amuck-landowner

RamNode OpenVZ SSD-Cached 256MB (ATL)

wlanboy

Content Contributer
Provider: RamNode
Plan: OpenVZ SSD-Cached 256mb VPS
Price: 12.75$ per quarter
Location: Atlanta, GA

Purchased: 04/2013

I did a review on the SSD OpenVZ in Seattle and the the SSD OpenVZ in NL offers too.

Hardware information:

  • cat /proc/cpuinfo

    processor : 0
    vendor_id : GenuineIntel
    cpu family : 6
    model : 45
    model name : Intel(R) Xeon(R) CPU E5-2630 0 @ 2.30GHz
    stepping : 7
    cpu MHz : 2300.069
    cache size : 15360 KB
    physical id : 0
    siblings : 12
    core id : 0
    cpu cores : 6
    apicid : 0
    initial apicid : 0
    fpu : yes
    fpu_exception : yes
    cpuid level : 13
    wp : yes
    flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm dca sse4_1 sse4_2 x2apic popcnt aes xsave avx lahf_lm ida arat epb xsaveopt pln pts dts tpr_shadow vnmi flexpriority ept vpid
    bogomips : 4600.13
    clflush size : 64
    cache_alignment : 64
    address sizes : 46 bits physical, 48 bits virtual
    power management:

  • cat /proc/meminfo
    Code:
    MemTotal:         262144 kB
    MemFree:          174268 kB
    Cached:            34820 kB
    Active:            23232 kB
    Inactive:          57364 kB
    Active(anon):       8300 kB
    Inactive(anon):    37476 kB
    Active(file):      14932 kB
    Inactive(file):    19888 kB
    Unevictable:           0 kB
    Mlocked:               0 kB
    SwapTotal:        262144 kB
    SwapFree:         260044 kB
    Dirty:                 0 kB
    Writeback:             0 kB
    AnonPages:         45776 kB
    Shmem:              3548 kB
    Slab:               7272 kB
    SReclaimable:       4220 kB
    SUnreclaim:         3052 kB
  • df -h
    Code:
    Filesystem      Size  Used Avail Use% Mounted on
    /dev/simfs       90G   32G   59G  35% /
    none            128M  4.0K  128M   1% /dev
    none             26M  992K   25M   4% /run
    none            5.0M     0  5.0M   0% /run/lock
    none            128M     0  128M   0% /run/shm
  • dd
    Code:
    dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync && rm -rf test
    
    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 1.8499 s, 580 MB/s
  • wget
    Code:
    wget cachefly.cachefly.net/100mb.test -O /dev/null
    
    Resolving cachefly.cachefly.net (cachefly.cachefly.net)... 205.234.175.175
    Connecting to cachefly.cachefly.net (cachefly.cachefly.net)|205.234.175.175|:80... connected.
    HTTP request sent, awaiting response... 200 OK
    Length: 104857600 (100M) [application/octet-stream]
    Saving to: `/dev/null'
    
    100%[================================================================================================================================>] 104,857,600 52.0M/s   in 1.9s
    
    2013-05-19 02:37:26 (52.0 MB/s) - `/dev/null' saved [104857600/104857600]
What services are running?

  • MySQL
  • Backup space

Support:

Well I did not need support yet. But I am using this vps for only one and a half month.

Overall experience:

You get a lot for what you pay. I am a happy customer. I did not trust that "we are fast" but the vps does have a fast I/O. Altlanta has got good routings to Buffalo, NY, Chicago, Austin, etc.

I use it for my own hosted "offloaded" MySQL server and for my encrypted backups of the other vps.
 
Last edited by a moderator:

Oxide

New Member
We've got a KVM VPS with Ramnode for a PowerDNS server. Never had to contact support either. Solid stuff!
 

drmike

100% Tier-1 Gogent
Thanks for this one @wlanboy.

Nick runs  some seriously fast servers.  Disk IO  is super.  

I use RamNode in both their locations and haven't had a bad day with them.
 

KuJoe

Well-Known Member
Verified Provider
<3 RamNode! I have a VPS that I don't use but I know if I ever cancel it I'll need it so I keep it around just in case. :)
 

Tux

DigitialOcean? lel
I've only contacted support only a few times and those were things I couldn't do myself. :)

Also, the SSD nodes have wicked I/O:

Code:
tux@disco:~$ dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync && rm -rf test
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 1.52441 s, 704 MB/s
 

acd

New Member
Ran some tests on one of my RamNode ovzs. This one is on atlcvz5 (a CVZ-E3 node), which, from what Nick says, ramnode is phasing out.

wget cachefly:


root@atl:~# wget cachefly.cachefly.net/100mb.test -O /dev/null
--2013-06-10 15:37:47-- http://cachefly.cachefly.net/100mb.test
Resolving cachefly.cachefly.net (cachefly.cachefly.net)... 205.234.175.175
Connecting to cachefly.cachefly.net (cachefly.cachefly.net)|205.234.175.175|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 104857600 (100M) [application/octet-stream]
Saving to: `/dev/null'

