Web Hosting Talk : Web Hosting Main Forums : VPS Hosting : Krypt VPS Review - Few months...

[Ron D]

I've had a Krypt VPS for a few months now, started with the VPS400 package and moved up to VPS800 over time. Initially (I'm guessing they had just gotten into providing VPS's) the service was EXCELLENT! My VPS was located in their LA datacenter on the Level3 backbone. First two months the VPS was rock solid, I was running a full LAMP (no optimizations) install along with Webmin. By the 3rd month I started having issues with sluggishness, both on the site and SSH. Put in a support ticket which got responded to within 10 minutes as usual (fantastic support!) and their techs moved my VPS to a less loaded VMware node. Things were again pretty solid for about two weeks, then the same thing happened again. I let it be for a few days thinking it may just be a spike, but even after a week it was still pretty horrid. I had even turned off the LAMP install and Webmin in favor of a lone Lighttpd install to serve static html pages. Mind you by this time I had been upgraded to above their VPS800 package, yet I was still having performance issues. Put in another ticket and asked to be moved to another node, they said sure and asked a timeframe, I told them asap and let the issue be. After 4 hours of not hearing anything back, I update my ticket to find out that they were having some issues with the VPS, ok bad things happen, give them time. A few hours later I ask again and am told that they formatted the VPS without any authorization from me...Its a VMware host, they could at the least have attached the virtual hard drivei mage to the new install of my VPS, but they prefered to just delete it. I run several ESX servers at work and even with a nonbootable (according to them) image, the hard drive can still be mounted and recovery can be attempted. I was angry but I have a few dedicated servers with Krypt so I gave them another chance. I took the freshly installed VPS and started hunting down backups to restore from. Installed LAMP and Webmin, gave it a week and once again: slow. I'm fairly certain that CentOS repo apps and Webmin arn't enough to cause a VPS to dog down this bad, even with *everything* turned off (top shows 24MB ram usage) SSH and pages being served by lighttpd are taking 3-10 seconds to come up, as before they came up before I even let go of the button.
Ratings
Support: 10/10
Service: 6/10
VPS Quality: 5/10
Network: 5/10 (great in the beginning, after a week less than 400KB/s)
Overall they're a great provider, but they need to monitor their nodes more carefully and find abusers. It seems to me like they're either overloading their nodes or have something misconfugured thats causing issues. I was told a few times to restart my VPS which didn't help at all. Orderded a VPS with them about 2 weeks ago on a different account, havn't even touch it since I got it, as of right now, SSH is sluggish, apps like top take an average 5 seconds to come up, max throughput is ~300KB/s from Seattle (hit ~7MB/s when I first got it). Same issue as my other VPS. I'll provide my domain to the mods for authentication.

[RavC]

Hi thanks for your review,
Did you run unixbench on your VPS to test performance? Unfortunately I can't link to it directly argh.. search this forum for the download.

[Ron D]

I did, I got a score of 28, which compared to other packages (even low end ones) is pretty horrid. My new VPS with the same specs has a score of 164, go HostChum!

[drumman24]

28 is very low indeed.

[foobic]

I'm always a bit wary of unixbench scores on a VPS. If it scores high then it's being allowed to burst its resource limits - nice in the short term but potentially a problem once the node fills up if it's not managed and controlled properly, like when some other user's busy VPS grabs all available resources and slows down everyone else, including you! OTOH, if a VPS scores low then it might just mean that the host has them all securely locked down, leading to a stable and reliable experience for all. Or not...
Real world experience is a better indicator, IMO, so thanks Ron for the detailed review. Another easy test on a machine that seems slow is a simple "time ls" - eliminates the possibility of network issues and can give interesting (scary) results.

[Ron D]

Still have the VPS (paid for until the end of the month)...here's the time ls output:
real 0m0.818s
user 0m0.703s
sys 0m0.814s
New VPS:
real 0m0.002s
user 0m0.000s
sys 0m0.002s
Equal amount of files in the /home (rsynced). Have noticed this trend among VMware VPS's (run a few ESX boxes at work)...if a VPS isn't accessed for a while, it gives up its resources to other VPS's on the box and goes into a warm standby mode. On first connect the VPS can take upwards of 30 seconds to go from warm to hot, once its good it'll last until the "burst" runs out, which on Krypt seems to be about a minute long.
The unixbench scores are performed right after I make the VPS do some heavy lifting (rsyching over ssh, tarring, untarring, multiple scripts, etc).

[tritium]

