Highlighted
dingding1
3 Argentium

the difference between ALUA and symmetric active/active

HDS claims their AMS 2000 storage are symmetric active/active, EMC CX now support ALUA.

it seems they both can route I/O from the controller not own the lun to the owner controller.

so i'm really confused at what's the major difference between them?

0 Kudos
14 Replies
bjpower
1 Copper

Re: the difference between ALUA and symmetric active/active

From my understanding of the symms they don't have the same "ownership" set up as the clariions.

You provision/mask the luns to ports on the symm it does not matter what card they are on.

So you get the same effect as if you zone your hosts to multiple ports in a single SP on a clariion

ALUA on a clariion is different.

The host will only access the lun owning SP as any other failover setting.

If an issue occurs....

the IO request is sent to the non-owning SP this then gets routed through the CMI bus ( a connection between SPA and SPB  using PCI or PCI express connection)

the owning SP processes the request and sends the response back through the CMI bus to the non-owning SP. the non owning SP then responds to the host.

dingding1
3 Argentium

Re: the difference between ALUA and symmetric active/active

it seems the difference is, in symm mode, hosts can access lun via both SP. but in asymm mode, host can only access via the not owning SP only when the optimal path is failed.

0 Kudos
jps00
3 Argentium

Re: the difference between ALUA and symmetric active/active

bjpower's explanation is correct.  ALUA is not a true active/active.  Using ALUA, an I/O is serviced by the original SP which passes it to its peer for actual execution.  This can occur for some time before the original SP's LUN is trespassed to the peer, because the ALUA path in 'not-optimal'.  It uses more storage system resources and has a longer response time per I/O.

ALUA is described in the EMC CLARiiON Storage System Fundamentals for Performance and Availability whitepaper in the Availability section.  This paper is available on Powerlink.

0 Kudos
RRR
5 Osmium

Re: the difference between ALUA and symmetric active/active

So is the HDS AMS2k using ALUA or is it a true active/active system ?

0 Kudos
jps00
3 Argentium

Re: the difference between ALUA and symmetric active/active

The ams2xxx family is true active/active.

RRR, you keep on trying to get free consulting out of EMC on your HDS gear, and we're going to need a P.O.?

0 Kudos
dingding1
3 Argentium

Re: the difference between ALUA and symmetric active/active

but in AMS2k, there is also optimal path. LUN is also owned by CPU core, since CPU is owned by SP, it seems no difference compared to CLARiiON SP owned mode.

if lun is accessed via optimal path, HDS call it direct access, if not, HDS call it cross access.

after view the competition video "Trn_HDS_AMS2000-01_Symmetric_Active_Active-1209.mp4" in powerlink, i'm more confused at what's the real difference.

0 Kudos
RRR
5 Osmium

Re: the difference between ALUA and symmetric active/active

Hahahaha, I am just curious ! I'm 100% for EMC, so if it was up to me, I'd replace the USP-V and the AMS1k and AMS2k3 for a CX4-960 !

Whenever HDS is here, I'm trying to figure out about their technology and compare it to EMC.

In fact I'm very curious about FASTv2 on Clariion, since we now could use that in this HDS empire ... HDS doesn't have a similar thing...

So Clariion is Asymetric A/A and AMS2k is Symmetric A/A ? Hmmmm... so the HDS guys were right then

0 Kudos
dpm011
1 Copper

Re: the difference between ALUA and symmetric active/active

I heard Vmax isn't truly active-active but uses something similar to ALUA internally.  Can anyone verify this?

It is reasonable to me since the previous Symms had a global cache, while the Vmax is distributed nodes like the CX.  In a maximum configured Vmax, cache coherency would be very interesting if it truly maintains an active-active pathing method.

0 Kudos
AranH1
4 Beryllium

Re: the difference between ALUA and symmetric active/active

That is completely wrong. The confusion lies in the fact that the CX4-960, VMAX, and VPLEX share a common base enclosure for the engines/storage processors. But the VMAX is truly Active/Active just like the DMX and previous Symms. Beyond the fact that the CX4 and VMAX arrays run entierely different Operating Code, the VMAX is implemented differently and has additional interconnets used for the cache coherency. The VMAX product guides are very clear on this and are easily downloaded from Powerlink.

0 Kudos