StarWind Software

Ok so I have some quick questions that I need to answer as we start upgrading our production cluster this year. If these questions are answered elsewhere, please let me know and I will update the topic.

So I think that the 3 node cluster has 2 real flavors, Active-Active-Active and Active-Active-Passive, and have questions/confirmations about each setup.

Active-Active-Active:
1. Latency: For best performance latency of <1 ms, but can handle up to <10ms.
2. Writes: The writes are confirmed on all 3 devices before released back to the system that performed the write.
3. Reads: Using Round Robin iSCSI connections all of the devices can be read from at any time.
4. Network: All active machines need to be on the same network.

Active-Active-Passive:
1. Latency: For the 2 active nodes best is same as above, but the Passive node can handle >100ms(WAN).
2. Writes: The active nodes are are written as above, but the Passive is shipped writes on a schedule/timer.
3. Reads: Round Robin connections are used for the Active nodes, but the Passive node is not actually added to the cluster.
4. Network: Active machines must be on the same VLAN/Network segments, Passive node can be on different networks.
4. Can someone provide details on how the Passive device could be used in a DR scenario?

It appears that there is a third option, or this might replace option 2.

Active-Active-Active-Passive:
1. Latency: For best performance latency of <1 ms, but can handle up to <10ms. Passive node can handle >100ms(WAN).
2. Writes: The writes are confirmed on all 3 devices before released back to the system that performed the write. Passive receives writes as configured based on network type and latency.
3. Reads: Using Round Robin iSCSI connections all of the devices can be read from at any time. Passive node is used to rebuild the cluster if there is a complete HA failure.
4. Network: All active machines need to be on the same network. The passive machine can be on separate networks.

Hold one a second:)

For now there is only two options:
*Active-Active (2 nodes config)
*Active-Active-Active (3 nodes config)

The Active-Active-Passive Active-Passive scenarios will be available starting from next releases. To be honest I cannot give you details on how it will work exactly now, but I will gladly provide you with all the info as soon as the final product design will be known.

Any suggestions then for using the current product to do off-site replication(maybe CDP)?

You'll need to have sync channel being fast enough to handle cross-site transactional workload. How fast is your site-to-site link is?

jeddyatcc wrote:Any suggestions then for using the current product to do off-site replication(maybe CDP)?

I'm curious about this as well, I'd like to have an offsite backup of my HA volume and have been toying with ideas.

Have a 100mbit link to the other site.

Suggestions?

You`ve got three option to accomplish this:
1.Store the backups on the HA device, put one node localy, and the other one on the DR site. Please note that you need to have low latency (<5ms) and high speed (Gbps or more) channel between site to achieve this plan, so upgrading of your data link is necessary;
2.Wait for the next version of our software which will have technology that allows to easily achieve your goal;
3.Look at the 3d party software that can replicate data on DR site or maybe some network accelerators.

I hope it helped.

Anatoly (staff) wrote:You`ve got three option to accomplish this:
1.Store the backups on the HA device, put one node localy, and the other one on the DR site. Please note that you need to have low latency (<5ms) and high speed (Gbps or more) channel between site to achieve this plan, so upgrading of your data link is necessary;
2.Wait for the next version of our software which will have technology that allows to easily achieve your goal;
3.Look at the 3d party software that can replicate data on DR site or maybe some network accelerators.

I hope it helped.

I've been waiting on the next version for a while. It keeps getting pushed back. Any timeline on the release?

Beta before TechEd 2013 (end of May), release +1(2) months. Sorry for delay, more stuff is coming then promised (for example automatic tiering between flash and spindle).

eickst wrote:

Anatoly (staff) wrote:You`ve got three option to accomplish this:
1.Store the backups on the HA device, put one node localy, and the other one on the DR site. Please note that you need to have low latency (<5ms) and high speed (Gbps or more) channel between site to achieve this plan, so upgrading of your data link is necessary;
2.Wait for the next version of our software which will have technology that allows to easily achieve your goal;
3.Look at the 3d party software that can replicate data on DR site or maybe some network accelerators.

I hope it helped.

I've been waiting on the next version for a while. It keeps getting pushed back. Any timeline on the release?