IT Employment

Ask something
Discussion Locked

Best practices advice for Data Center

0 helpful
5 replies

by nathanh 1 view

Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.

Please give me some suggestions if you have experienced this decision in the past. We have the choice in our budget to either shore up some redundancy in our local data center, by adding an additional T-1 and load balancing it, putting in a SAN with active/passive dB, and adding redundant switches and routers with HSRP (plus various other NSPOF features).

OR

We could beef up another location with the same current set up here. We will NOT globally load balance though; it would only be used for disaster recovery. (We do have a _minimal_ system for disaster recovery at this site now.) The new system would basically be the same as what we currently have in the Primary DC.

So the argument basically boils down to increasing redundancy in one location or adding more to our disaster recovery site?

5 replies

by timwalsh

All the on-site redundancy in the world would not help you in a World Trade Center type of disaster scenario. If your present data center were to become totally non-functional due to a large-scale disaster (man-made or natural), how long would it take for your present back-up site to resore operations to your present level?

You need to do a detailed risk analysis based on your location (likelyhood of a natural disaster), your visibility (likelyhood of a terrorist attack), and criticality ofdata/processes (time requirement for restoring operations) before spending significant amounts on either option.

You may find that your best bet would be to spend SOME money on both options instead of either/or.

by nathanh

It would take as long as your DNS refreshes after we changed the entry (we use short ttl's for this reason) or immediately if you use the alternate path to the back-up site as we direct our customers in an emergency. The data is replicated every 10minutes, so there could be a loss up to 10 mins. The current backup site has 100% capability to support usage, but basically has no redundancy. So, the question is should we bulk up the redundancy for longer term availibility in the Backup Data Center or eliminate more failure possibilities in the Production Data Center? For arguments sake, lets say the answer can not be both, though I can understand why one might recommend this approach.

by lordinfidel

You have to look at some different areas.

Are you a 24/7/365 shop that's entire business depends on people getting to your sites?

If so, redundancy at your local data center may not be the "best" answer.

Sounds odd I know. But your best answer for that scenario is a co-location facility that sit's across Multiple Providers (not just a att, sprint or bell).

When we found that having multiple connections into our facility was not enough, we had to switch to a colo facility.

The problem that we faced was, Our primary circuit went down and our back up circuit was not reliable. This combination was/is disastrous.

So we went with the "Highest Availablity" possiblity by co-locating at an Internap facility.
Which basically sits on Most of the tier1 providers.

Then we have private lines to the colo from our facility, plus multiple backup circuits coming into our facility in the off-chane that our pvt lines go down. We also off-loaded all user traffic onto their own t1's.

We also went a step further and made sure that our circuits used different media; we have fiber and copper coming in. Which reduces the risk of something getting cut taking all of our connections out.

You can never have 2 much redundancy. It boils down to the "What happens to your business if you go down"

If you are a small company that just has a website for informational purposes and the core of your business is on your local lan. Then take as much precautions at the local lan.Off-site storage, redundant servers, halon fire supression, etc.

If your business is totally dependant on people getting to you, then you need to take the "High Availablity" route.

by nathanh

Let me ask you a follow-up question to your response that is more to the heart of my question, would you move the Production Data Center to a Colo (which I didn't say ours was not) and build in additional redundancy (more $) or spend the money on a better disaster site.

If you like, lets say it like this:
Is is better to have?
ProDC 99.9% / DR DC 99.9%
OR
ProDC 99.999% / DR DC 95%

`Nathan

by lordinfidel

I would concentrate on the colo side.

Generally, A *good* High availabilty multi million dollar co location site will have lot's of precautions already in place.

(mulptiple power grids, generators, fire proof, electronic survellance, controlled access, etc)

Depending on the size of the org, the DR plan will be different.

Some companies approach this with the thinking of absolute natural disaster type of destruction and colo in several georgraphically different locations.

Others look at it from a data standpoint and will replicate the production data back to corporate or to another colo facility.

But your main line of defense right now is to get the redundancy built into your colo now. That is your most vulnerable spot. It's more likely that your server/router/switch will go down than it is a hurricane blows thru and destroys the center.

This discussion is closed to new replies.