This post is more than 5 years old
186 Posts
0
1959
Dynamic Drive Sharing
Hi guys,
I am thinking of implementing Dynamic Drive sharing and what to know more about real world workings of it.
Do you guys / Gals use it in your enterprises?
What problems if any do you have?
Do you find better speeds ect using several storage nodes connected to one library?
Would you recommend using it?
I am thinking of implementing Dynamic Drive sharing and what to know more about real world workings of it.
Do you guys / Gals use it in your enterprises?
What problems if any do you have?
Do you find better speeds ect using several storage nodes connected to one library?
Would you recommend using it?
jkernagh
35 Posts
0
September 22nd, 2009 22:00
It's possible though that you can confuse your backup environment unduly when implementing it so some considerations must be taken:
- If limiting pools to specific devices, ensure that sufficient devices are available to appropriate pools based on backup window utilisation of pools at any given time; ie. you will need to understand what pools are used during which periods (and plan for which SNs you wish to use, and when, to which devices)
- Ensure that Target Sessions for devices are set to an appropriate number to ensure a single SN host does not claim more drives than necessary, thus denying other requesting SNs access
- Consider using Group Parallelism or Pool / Device restrictions, for the reason outlined in the last point
- Ensure that if you only intend to share any given drive with specific SN hosts (not all), only those hosts are zoned to see said drive
- For the reason mentioned above, review your zones for non-NetWorker hosts which should have no access, and your NetWorker hosts which do have access for software which may attempt to access drives (leading to SCSI resets)
- Before you attempt to configure this, either persistently bind or persistently name (Windows / Linux) your drives to preclude drive ordering issues, which are more difficult in DDS environments
- Ensure you do not set up conflicting parameters in your NetWorker configuration; eg. a client whose storage node is listed as 'SN-A', but whose group is backed up to a pool which only 'SN-B' devices are allowed to access
The benefit of DDS is that it may improve performance by allowing you more customisation in how you route your backup traffic; the downside is the potential complexity. If implemented properly (on a number of levels, including NetWorker, your network, hosts and SAN) it should allow you better performance.
HTH, James.
ble1
14.3K Posts
0
September 20th, 2009 15:00
dugans1
186 Posts
0
September 20th, 2009 15:00
dugans1
186 Posts
0
September 23rd, 2009 23:00
I think i might have a backplane issue and am looking at setting up either Library sharing or DDS.
We don't run ant VTL's only PTL's. I have some pools that backup accross 2 data centers and other that only use one. when the job that uses both DC's all the tape drives go nuts but when massive exchange backups run and take 16 odd hours to complete i find the stream is slowed by other backups that execute diring the window.
It doesn't seem like it picks the speed back up again.
i was thinking of isolating these through a seperate S/N and seeing how they go.
any thoughts?
HacknDos1
66 Posts
0
September 30th, 2009 12:00
We run, a 200k+ Exchange mailbox Environment.
The way we get our backups done, completed, and in a timely manner was by using DDS and DSN.
Our Exchange servers are Dedicated Storage Nodes. This allows the server to back itself up. Since we use DDS, the tape drives are presented via fiber directly to the Exchange servers. Thus we backup around 3tb to 4tb in around 12 hours.
Each cluster of exchange servers are in different domains, use different pools, and are scheduled so they don't have any major conflicts with each other. But our Exchange clusters share 10 fiber LTO4 drives amongst each other.
It works flawlessly in our scenario, and the datacenter from a Friday evening to Monday morning backups around 80tb to 90tb of data over the weekend.
I don't backup any of my exchange servers across datacenters over copper. I have libraries in each datacenter. And not all Exchange servers backup via DSN. Some do via copper, but are much smaller in size.
I have also moved to a model of, a Networker Server in each datacenter instead of 1 networker server talking to all the storage nodes in multiple datacenters. What this does, is if a network outage should happen between any datacenter, but doesn't effect the machines in the datacenter talking to each other, my backups keep going.
Hope my info helps.
Paul
HacknDos1
66 Posts
0
September 30th, 2009 14:00
If you use the dedicate storage node method, remember although your regular network traffic is no longer moving the data a couple things are still taking a somewhat high utilization hit.
