This post is more than 5 years old
7 Posts
0
3092
EMC E20-885 mirrorview resynchronization calculation
Hi, i need help with this question, i´ll explain my procedure to calculate it and let me know if i´m wrong:
A customer has two sites with a VNX system at each site. The local site has four application LUNs of 512 GB each. They are replicating to a remote site with MirrorView/S. The link between the sites has an effective bandwith of 100 MBPS. The application generates 1000 random IOPS of 4 MB with 3:1 R/W ratio of the LUNs. The link fails for five minutes.
How much data is copied to the secondary LUN as part of the resynchronization process?
- 4.6GB
- 18.3GB
- 36.8GB
- 73.2GB
First i look for the size of the Lun is 512GB so the track size is 256KB
The aplication generate 1000 random iops 3:1 so 250 iops are for writes
They mention that are 4 luns so 256KB * 4 = 1024KB
1024KB*250iops=256000KB/s / 1024= 250MB/s
The link between the sites is 100MBPS so there is a bottleneck here, each lun can use a bandwith of 62.5 MBPS.
So if i work with only one lun the answer will be 62.5 MBPS*(5min*60sec) = 18750MB/1024=18.3GB
is this correct?
andre_rossouw
62 Posts
2
May 18th, 2015 13:00
The question contains a fair amount of data that you won't need as part of your calculation.
Things you need to know: there are 4 identical LUNs of size 512 GB each, and the extent size is therefore 256 kB. [Note that the question would be very different if the LUNs had different sizes].
The total workload is 1000 IOPs, with a 3:1 R/W ratio, so that's 250 writes/s. Each write marks an extent as dirty, so the total amount of data on the dirty extents [and therefore the total amount of synchronization data that crosses the link] is 250 writes/s x 256 kB/write x 5 minutes x 60 seconds/minute = 19,200,000 kB / 1024 = 18,750 MB / 1024 = 18.3 GB.
The link speed is only important if you need to calculate how long it would take to perform the synchronization, or, in MV/A environments, what the upper limit of unique writes/s on the primary LUNs would be.
jony_pi
7 Posts
0
May 18th, 2015 13:00
Thanks for the explanation
jony_pi
7 Posts
0
May 19th, 2015 13:00
andre_rossouw
62 Posts
0
May 19th, 2015 13:00
The question says that the stated load is generated on each of the LUNs. You need to multiply your answer by 3.
jony_pi
7 Posts
0
May 19th, 2015 13:00
Look at this question from the practice test:
A customer has two sites with a VNX system at each site. The local site has three application LUNs of 128 GB each. They are replicated to the remote site with MirrorView/S. The link between the sites has an effective bandwidth of 100 MB/s. The application generates 800 random IOPS of 4 kB with a 4:1 R/W ratio on each of the LUNs. The link fails for 10 minutes.
How much data is copied to the secondary LUNs as part of the resynchronization process?
5.9GB
8.8GB
17.6GB
35.2GB
____________________________________________
3 LUN 128GB so 128KB Extend Size (im not sure about this)
800 random IOPS of 4KB with R/W 4:1 Write IOPS: 160 IOPS (Not sure again)
Data=160*128*10*60=12288000/1024=12000/1024=11.72GB
Where is my mistake?
andre_rossouw
62 Posts
0
May 20th, 2015 10:00
In the first question, the total number of I/Os is mentioned. In the second, it's the I/O per LUN, so you need to multiply by three.
jony_pi
7 Posts
0
May 20th, 2015 10:00
Hi Andre
One last question, why in the first question where: The local site has four application LUNs of 512 GB each. The answer is only using the 256KB extend size from one Lun.
And in the second question where: The local site has three application LUNs of 128 GB each. You said i have to multiply by 3 the answer
Why one is mutiplied by the quantity of Luns and the other no?