I've been a customer of Krypt VPS for the same time frame, though I've been using a Windows VPS. To that end, I've had precisely the SAME experience with Krypt and sluggishness -- I've even been OFFLINE for 24-48 hours each month while Krypt has worked on my VPS.
They even moved me to a new node too, just like the original poster. Just like in his scenario, they didn't move my IIS settings or my web data, so I had to reconfigure absolutely everything. Then, I realized that I had a new IP, so I had to update my external DNS too.
At the moment, I'm dealing with them on the sluggishness issue, and they think its my route. Through my employer, I have a 10MB Level3 Ethernet drop (I can't host personal things here) and the route right up to my VPS looks good. The last hop to my box is typically running around 500-4500ms however, so that doesn't sound like a routing problem to me. It sounds like a misconfigured or overloaded box. My bet is on overloaded box.
Either way, I'm disappointed with the Krypt service, and will be writing my own review soon. For the record, here's a sanitized tracert from my local network.
-------------
Tracert Data:
-------------
Tracing route to customer.vpls.net [98.126.15.X]
over a maximum of 30 hops:
1 <1 ms <1 ms <1 ms 10.10.32.1
2 1 ms 1 ms 1 ms 192.X.X.X
3 2 ms 2 ms 2 ms X.hsa1.X.level3.net [64.X.X.X]
4 2 ms 2 ms 2 ms ge-7-0-X.mpls1.X.level3.net [209.244.22.1]
5 * 6 ms 4 ms ae-2-0.bbr1.washington1.level3.net [4.68.128.201]
6 5 ms 6 ms 5 ms ae-24-79.car4.washington1.level3.net [4.68.17.70]
7 5 ms 5 ms 6 ms 10ge-3-1-0.was10.ip.tiscali.net [4.68.110.98]
8 70 ms 70 ms 69 ms so-4-0-0.lax10.ip.tiscali.net [89.149.187.93]
9 76 ms 74 ms 74 ms vpls-gw.ip.tiscali.net [77.67.68.98]
10 74 ms 74 ms 75 ms vlan2099.br1.lax3.vpls.net [67.198.200.18]
11 1353 ms 227 ms 383 ms customer.vpls.net [98.126.15.X]
Trace complete.

[Ron D]

tritium, same results here. They tell me its my network, its my bandwidth carrier, its this or that when its either a heavily loaded box or their network. They've even kindly moved to to different bandwidth carriers in hopes of resolving the issue to no avail. I have a gigabit pipeline on Savvis from Seattle (I'm a datacenter tech in Bellingham) and I can hardly pull 13KB/s right now off my VPS.
Either through Cogent, Level3, Tiscali, etc. The last hop varies between 40ms (normal) and 8000ms (uhhh?). Can't even run two pings to the server without the latency spiking, much less have two people viewing any site on the box or downloading anything. 1 person, sure I get 3MB/s, as long as theres *nothing* else happening on the box. Soon as another download starts or a crawler decides to visit a site (mind you, the crawler is set on low priority), that speed drops like a rock!

[patord]

Just curious if both of you are maybe on the same vlan and possibly the same node?
It seems I am one of their newer E5450 nodes, and currently on the same vlan (2099). Traceroutes for me appear fine (from my servers in GNAX, PEER1, etc), no latency between the vlan hop and my node as indicated in the posts above. Again, I'm probably on a different node from you all.
Right now the performance of my VPS with them seems to vary. I too am running in to the initial process lag that one of you mentioned. Now if that is a function of our vCPU getting reassigned (or idled) due to their minimum processor guarantees I'm not sure. I have other ESX VPSs elsewhere with processor guarantees and do NOT see this kind of lag. However, once this lag is gone, I can burst to the max of a single pCPU and given I'm on a E5450 - that's pretty good.
My other issue is they advertise possible burst to 8ghz and to me, I would assume I would require at least two vCPUs to get anywhere close to that. My other ESX providers assign me multiple vCPUs to allow for X bursting - and I'm ok with that. There are plenty of debates about going this route due to VMWare's SMP handling. Just like there are plenty of threads on WHT of how the WHT Unixbench is NOT a valid way to benchmark a VMWare VPS.
Hope you gents can get things squared away with them.

[Ron D]

I'm not judging the performance based on unixbench alone, I never really judge anything based on a single benchmark application, real-world performance/benchmark trumps application based benchmarks any day. I was moved to their newer nodes a few weeks back, but as always, wasn't long before it and I were suffering again. cpuinfo shows E5450 @ 3.00GHz, so I am on their newer nodes. It starts off with varied performance issues, then a week later your VPS is struggling to keep up with a snail. By no means are Krypt as a company bad, theyre GREAT, but their VPS platforms configuration needs some optimization, or hey, they're probably overselling like everyone else
Edit: Where did you happen to see the vlan id? The only control panel I was granted access to was their Vault (previously another system).

[patord]

Ron- no worries on the benchmarking topics. I mainly run some php benchmark tools since that's what my sites are usually geared for. Those seem to be pretty good, but running PHP scripts is only part of what I need a VPS to do. The lagging/sluggishness overall "feel" of the VPS is more apparent as I've started to put more activity in to the server. Again, I have other ESX VPSs elsewhere and they seem to run smoother then Krypt's setup.
The vlan route usually appears in the hop prior to hitting your VPS IP during a traceroute. From what I've seen it depends on the ingress transit provider.
So have either of you received any resolution and/or responses from Krypt on your various performance and network concerns? I've emailed Tim a few times, but it seems discussions have stalled.

[Ron D]

Hi patord, the staff at Krypt are aware of the issues and currently don't know why. I have canceled my VPS with them and am going to transition everything over to either my dedicated servers with them or to another VPS provider. Last I checked my VPS had no actual services running--just SSH and a base Cent install. Although it was speedy at 2am, the lag issues were extremely apparent during normal hours. A simple zip of ~1MB txt files took about 5 minutes, the transfer (over http/apache) went ~3KB/s to my pc (on a gbit line at home).

[Ron D]

Received an email a few days ago that said the Krypt guys were looking into the issue and would be making some changes to the way their VPS services work to fix all the performance issues. They've offered me another month (free, gotta love the Krypt support!) in which time they hope to have everything mostly ironed out. I'll update my review at the end of my free month if things change.