Network restructuring advice
-
I would rebuild that CentOS 5 machine from scratch while doing this. Don't migrate it. Time to be on CentOS 7.
-
CentOS 5 is roughly like being on Windows Server 2003 R2.
-
@scottalanmiller said:
@whizzard said:
So with a backup host in a co-location site and better email server software....
What's better than Zimbra?
MDaemon?
-
@whizzard said:
MDaemon?
Not a bad product and really nice guys, we know them well (we hang out in their offices and go drinking with them.) But it's a small, SMB focused Windows app not on the same level as Zimbra. MDaemon makes a ton of sense for people needing to run email in house who can't manage Linux machines and are stuck on Windows and don't need the complexity of Exchange. But you'd be moving from top end enterprise to SMB focused and from free to not free. You'd be moving away from "makes sense to run in house" in every way.
-
Zimbra, for example, has full HA built in.
-
@scottalanmiller beating the dead horse of in house email is like @scottalanmiller beating the dead horse of RAID 5 on spinning rust. It is something you learn to listen to in the proper context.
The world does not run in the nice pretty black and white that his arguements for the subject try to make them. Yes, he is 100% technically correct and it IS the job of IT to recommend the best solution for the business.
But it is not our job to run a point into the ground once a business makes a decision. This does not mean that the business is going to go under. In fact the costs clearly show that moving a from EXISTING in house email to a hosted solution is expensive. Until such a time that an email server is needing to be replace or significantly changed the cost in employee time on top of the new monthly costs for the hosted service typically do not balance out.
Yes, you need to factor in what expected labor costs misc. down times would have cost. Yes their are other costs that begin to go away over time. Yes, the advantages of hosted are large. No that does not mean that every business should just move all their shit right now.
-
@JaredBusch said:
But it is not our job to run a point into the ground once a business makes a decision. This does not mean that the business is going to go under.
Which I did not suggest it was. But I was pointing out that hosted should be considered and probably recommended, that there are reasons to be in house and that the reasons he was wondering about for avoiding hosted were the opposite of the reasons you would use to move in that direction.
In no way did I paint a gloom and doom scenario, only that one solution appears to be the clear winner in both the current and proposed cases. No matter how reliably one solution wins as "best" does not imply that second place is dysfunctional (RAID 5 is the same way, there are plenty of cases where it is not a disaster even though there are none where it would be deployed following good analysis.)
The idea here was simply that he was feeling that he might have factors that make hosted not ideal, but those factors are generally not resulting in the direction he was envisioning.
-
@whizzard said:
SSD
@scottalanmiller said:
@whizzard said:
Is it advisable to mix HDD types in a RAID? i.e. SSDs with non SSDs?
It is not advisable to mix drives in any way if possible. You want absolutely identical drives in every way for optimum value, wear and performance. Even two drives of the same basic specs (size, spindle speed) is not as good as totally identical drives. You want all spindle and arm movements to align whenever possible.
I assumed it was a best practise to mix drive brands or models for HDDs in case of manufacturer defects, all you drives wouldn't suffer from the same issue which otherwise may make them likely to fail at once. I thought the firmware issues and other shortcoming of SSDs would have cause for the same rational.
-
@whizzard said:
I assumed it was a best practise to mix drive brands or models for HDDs in case of manufacturer defects, all you drives wouldn't suffer from the same issue which otherwise may make them likely to fail at once. I thought the firmware issues and other shortcoming of SSDs would have cause for the same rational.
Neither is the case. All drives should match. The idea of mixing has been one of those enduring myths of the small business market. No enterprise IT shop does anything like this and no enterprise vendor supports it. What they sometime encourage, and makes sense, is mixing order dates and manufacturing runs for this reason, but never drives of different models or manufacturers. As it is, drives do not die all together as often imagined as long as they aren't manufactured in the same run.
The problem here is that there is a myth of drives having a shared failure rate which has been used as the basis of a myth around how to mitigate that imagined risk.
-
@scottalanmiller said:
@Dashrender said:
What are you thinking about for backups? Once you're completely virtualized you could use Veeam, or you could look at getting a Unitrends box (though be prepared for several thousand dollar bill on that one). It doesn't appear that you can use the Drobo as a backup target, as you mentioned it's already being used for live data.
Unitrends does not require an appliance. He could run it as a VM on HyperV or XS and backup to the Drobo.
