This post is more than 5 years old
295 Posts
0
928
November 2nd, 2010 11:00
NS120 - for replicator to get performance and space what is better?
replicate 1 big file system of 10 teras
or replicate 5 file system of 2 teras each one,
how the space of savvol and checkpoint is estimated?
No Events found!
umichklewis_ac7b91
300 Posts
0
November 2nd, 2010 13:00
As with all things in life, "It Depends." First, are you talking about a one-time replication (migration) or are you talking a regular replication? I'm assuming you're talking about regular replication, as one-time replications will only perform as well as they'll perform.
When you create a checkpoint for replication, the initial SavVol will be a minimum of 20GB. So, to replicate your 10TB filesystem, you'll immediate use 20GB. However, if you're replicating (5) 2TB filesystems, each filesystem has its own replication session, hence, it's own 20GB SavVol. Now, your total initial SavVol usage is 100GB (5 x 20GB). So to save initial space, it appears than replicating one large filesystem initially uses less SavVol space than multiple, smaller SavVols.
In terms of performance, however, the larger the filesystem, the longer it usually takes to create a checkpoint. Logic follows that a 1TB checkpoint is faster to create than a 10TB checkpoint. Also keep in mind that larger checkpoints take longer to be checked/mounted following a failover or reboot (there are several threads about reboot performance being impacted by many, large checkpoints - take a look at them).
All things being equal (which they never are!), you might find replicating several small filesystems will probably perform acceptably, but may use more SavVol space than replicating a single, large filesystem. But after the initial replication has occured, you'll find that replicating the changes will take the same, regardless of the size of the filesystem.
If you have the free space, you might try a local replication to estimate how much space you'd use up for the replication process.
Thanks!
Karl
cxCpyIzSVS12414
43 Posts
1
November 2nd, 2010 14:00
It is always a best practice to keep a smaller file system and replicate them.
replicate 5 file system of 2 teras each one
Regarding how the space of savvol and checkpoint is estimated.
For any production file system greater than 10GB, the default SavVol size is 10GB. For any pfs under 10GB, the SavVol will
equal the pfs size, unless otherwise provisioned. Each pfs will have its own dedicated SavVol.
Default creation and extension size of the SavVol is 20GB beginning with NAS 5.6.44
In case of replication, on assuming the filesystem is full. The SavVol will be equal to the size of the filesystem.
Also refer the below content:
https://community.emc.com/message/383536#383536
jelucho
295 Posts
0
November 2nd, 2010 18:00
i will do regular replication,
1. what i dont understand is how much space do i have to have in free space pool.
in order to replicate these 5 systems of 2 teras bytes, and to avoid that the replication stops with a warning of full savVol
2. i read in somewhere that when i configure to replicate a file system from one nas to other nas
automatically that process creates 2 checkpoint in every side of the replication.
how this works?
3. other questions: how many file system of my configure can i to replicate at the same time.
thanks
mc.
umichklewis_ac7b91
300 Posts
0
November 3rd, 2010 08:00
If you're looking for absolutes, i.e., "I must have XXX TB free to ensure continuous replication", I'm not sure anyone can tell you - especially without information about your daily change rates and so forth. Why? If one of your users wipes out 5TB of your 10TB filesystem, those changes get pushed into the SavVol (the changed blocks). If they delete 1TB then replace it with a new 1TB, those changes get pushed into the SavVol. Do you have a hint of how much your data changes over the course of a day?
Both of these questions are answered in the guide Using Celerra Replicator v2, available in PowerLink. Page 37 has a nice diagram, with step-by-step details about how and why the two checkpoints are created and so on. In terms of how many simultaneous replications you can run, one source can replicate to up to four destinations. From a performance perspective, however, you have to know a little bit about your change rate. Any idea about how fast things are changing?
In our environment, we have four VDMs and eight filesystems being replicated at 5min and 10min intervals. Our VDM deltas are only a few MB in total; filesystem deltas are anywhere from 20MB to as much as 30GB in that time frame. I have yet to see a performance impact from replication. Unless you're replicating hundreds of GBs every few minutes, it's hard to say if you'd see any impact. As with all things, it depends on your replication bandwidth and change rate. Replicating 1TB over a T-1 will probably not impact performance (since your changes are just accumlating in the SavVol), but it certainly take a long while to complete the replication process.