Experiences with the Dell/EMC AX150i or CX300i

[Kevan]

Hi

Thank you to all the replies in my last post regarding setting up a ESX VM Environment for a maximum of 20 virtual Servers.

As we have always been a Dell shop I am being pushed in their direction and have been quoted for a Dell/EMC AX150i and a CX300i.

From the perspective of total Disk space I can not see us getting anywhere near the AX150i maximum of 6TB so logic says that this should do for us. However I am told that the CX300i will give us an infinitely better performance because of it's fiber channel architecture.

Have forum members had good a experience with these SAN's. Should we opt for the faster CX300i? or will the AX150i be perfectly OK for what we want to do? Or should I really look out side of Dell/EMC at the Equallogic Iscsi SAN's which I can see from these forums are very popular.

Driving some of this is the fact that ideally we would like to get everything from one shop. However it has to be right and if people consider that the Dell products are not right I would appreciate there comments so as we can make an informed decision.

Another determining factor in our choice of SAN will be that we intend to use some of its available storage to migrate about 750GB worth of Department and user data from an old PowerEdge 740 NAS box that is currently our Primary file server.

We will also use some more space for a SQL database. However the SQL database is not particularly busy.

All comments and suggestions appreciated. I am quiet new to this and still getting to grips with what I need to do.

Kevan

[CoreyIT]

haha no kidding, eh. Two Londons located in a Middlesex. What are the chances. Even an Essex over here to and a sussex around here somewhere.

[TomHowarth]

good greif!!!! could be that Canada once had some Brits on tour over there. lol

[eolson-uf]

Fascinating conversation about iSCSI SANs.

We are about to deploy our first ESX servers and our first SAN. Very excited.

I am leaning toward installing either EqualLogic or LeftHand for my virtual machines that require higher performance.

However, I also have a couple of file servers that provide about 6TB of storage to our faculty and staff. The traffic to these servers is quite low -- under 400 IOPS[/b] for 99% of the time. I was considering purchasing a couple of EMC AX150i units each with 12x500GB SATA2 disks to serve that user population. Even at a conservative 50 IOPS per SATA-2 spindle, it seems to me that the AX150i ought to crank out at least 600 IOPS.

The cost per usable terabyte is VERY different for EQL/LeftHand vs. AX150i... EQL/LH over $10K per TB. AX150i is about $4K per TB. Over time, as my pool of light-use home directories grows, the savings could be enormous.

So I have a couple of questions for the group:

1) Is the AX150i up to that level of load (max 400 IOPS) ?

2) Is the proposed solution of using two different SAN vendors to provide high-performance vs. low-performance storage reasonable? or am I mad?

Best,

Eric

[doubleH]

not sure, but the equallogic PS100 and PS300 provide 60,000 IOPS and 300mb/sec xfer rate

2) Is the proposed solution of using two different

SAN vendors to provide high-performance vs.

low-performance storage reasonable? or am I mad?

I prefer standardization, but what ever works for you. You may get better pricing if you buy 2 products from the same vendor vs. 1 from each.

Best,

EricFascinating conversation about iSCSI SANs.

We are about to deploy our first ESX servers and our

first SAN. Very excited.

I am leaning toward installing either EqualLogic or

LeftHand for my virtual machines that require higher

performance.

However, I also have a couple of file servers that

provide about 6TB of storage to our faculty and

staff. The traffic to these servers is quite low --

under 400 IOPS[/b] for 99% of the time. I was

considering purchasing a couple of EMC AX150i units

each with 12x500GB SATA2 disks to serve that user

population. Even at a conservative 50 IOPS per

SATA-2 spindle, it seems to me that the AX150i ought

to crank out at least 600 IOPS.

The cost per usable terabyte is VERY different for

EQL/LeftHand vs. AX150i... EQL/LH over $10K per

TB. AX150i is about $4K per TB. Over time, as my

pool of light-use home directories grows, the savings

could be enormous.

So I have a couple of questions for the group:

1) Is the AX150i up to that level of load (max 400

IOPS) ?

[christianZ]

1) Is the AX150i up to that level of load (max 400

IOPS) ?

I would say the AX150i is for low io a good choice (alternative could be here the Siemens FibreCat SX60/80 - sata/sas with expanding shelves - Siemens wanted to make it with iscsi too-not sure when).

