Operating System - OpenVMS
cancel
Showing results for 
Search instead for 
Did you mean: 

Not a VMS question but a questions for VMSers

 
Wim Van den Wyngaert
Honored Contributor

Not a VMS question but a questions for VMSers

1. If you had a company with let's say 500 employees and a few (windows) servers, would you feel safe with MSFT Azure as a replacement of those servers (yes, running Windows, not VMS) ? I would think that when I use a server in the cloud near my company it could be ok but when I fail over to the US or India ... would the internet still give acceptable performance (speeds are higer over here) ?

2. MSFT says you would save about 50% on TCO. Anyone has real data or reasons to think it would not be so (I would say migrations are always more expensive than the planned number of days and MSFT is never cheap).
http://www.microsoft.com/windowsazure/offers/
will let you see a MSFT estimate of TCO.

3. Any other remarks are welcome.

Wim
Wim
4 REPLIES 4
Robert Gezelter
Honored Contributor

Re: Not a VMS question but a questions for VMSers

Wim,

My concerns with "clouds" are generic, not specific to any vendor.

While clouds are useful, there are a number of challenges. Some challenges are technical, others are legal and operational.

In "Why Settle on a Hosting Provider? Bandwidth Liquidity and Other Issues" (see http://www.rlgsc.com/blog/ruminations/why-settle-on-a-hosting-provider.html ), I noted that there are incentives for cloud providers to minimize hardware provisioning. For non-mission critical tasks, this is one thing. For mission-critical (e.g., "shut the business down" computing), this is a whole different story. Note that I am not writing of real-time process control, if employees cannot accomplish their tasks, I define that as "mission critical".

Locality is one thing, transcontinental failover is also a different issue. The other year, India was substantially offline for several days (I would have to check the precise time), when a Red Sea cable was severed by a dragged anchor. The challenge is "What happens when my data is on the other side of the cable break?"

Legally, there are often issues when data is out of one's control, both from a security standpoint and from a process standpoint. I do not know the particulars of the law in your jurisdiction, but there is often a world of difference between data residing on a server under your control and data on someone else's server. Whether a cloud provider will resist as diligently as your own counsel is, to put it politely, an open question.

TCO statements without full (emphasis: FULL) backup are not meaningful. If the "savings" is realized by fractional provisioning, and over-allocation, I am not sure if I would consider the "savings" real.

In general, caution is advised. I also presented a session entitled "Agility, the Cloud, and Accountability: What you can't know can kill you" (see http://www.rlgsc.com/trentoncomputerfestival/2011/agility-the-cloud-accountability.html ) at the 2011 Trenton Computer Festival, which may be of interest.

- Bob Gezelter, http://www.rlgsc.com
Wim Van den Wyngaert
Honored Contributor

Re: Not a VMS question but a questions for VMSers

I will share what I find. Correct me when I'm wrong.

http://www.microsoft.com/downloads/en/details.aspx?displaylang=en%20&FamilyID=d32702dd-a85c-464d-b54d-422a23939871 at the bottom.
If Azure doesn't give 99% availability you get 25% reduction on your invoice. Poor SLA IMO. Will not motivate MSFT to do
their job properly. SLA's may be updated every term by MSFT. Customers must post a claim to get a refund (no automatic price reduction).
Only 1 component (storage/network/cpu/sql) can be the subject of a reclaim and for claims on availabilty of cpu you have to
take (and pay) at least 2 instances/roles. Performance may be bad during maintenance on the servers without compensation.
MSFT may take the systems down for 10 hours per year without compensation.
And of course, MSFT can not be responsible for third party failures. No internet to India equals bad luck if your node must run
over there. No computing nor compensation for you.
And data loss ? No compensation according to the SLA.

You pay for the hours you use. But you have to do a shutdown of some kind to stop the usage. Just not using it is not enough.

DB size limited to 50 GB. Want more ? Split your db. Additional work. And maintenance. And costs.

Using VMWARE on your Azure server ? Bad luck. Only Hyper-V is supported. And Windows 2088 SP2 and higher.
Foget clouding your legacy systems running on e.g. XP or Vista.

How do the servers look like ?
http://www.datacenterdynamics.com/focus/archive/2011/04/microsoft-optimizes-servers-for-azure,-bing-clouds

I've read forums where claims are made that Azure SQL is up to 10 times slower as local servers. Depending
on how it's coded. But if you stay local (the same Azure wathever that means) it could be equal performance.
But in case of a disaster you might get awful performance.

In terms of availability in the Asia-Oceania region, Google App Engine has the lowest overall availability among
the 18 cloud service providers being monitored at CloudSleuth, averaging at 82.88% in the past 30 days (even 7% in Beijing).
(calculated on 23rd March). Azure had no delays, which is good.

Wim
Wim
John McL
Trusted Contributor

Re: Not a VMS question but a questions for VMSers

Wim,

I think Bob's on the right track. To me the situation comes down to
- capability
- continuity of service
- performance
- data security
and
- control

I've not used it but my gut-feeling is that cloud computing could be used effectively within the one company, utilising IT resources scattered around the country or the world, and with some degree of internal company control over responsibilities and operating procedures.

I really don't like the idea of handing the data and processing to total strangers that you have virtually no control over.
Lucifer Megacruel
Valued Contributor

Re: Not a VMS question but a questions for VMSers

Hi,

Cloud computing is a new marking fad :P. There is nothing tremendously innovative. For me at least it does not really provide more than what traditional systems provide. The only advantage is that it makes lot of things easier for the developer , lot of deployment related activities , setting up machines etc ... taken care by the "cloud". It is just transfer of responsibility from one place to another by providing a slightly more convenient interface.

--Lucifer
"To Denouce the Evils of Truth and Love. To cause may hem and destruction and necromancy , Lucifer is here"