what did you learn today? (part 2)
- Thread starter continuum
- Start date
Ah, so the majority of the airflow in there is from the crack under the door? That may be another reason for the overheating.
According to my colleague who is actually working on that travesty the inlet temperatures are okay so it must be getting fairly decent ventilation.
Those Rx40s are such workhorses they probably wouldn't care.
Our old R720xd was also a beast. Only failure ever was one or two hard drives.
Apparently if you pay top dollar for Palo Alto gear and buy NBD support and open a support case at 5:15pm EDT it's really too late for them even though business hours for support are allegedly 9am-6pm and you won't get your replacement gear until the second next business day.
(but if you want to buy 4 hour premium support you could get it faster next time!)
(but if you want to buy 4 hour premium support you could get it faster next time!)
If you read the TOS of any support like this the NBD refers to after the dispatch occurs, not from the moment you contact them.
Oh I'm sure they're well within the letter of their TOS - just sucks to have an actual person tell you stuff will be there the next day then walk that back behind some bullshit after you've already lined up travel, access, etc.
I'm sad about what's happening with PAN, between their support getting worse and a dramatic increase in code quality issues. And Fortinet's no better (and arguably worse) so there's nobody to switch to unless you're a masochist and wanna deal with the dumpster fire that is Cisco.
Actually I'm entirely wrong here. The case was opened at 10:15am, we were told at 5:15pm that new hardware would be there the next day.
It has been a rough year for sure. It's increasingly hard to justify paying premium prices for this kind of service. The end game for us is probably public cloud and no more PA. That comes with its own set of problems but at least I don't have to care about when hardware gets shipped (or doesn't).
You can run a surprising amount of seevices on a 4 node kube cluster. Firewall / HA ingress: built in on cloud. Geographical spread / disaster recovery: covered by multi zone cluster, one node in a, b, c and d. Database replication: purchase it as a service instead of rolling your own. Backups: buckets with retention rule set.
If you have services built for it, it's cheaper than attempting to DIY renting space in multiple locations and interconnects between them. Only thing stupid expensive is egress, used to have many TBs included in the rack, now paying €€€ per GB, but not a lot of GB fortunately.
If you have services built for it, it's cheaper than attempting to DIY renting space in multiple locations and interconnects between them. Only thing stupid expensive is egress, used to have many TBs included in the rack, now paying €€€ per GB, but not a lot of GB fortunately.
Maybe it's different on this side of the pond but colo racks here almost always include unlimited transfer.
Speaking of, does anyone have any favorite colo providers in NC? Probably only need a quarter cab, and being on-net w/ Lumen is important.
What part of NC? We used Immedion, now DartPoints, here in SC who have a datacentre in Asheville. I don't know what all it has there though.
We're already making plans to move our ERP to SaaS in the next 5 years, but we're never going to be able to move everything else to the cloud too -- just too many niche servers that require being on-prem for various reasons. So it'll cut my virtualization footprint in half, which is nice, but I'll never get to pare things down to just routing/switching with all the traffic going to the Internet.
We were looking around Charlotte I believe but that could work too. Thanks.
Believe me I wish it were different but when they Bought L3 there wasn't (and still isn't really) much choice.
I asked someone I know in Charlotte: Tier point and Flexential come to mind immediatley
The middle ground is tough. Reducing your virtualization footprint by half doesn't reduce your support burden by anywhere close to half - you still need to deal with firewalls, switches, SANs, etc. And then you have to manage cloud/SaaS on top of that. For a sufficiently big company this is probably manageable, but it's really tough at a smaller outfit like ours where you simply don't have the bodies to cover all those areas of expertise AND handle the strategic work of moving things into cloud providers.
It's difficult to argue for anything other than "cloud for things that need it, rest on prem" or "everything cloud".
So long as you have something on prem, you're eating costs for that anyhow so you may as well use it. There are certainly some things that are better off cloud based, but there's little point in trying to transfer half your stuff to the cloud. You just won't see the benefit.
Our main application is a bit of a beast and can't really run in cloud. So long as we have to have a stretch cluster to handle that, it's easier to add a few nodes each site to handle everything else than try to cloudify everything we can.
So long as you have something on prem, you're eating costs for that anyhow so you may as well use it. There are certainly some things that are better off cloud based, but there's little point in trying to transfer half your stuff to the cloud. You just won't see the benefit.
Our main application is a bit of a beast and can't really run in cloud. So long as we have to have a stretch cluster to handle that, it's easier to add a few nodes each site to handle everything else than try to cloudify everything we can.
If half your stuff in the cloud means you can close a facility or get rid of colo racks it can be a massive savings in cost. For most, the cost of the VMWare/Broadcom licensing is the cheapest part of all of this.
Last edited:
If you can close a facility, you probably never needed it. I worked at a place that had at least a dozen, ranging from proper DC to a closet with a raised floor. But it was mainly due to acquisitions and the solution was to move everything to the main locations and shutter the rest. Then look at moving some things to cloud. (Although said employer decided to be a cloud provider instead and buck the trend)
Never is a long time. What you needed 10, 20, 30 years ago may be very different from today. My current org has been around since 1852. What we needed in 1980 and what we need in 2024 is pretty radically different.
DR can be a great starter target. We are contractually obligated to have a geographically distant DR site. In the on-prem world this means a colo with a bunch of expensive equipment that doesn't get much actual use (which is mostly a good thing!) and that is a pain in the ass to visit in person. Replacing this with cloud-based DR means we will have much lower fixed costs (mainly connectivity, redundant storage, backups, etc) and can quickly ramp up compute if we need to.
Our main datacenter was built in 1990 and we've got this sweet video they made about it, full of synth music and stylin outfits. I'm confident that our needs are the same 
What's the threshold for legal claims against Broadcom for failing to provide agreed upon services?