Doesn't mean the cost would be much less. Last time I looked, their licensing was pretty stiff. Of course, if the free version would work for him.... he might be set...
-
@Dashrender said:
Agreed, must of us can't. But even so, years of uptime make this a hard sell at best, impossible at worst.
It should cause no problem at all. It's the seatbelt issue. You simply need to explain that getting lucky is very different than being reliable. You don't drive without a seatbelt for a few years and claim that wearing one isn't needed despite years of not going through the windshield, right?
No remotely rational adult is going to argue with reliability over getting lucky if it is presented properly.
-
@scottalanmiller said:
@Dashrender said:
Agreed, must of us can't. But even so, years of uptime make this a hard sell at best, impossible at worst.
It should cause no problem at all. It's the seatbelt issue. You simply need to explain that getting lucky is very different than being reliable. You don't drive without a seatbelt for a few years and claim that wearing one isn't needed despite years of not going through the windshield, right?
No remotely rational adult is going to argue with reliability over getting lucky if it is presented properly.
No one's saying they are going to argue over the rationality, but I've seen many continue with the "we're lucky" mindset.
-
@Dashrender said:
No one's saying they are going to argue over the rationality, but I've seen many continue with the "we're lucky" mindset.
Isn't that arguing in a way?
So you are saying that they know that they are just lucky, know that they are being risky and do it anyway when they have a clear understanding that they are causing risk?
-
@scottalanmiller said:
@Dashrender said:
No one's saying they are going to argue over the rationality, but I've seen many continue with the "we're lucky" mindset.
Isn't that arguing in a way?
So you are saying that they know that they are just lucky, know that they are being risky and do it anyway when they have a clear understanding that they are causing risk?
That's one way to look at it - another is that the expense of going hosted isn't worth it to them, especially when there is no hardware costs associated with locally hosted because you stand up another VM on an existing host. If you're talking a stand alone email in terms of hardware, then you're probably rarely/never getting any push back going hosted.
If you need only email and no shared calendars, something like Rackspace $2/month/user (normal price - what one of my customers is being billed through NTG) it definitely gets hard to locally host, but even if you need shared calendars, O365 at $4/m/u does add up quickly.
-
@Dashrender said:
That's one way to look at it - another is that the expense of going hosted isn't worth it to them, ....
That's unrelated. Don't mix a discussion around understanding risk with valuing risk. That's completely different. If email isn't seen as that important that's absolutely fine, for many businesses it is not. But that's wholly different than feeling that they've gotten lucky and ignoring rational thought. In one case, they lack logical thinking and feel that being lucky is better than being reliable. In the other they don't see reliability as valuable. Completely different concepts.
-
@Dashrender said:
If you need only email and no shared calendars, something like Rackspace $2/month/user (normal price - what one of my customers is being billed through NTG) it definitely gets hard to locally host, but even if you need shared calendars, O365 at $4/m/u does add up quickly.
Doesn't Rackspace email have shared calendars?
Edit: Sure does. I was sure that we had always used those.
-
OK, the shared calendar is news to me - not that my client wants it. I asked, they say 'meh'.
Though I have no idea how that shared calendar works when it comes to locally installed Outlook, not the web portal - and does it support the sharing of a person's calendar or is it just a shared calendar for the company?
-
@Dashrender said:
OK, the shared calendar is news to me - not that my client wants it. I asked, they say 'meh'.
Rackspace Email is full featured. Not aware of any enterprise feature that it doesn't have. That's always been their thing, all the features at a much lower price for people not addicted to Exchange itself.
-
@Dashrender said:
Though I have no idea how that shared calendar works when it comes to locally installed Outlook, not the web portal - and does it support the sharing of a person's calendar or is it just a shared calendar for the company?
Normal shared calendars. It's actually the other way around, ALL of the functionality of Rackspace Email is in the web client, not Outlook. It's a web-native service, that isn't a second class interface, Outlook is the second class service here. OWA is starting to replace Outlook as the main interface for Exchange too, especially now that services like Clutter can only be controlled in the web client.
So yes, individual shared calendars. We were doing this all of the time. The feature goes back at least to 2009. Probably long before that.
-
Good to know.
Hmmm... While I appreciate the anywhere access to systems like this, there are times I just prefer an app instead of a web interface - frankly sometimes if only because the icon in the start menu is dedicated to function (though I think on Windows now, if you launch a site from a .website shortcut it does create it's own icon while running on the start bar).
