU metric. In the graph below it is done for Layer7 RPS (Inf-Inf) but you could use the same math for any metric. Note the graph below is not derived from a per vCPU test, it is taking a published appliance metric and dividing it by the number of vCPUs (minus the platform vCPUs) to come up with a per vCPU metric. As mentioned above using the rSeries metric which is (minus the platform CPUs) is the most realistic. As you will note below, migrating from an i5600 to an r5600 will have better per VCPU performance. This is also the case when migrating from an i7600 to an i5900. There are two cases where the per vCPU performance is lower. When going from an i5800 to an r5800 or when going from and i7800 to an r5900 the per vCPU metrics are lower on iSeries. The per vCPU metrics are lower on rSeries **even though the aggregate performance is higher for the entire appliance**. This is due to the speed of the processors, but since there are more processors, the aggregate performance is higher.
 
 .. image:: images/rseries_performance_and_sizing/image15d.png
   :align: center