Fundamentals of EMC

[Madhu Sudhan_1]

Hi guys !
Can anybody guide me how start of to know the fundamentals(Like the one mentioned below) of EMC.

1) Architecture.
2) How is it different from having multiple disks.
3) How do I interface with HP9000 N-Class.
4) How do I create logical Volumes (keeping emc in mind. With traditional disks I have done several times.)
5) Any special commands for EMC operation?


etc.,

Thanks in advance.
Happy Holidays!

...Madhu

[Darrel Louis]

Madhu,

I see a EMC or XP256(512) nothing more than a diskfarm(Fault Tolerant Disk Subsystems)

- The plan is based around a ?RAID? disk array system from EMC Symmetrix which is capable of automatically duplicating (?mirroring?) data. Furthermore the unit is fully fault tolerant when configured as proposed here. In other words, in the event of a hardware failure within the disk system (controller, cable or disk), there should be no interruption to normal operations. The failed part can, with such a RAID system, be replaced while the system is ?live? - avoiding downtime and disruption within an operational environment. The array also has a number of other avantages over convential disks in the areas of reliability, performance and repair times.

- The disk array also has the ability to be shared between multiple HP systems - extending the protection it offers to all of our production systems in time.

-One final advantage over conventional disks, which are also able to mirror data, is that the mirroring is done within the RAID system itself and does not therefore impact upon system performance.

The LV you create as on normal disks, but you should keep in mind which disk is the primary and which one the Alternate Disk(Raid Level Mirroring)

Some LVM tips:
1. Turn OFF Bad Block Relocation in LVM for all volume groups. It does not work correctly and can corrupt the customers data.
2. EMC Arrays MUST always be on it's own interface card. No other devices. Period. No exceptions.
3. Always have a LUN 0 configured.
4. Always have sequential LUNs. i.e.,
0,1,2,3,4,5,6,7. Do not skip any LUN
numbers.
5. Still only 8 LUNs per target.



http://www.enterprisestorage.hp.com/products/disk_array/index.html
http://www.emc.com

[Madhu Sudhan_1]

Congratulations on your new hat.
A Christmas gift.

Enjoy !
...Madhu

[Dave Wherry]

Madhu,
To go over your 5 points.
1) Architecture.
Darrel and Tim pretty much covered the Raid and redundancy parts. Very few single points of failure. You can connect multiple hosts using either SCSI or Fibre Channel. Several ways to build your system.

2) How is it different...
This also goes along with the first question on architecture. They are hot swappable Raid disks. Not that much different from an AutoRaid array. The big difference is the cache. It has lots of cache. All writes go to cache and are acknowledged to the host right away. The host moves on to its' next task. The data is then de-staged to the disks later when the EMC is not busy or cache is full and has to be flushed.
When doing sequential reads the EMC software has a pre-fetch routine which tries to anticipate what data the host will ask for next. It then reads that into cache and often times the next read request is serviced from cache. Much faster than waiting on the disk. I have had systems which had a 90% cache hit rate. I/O did not slow us down.

3) Connecting to hosts.
You can use SCSI or Fibre Channel. I suggest the Fibre Channel for speed and the larger number of LUNs it supports. Darrel's comment about only 8 LUNs per target is correct. However, if you use Fibre Channel you can get multiple targets per interface. If you have SCSI it may be say c1, controller 1. Then you will have 1 target on it, say t1. You have 8 devices, 0-7. You will end up with c1t1d0 - c1t1d7.
With Fibre Channel you can have I beleive up to 8 targets on that controller.
c1t0d0 c1t1d0 ... c1t7d0
.
c1t0d7 c1t1d7 ... c1t7d7
I think it is 120 LUNs per controller. Much greater flexibility.
You can also use Fibre Channel switches to connect multiple hosts to 1 port on the EMC. There are many way to build it.

4) LVM commands.
As they said, regular LVM commands. I agree with using the command line and staying away from sam.

5) EMC commands.
I last worked with an EMC over a year ago. The GUI was OK. I heard there is a new version out with many improvements. From the command line there was "inq". It is character based and gives a lot of good information you can use in scripts on the hosts.

Finally, I agree with Tim, if you are looking to purchase take a strong look at the XP products. I have been pretty happy with my XP256. The XP512 and XP48 systems are out now and give you more choices.
The biggest advantage the XP's have over the EMC's in my mind, is that you can do much more on-line and do it yourself. If you want to move a LUN from one host to another on an EMC, you need to lock the system and not make any changes for 5 days. EMC will take the bin file, which is the configuration file and modify it. Then the EMC tech will shutdown the Symmetrix, you also have to have your hosts down, and load the new bin file. Then the changes will take effect. On an XP I do all of that on-line myself.
There are several other pros and cons. In the end I prefer the XP. Of course as I said earlier, it's been over a year since I was working on EMC. Maybe they have changed some.

[Darrel Louis]

Madhu,

Two good books to look for:
- storage area networks: designing and implementing a mass storage system
ISBN 0-13-027959-5

- Building Storage Networks by Marc Farley
ISBN: 0-07-212050-9

Maybe two late Christmas Presents to ask for.

[Madhu Sudhan_1]

Thank you all.

...Madhu