2) Is the proposed solution of using two different

SAN vendors to provide high-performance vs.

low-performance storage reasonable? or am I mad?

I wouldn't do it - you must learn 2 different systems - then by problems always 2 different sources.

[UMinventor]

Does anybody actually believe "not sure, but the equallogic PS100 and PS300 provide 60,000 IOPS and 300mb/sec xfer rate"? These numbers are arrived at by a marketing person who summed up the maximum capacity of the three GbE ports. It is an "up to" number that no one will ever come close to attaining.

And it is our experience that nearly all of the eql claims are ridiculous. The supposed 3rd party tests were laughable when one read the fine print. It was obvious to most that these were paid-for tests where the paying vendor got to break the knees of their competition first. There has been an on-going battle between NetApp and EMC on these "cooked" numbers that EQL then uses. Turns out that the EMC CX500 was handicapped by forcing it to do all the log file writes to a 4+1 RAID5. Four times 150 IOPS per disk is only 600 IOPS. The box got only 1750 IOPS because of the log file bottleneck. The benchmark wrote a couple records and then a log write. Well duh! Looks like the max any box would get is about 1800 IOPS with that handicap. The NetApp and the EQL boxes spread the log files over all the disks to prevent this bottleneck. Yes, the CX500 could have done that too.

One can estimate the max IOPS by taking the max IOPS per disk and summing. If a vendor says anything higher to you, then you are probably talking to a non-technical person. Use these for your estimation:

15K rpm disk 200 IOPS

10K rpm disk 150 IOPS

7200 rpm SATA 60 IOPS

So, the max a SATA based EQL is going to get is 14x60 or about 840 IOPS, not 60,000.

The lesson: some vendors try to create myths about their products. Say it often enough and people will start believing. Don't buy based on myths.

[nolent]

I have always wondered why there were so many EQL fanbois here, when it comes down to it, they really are on the outside looking in as far as performance goes.

[doubleH]

pleeeeeese. if it was so poor performing don't you think it would have come out in these fourms?

[Ken_Cline]

Don't want to sound cynical, but this is your first post and you're "bashing" a vendor product. Do you by chance work for a competitor?

[femialpha]

For your information, most of us that purchased EQL did not do it based marketing information but on hands on testing. Equallogic has been very comfortable in dropping off their appliances with customers without any hassle and it usually sells itself and i can say confidently it is a well engineered system. You can not judge the performance of a SAN or any technology by just one component but you have to consider the design of the entire solution.

All vendors, within and outside technology use unrealistic numbers for marketing and we as consumers have a responsibility to respond to such. I have expressed my feelings to equallogic about some of the numbers they claim but that does not negate the fact that it works well or they will not have the success they have attained.

[femialpha]

840IOPS? Now that you've opened our eyes to the truth, can you lead those of us who have made the mistake of buying EQL in the right path of what to buy next time?

[christianZ]

Well, we don't trust the numbers too - and made ourself tests:

http://www.vmware.com/community/thread.jspa?threadID=73745

You are welcome to make those tests with e.g. EMC box too.

[woharrow]

Nicely put Femi!

I work for EQL and I'll be the 1st to say I can't stand marketing performance numbers. Having been in the storage business on the manufacteror side for over 12 years there is one consistant. Any marketing Slick number you see will be a "100% Cache hit number" the EMC person above should know better.

Femi is dead on. When you are getting a SAN, you're NOT getting a Single Drive so to look at Drive marketing numbers isn't a wise idea. You need to look at the overall architecture as a whole and understand how it really operates and what is a real world expectation.

EQL is always willing to show up with a system to compare against anybody. We realized early on that what we are telling people sounds too good to be true, given all the "slide-ware", "Vapor-ware", and other "marke-techures" that vendors have tried to sell in the past. So we determined why should you take our word for it? We'll just bring the real deal with us and you as customers can judge for yourself. If all the other vendors don't have anything to hide and thier products are so easy, why don't they do the same?

[turobordecocina]

I don't like Dell so much, but some of my computers are from this brand. I can say something good, and it is about the quality of the material, so robot de cocina, it is something important nowadays and cómo saber si le gustas a una mujer.

[robotdecocinato]

I don't have any problem with http://www.dell.com‌ brand.

And im using it from much years