The following notes were taken during an interesting meeting we had with a delegation from Dell. The post is composed in the form of a question and answers session for clarity of presentation – even though most of the information was intelligently proferred without any initial question – and the questions are not put in any particular order. The answers are not the words of the Dell delegates themselves, and I apologize in advance if there is any mis-representation.
Q: Does EqualLogic support dedupe?
A: No, dedupe is not a feature provided by EqualLogic. If you are looking for an excellent, cost-effective dedupe appliance, we can highly recommend the DR4000. The DR4000 uses inline dedupe, and it is packed with high-levels of processing power to enable high-throughput levels. This storage is highly recommended for backup storage and allows for replication of deduped backups off site.
Note: Not recommended to run production/live virtual machines off this.
Q: What snapshot mechanism does EqualLogic use?
A: EqualLogic uses the Allocate-On-Write (AOW) snapshot mechanism. This has performance advantages over the Copy-On-Write (COW) mechanism, and also allows for the live mounting of volume snapshots without affecting the primary volume.
Notes: The following extract explaining Copy-On-Write versus Allocate-On-Write is taken from http://www.dcig.com/2011/03/another-look-at-what-is-the-best-snapshot.html
A copy-on-write snapshot incurs a performance penalty anytime previously written data is changed. In these circumstances, the previously written block of data first has to be read, re-written to another block of storage and then the new data written where the other data previously resided.
Allocate-on-write conversely "freezes" data that was previously written and writes changes to that data elsewhere on other disk. This eliminates the three step "read, rewrite and write" process associated with copy-on-write and reduces snapshots to just one step - a write.
Q: Does EqualLogic support Synchronous Replication?
A: Not yet, this is in the pipeline.
UPDATE: As of FW 6.0, EqualLogic does now support Synchronous Replication.
Q: Does EqualLogic support a Thin Provisioned LUN Space Reclaimer?
A: Not yet, this is in the pipeline.
Q: What best practices do you recommend as regards RAID setup?
A: This all depends on the environment and use scenario, which is why Dell have invested $1'000'000s in developing DPACK (Dell Performance Analysis Collection Kit) which is available to partners. The DPACK can be run over a typical working day to pick up various disk utilization statistics, and will then recommend which EqualLogic model(s) to go for, and the recommended RAID setup.
Q: Does EqualLogic support data auto tiering / load balancing (intelligent data tiering)?
Notes: The following extract is taken from http://up2v.nl/2011/10/30/dell-equallogic-version-5-1-firmware-overview/
Dell added a new loadbalancer called Automatic Performance Load Balancer or APLB. This automatically moves blocks of data between pools of different controllers, RAID set or disk types. This means you can create a pool of disk drives containing different characteristics, like SAS and nearline SAS or SATA. As a administrator you do not have to worry anymore where to place the data. Also when administrator finds out an application needs more iops, he does not need to move data manually to high performance tier of storage. APLB will make sure the most accessed data will be placed on the fastest drives (the drives with the least latency).
For consideration: Consider the intelligent moving of blocks of data by the storage array to maximize storage performance, versus VMware vSphere 5's Storage DRS which has to move the entire datastore – which one do you think is better?
Q: How expandable is an EqualLogic PS-Series Array?
A: EqualLogics are expanded by purchasing additional PS-Series Arrays. You can have up to 16 members in a group and 8 in a pool. The members can be different models of EqualLogic, with different types of disks. And there is no reason why you cannot continue to expand your group over time keeping the older EqualLogics in place and adding additional EqualLogics as need arises. The older PS-Series appliances going back over 7+ years continue to support the latest firmware and features, and the current models will continue to do the same, future-proofing and maximizing the return on your investment.
Q: Can I buy storage shelves to expand my EqualLogic PS-Series Array?
A: No. The reasoning behind this is that as you add more storage shelves to a controller unit, the demand placed on that controller unit goes up and up. Also, adding more shelves to a controller unit does not increase the number of I/O ports. With the EqualLogic you expand by adding additional PS-Series storage arrays, linearly increasing the number of high performance controllers and I/O ports.
Q: Can I mix and match different types of disk in an EqualLogic PS-Series Array?
A: No, each separate array will have a set of identical disks. Within a pool you can have different types of arrays with SATA, SAS, etcetera...
Note: An exception is the PS6100XS which has 7 SSDs and 17 SAS drives with intelligent data tiering.
Q: Which EqualLogic Array would you recommend for VDI deployments?
A: An excellent storage appliance to use for VDI solutions is the PS6100XS, this specialist unit has 7 SSDs and 17 SAS drives, and uses intelligent data tiering. You'd be looking at running 250 typical virtual desktops from just one unit.
Note (i): Don't take my word for this, please check with your Dell partner first. It may well be that you will get more much more than 250 virtual desktops running off one PS6100XS.
Note (ii): A certain world famous retail company uses 3 of the predecessor units to the PS6100XS, to service a near 1000 desktop VMware View VDI farm.
Q: Does EqualLogic do CIFS/NFS?
A: The EqualLogic FS7500 is designed for CIFS/NFS implementations. Essentially you just present the storage as shares, there are no LUN configurations required.
Note: Traditionally EqualLogic arrays have always been iSCSI.
Q: Why does EqualLogic not support having active/active controllers?
A: This is a very good question. EqualLogic runs the active and passive controllers connected by a thick I/O pipe that effectively maintains the passive controller as a mirror of the active controller, this allows for near instantaneous failover in the event of a RAID controller failure – there is no need for the controller having to seize ownership of the failed controllers disks. This is supported by write cache mirroring and the write cache is cached to flash memory.
Note: The process of controller failover uses MAC spoofing and needs portfast and rapid spanning tree enabled on switches.
Q: Nothing to do with EqualLogic per-se – are 3.5" disks faster than 2.5" disks?
A: Technically speaking, with the same disk spin speed, if your data blocks are on the edge of a 3.5" disk it will be slightly faster read/write than if it were on the edge of a 2.5" disk. In practice there is nothing really in it (especially if you can pack in more 2.5" disks and thus have more spindles which means more IOPs.)
Thanks for mentioning the supporting material from DCIG.com.ReplyDelete
Last week we released the following Interactive Guide that includes Dell Equalogic ranked and scored with 52 other products from 20+ vendors (end users have access free for 30 days)
Hello Joshua, thank you for the comment. I will be checking out your Interactive Guide - will be interesting to see where the EqualLogic ranks. Cheers!Delete
You should have my email, so drop me a note for access.
You and end users may request an account here:
This is THE BEST article I've come across on Equallogics- explanations are brief and precise.ReplyDelete
Thank you Vidad!
Hi friend, nice article. However you should consider updating it as synchronous replication and some other features that you mentioned are now available in the 6.0 firmwareReplyDelete
Hi Corey, thanks for the message. Excellent news about the synchronous replication being available in the 6.0 firmware. Cheers!Delete
Anybody aware how SyncRep of EQL is handling Split brain situations?ReplyDelete
If I've got it right, EQL SyncRep is more like a manual active/passive setup. The storage admin chooses which is the SyncActive pool with the other being the SyncAlternate pool, and if the SyncActive pool goes down, then the admin manually switches to the SyncAlternate.
I might have it wrong.
The EqualLogic Configuration Guide recently updated to v13.4 has a section (6.2) that gives a better explanation.
Nice article.Tesla and Elon MuskDelete