In our case we have our data on SAN. So our SAN's have about 80% utilization as an exchange server is being backed up. I would expect similar things if you are using iSCSI as well. Or your local disks to be peaked.
So you have to weigh how you will do this.
In our case, backups are done during non-peak times (weekends). Also we are doing it via SAN fiber. Our LTO4 drives are all 4gb as well as our servers.
Paul
HacknDos1
66 Posts
0
September 30th, 2009 14:00
You cant use DDS to present a drive to a server and expect it to backup localy. DDS wasn't designed for that. It was designed so storage nodes could all see and use the same drives.
When configuring a DSN, you need a free regular storage node license. Networker limitation. Really stupid one at that. Basically, when configuring a DSN you configure the drives to it first as a regular storage node. Then you go into the device list for that DSN and change the optionon the drive to show that it is a dedicate storage node. It then uses the DSN license and free's up the storage node license.
At least through the GUI, I do belive via the command line it is possible to set them all up without that hassel. But i find the command line more apt for typos and more time consuming myself.
Paul
dugans1
186 Posts
0
September 30th, 2009 14:00
Last question. If DDS is setup and you allocate the drives directly to the Dedicated Storage Node. Can the drives be used for other things like to backup a files server for example when the exchange job is not running.
HacknDos1
66 Posts
0
September 30th, 2009 14:00
Our farms are 6 active nodes with 2 passive nodes. We don't present the drives to our passive nodes. In our situation and active node down longer than an hour is a big problem. So that aside here it is.
In each of the farms(clusters) the 6 active exchange servers are presented with the tape drives so they seem them as local devices.
We configure each of these active exchange nodes as what is called Dedicated Storage Node.
In order to do this you need a Dynamic Drive Share license for each drive. Not each server.
Dedicated Storage Node license is needed for each Exchange server to backup directly to tape. Otherwise you have to go over copper to your regular Storage node.
By using the Dedicated Storage Node license with the exchange server it allows you to see and use the tape drive as a local device and use it. Thus the exchange server is backing itself up. No network traffic except back to the Networker server to say i have completed the task etc.
Hope this helps.
Paul
dugans1
186 Posts
0
September 30th, 2009 14:00
Wicked man thanks.
Think i might be putting in some overtime for the next week or so.
dugans1
186 Posts
0
September 30th, 2009 14:00
Noted thanks. In our environment we are using HP G4's as storage nodes. These have a back plane of 666mhz, I am trying to drive 6X SDLT600 drives (god knows why) in an ESL286 library. The performance i am getting is week. i think with each drive hitting between 10Mb/s and maxing out at 30Mb/s. I think this is a SN capacity issue and am testing a few more SN's to hopefully support bigger boxes being purchased.
So i am going to give your idea a go and see how it pans out. Thanks for the info i am sure this will help other and me to look back on.
dugans1
186 Posts
0
September 30th, 2009 14:00
Awesome Post thanks for the info. I just want to clarify one thing thow.
Your exchange farm is configured with the networker software as a storage node and you have presented the drives directly to the farm. Is this correct?
Sorry for the noob question. I have inherited networker and only been using it for 12 months. I have thought about this idea but have not tried it due to costs of the SN V's an exchange license.
Could you clarify please.
Thanks again, I really got a lot out of that post.
HacknDos1
66 Posts
0
September 30th, 2009 14:00
With DDS we not only share these 10 drives to our Dedicated Storate Nodes, but they are shared to our regular Storage Nodes.
So fileserver backups, system states, sql backups, and anything else non-exchange related goes back through our normal storage nodes utilizing those same drives.
When networker uses a drive, whether on a Dedicated Storage node or regular storage node, it tells all other storage nodes the drive is in use, thus they do not have access to use it.
Paul
ble1
14.3K Posts
0
October 4th, 2009 06:00
line it is possible to set them all up without that
hassel. But i find the command line more apt for
typos and more time consuming myself.
From CLI it works. For GUI I filled RFE long time ago... the best workaround approach is to use temp enablers if using GUI method to configure it.
dugans1
186 Posts
0
October 4th, 2009 14:00
I am geting some temp enablers as we speak.