DC vs AWS

[Andrew Bruno]

Hi all,

I am looking at comparing costs between a DC provided server and what
I can get with AWS EC2, but its hard to know exactly which instance
matches what we have.

For example, one of the servers I am looking at adding is:

HP DL 360 G6
1 Processors Intel Quad-Core 2.53 GHz
16 GB RAM
BBWC Upgrade P400 512MB
6 x 600 GB HDD 10K Serial Attached SCSI Hard Drives
RAID 1+0 = 2Tbyte storage
Includes 1T byte of bandwidth in and out

What would "Intel Quad-Core 2.53 GHz" map to in EC2?

http://aws.amazon.com/ec2/instance-types/

Thanks
Andrew

[Mark Bate]

Hi Andrew,

AWS measure the speed of their virtual cores in EC2 Compute Units, with 1 Compute Unit being ~1.0 - 1.2GHz.
So CPU & RAM wise, I'd say the servers you're looking at would be close to the Extra Large Instances (m1.large).

Extra Large Instance

15 GB memory
8 EC2 Compute Units (4 virtual cores with 2 EC2 Compute Units each)
1,690 GB instance storage
64-bit platform
I/O Performance: High
API name: m1.xlarge

The RAID array could be a achieved using EBS volumes.

Although, the cloud is a little different to setting up in a DC.

In the DC, you buy to where you expect things to be. Ie. you buy the most expensive servers you can afford, so that it's a while before you need to grow again.

In the cloud, you grow your infrastructure as your needs grow. got a product launch coming up? great, fire up another couple of instances for a week or two.

Prior to this position I was at Monash Uni, where I built their nice shiny bioinformatics Xserve cluster (100x 8-core Xeons, 10G Myrinet)... it was fun, but there's a lot of headache involved in setting up in a DC.

Although, there's pros & cons to DC & AWS... AWS grows dynamically (the auto-scaling stuff is really cool!), but DC may work out better for HPC stuff (AWS have HPC nodes, but you could still have to shift large amounts of data over the line and you have to pay for bandwidth).

Personally, I'd recommend starting m1.small or t1.micro instances on AWS, and seeing how that goes.

If your app runs nicely on those, great, you've saved yourself a lot in expensive hardware, if not, try scaling up, until you find a spot where things run comfortably.

Then compare the costs of that to what it'd cost you to run in a DC (initial hardware, software, rent, maintenance, etc).

Also, I found some nice articles the other day from Involver (social marketing guys) on how they moved out of AWS with zero downtime (quite impressive) - there's a bit of info on some of the cons of EC2.

http://invalidlogic.com/2011/02/11/how-we-did-a-datacenter-migration-with-no-downtime/

http://invalidlogic.com/2011/02/16/our-pain-points-with-ec2/

Hopefully that helps,

thanks,

mark

--

Mark Bate

Wrestler of Crocodiles and Infrastructure

Mekentosj Inc.

email: ma...@mekentosj.com

twitter: @markbate

[Andrew Bruno]

The invalidlogic.com blog is very interesting. I am actually in the
opposite situation, I have a DC with 6 servers already, and need to
get another 2. Yearly cost are at about 80k, so I was researching
whether AWS is a viable option. So far, costs are similar, but I get
great DC support. Thanks for the links!

[Mark Bate]

No worries on the links... glad they were helpful.

Hmm... interesting situation.

I've found the AWS response time is pretty good, but depending on the time it can take a few hours.

Would be keen to hear what you go with there.

thanks,

mark

--

Mark Bate

Wrestler of Crocodiles and Infrastructure

Mekentosj Inc.

email: ma...@mekentosj.com

twitter: @markbate

[Andrew Bruno]

Just to clarify, I don't have any experience with AWS response time.
I am still exploring options.

I just like the fact that I can pick up the phone and call in
emergency situations, and have DC engineers look at issue. They also
call me in emergencies, CPU spikes, low disk, etc.

Does AWS premium support offer this?

Cheers
Andrew