100%[======================================>] 104,857,600 55.1M/s in 1.8s

2013-06-10 15:37:49 (55.1 MB/s) - `/dev/null' saved [104857600/104857600]

dd write test:


root@atl:~# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync && rm -rf test
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 6.88498 s, 156 MB/s

beancounters:

root@atl:/proc# cat user_beancounters
Version: 2.5
       uid  resource                     held              maxheld              barrier                limit              failcnt
     3652:  kmemsize                  6822485             28995584            134217728            134217728                    0
            lockedpages                     0                  525                32768                32768                    0
            privvmpages                 15654                39424  9223372036854775807  9223372036854775807                    0
            shmpages                     1545                 3625  9223372036854775807  9223372036854775807                    0
            dummy                           0                    0                    0                    0                    0
            numproc                        29                   69  9223372036854775807  9223372036854775807                    0
            physpages                    8728                65529                    0                65536                    0
            vmguarpages                     0                    0  9223372036854775807  9223372036854775807                    0
            oomguarpages                 4526                 6418  9223372036854775807  9223372036854775807                    0
            numtcpsock                     15                   26  9223372036854775807  9223372036854775807                    0
            numflock                        4                   10  9223372036854775807  9223372036854775807                    0
            numpty                          3                    8  9223372036854775807  9223372036854775807                    0
            numsiginfo                      0                   42  9223372036854775807  9223372036854775807                    0
            tcpsndbuf                  263912               477552  9223372036854775807  9223372036854775807                    0
            tcprcvbuf                  245760              4464912  9223372036854775807  9223372036854775807                    0
            othersockbuf                36992               125256  9223372036854775807  9223372036854775807                    0
            dgramrcvbuf                     0               145656  9223372036854775807  9223372036854775807                    0
            numothersock                   52                   82  9223372036854775807  9223372036854775807                    0
            dcachesize                4144518             25840639             67108864             67108864                    0
            numfile                       455                  855  9223372036854775807  9223372036854775807                    0
            dummy                           0                    0                    0                    0                    0
            dummy                           0                    0                    0                    0                    0
            dummy                           0                    0                    0                    0                    0
            numiptent                      25                   25  9223372036854775807  9223372036854775807                    0
Interesting things to note:

  1. RamNode doesn't cap proc limits per vm. I assume they have other methods of mitigating crazy forkings in a vm from taking down a whole HN. I did not test this. (a max of 4-5k procs can be expected from the kmemsize limit, but that's a LOT of procs)
  2. Only about half your memory can be used for mlock, If you don't know what this is, you probably don't care.
  3. The amount of disk cache the kernel will hold for the OVZ is limited to 64MiB. On my other ovz providers, this is left to "fill unused ram". Granted, it doesn't appear I've ever hit this limit but I generally only use this ovz for openvpn.

openssl aesni is enabled:


root@atl:/proc# openssl speed -evp aes-256-cbc
Doing aes-256-cbc for 3s on 16 size blocks: 80523557 aes-256-cbc's in 2.95s
Doing aes-256-cbc for 3s on 64 size blocks: 21962813 aes-256-cbc's in 2.99s
Doing aes-256-cbc for 3s on 256 size blocks: 5542420 aes-256-cbc's in 2.99s
Doing aes-256-cbc for 3s on 1024 size blocks: 1397454 aes-256-cbc's in 2.99s
Doing aes-256-cbc for 3s on 8192 size blocks: 175106 aes-256-cbc's in 2.99s
OpenSSL 1.0.1 14 Mar 2012
built on: Tue Mar 19 19:15:10 UTC 2013
options:bn(64,32) rc4(8x,mmx) des(ptr,risc1,16,long) aes(partial) blowfish(idx)
compiler: cc -fPIC -DOPENSSL_PIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security -D_FORTIFY_SOURCE=2 -Wl,-Bsymbolic-functions -Wl,-z,relro -Wa,--noexecstack -Wall -DOPENSSL_NO_TLS1_2_CLIENT -DOPENSSL_MAX_TLS1_2_CIPHER_LENGTH=50 -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM -DVPAES_ASM -DWHIRLPOOL_ASM -DGHASH_ASM
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-256-cbc 436737.94k 470107.03k 474534.96k 478592.94k 479755.30k


That's about as good as it gets for aes. =) Less useful than on KVM w/ dmcrypt, but I suspect the option is available on kvm as well.

I should note that Nick offered to migrate me off this older node onto a newer E5 based node to speed up disk writes, which I declined, as this VM doesn't see heavy disk usage.
 

eva2000

Active Member
nice review, i have 3x ramnode VPS as well and they definitely perform better than my 4x buyvm VPS. I simultaneously updated Nginx and php via Centmin Mod shell based menu on all 7 VPSes at the same time. All 3x ramnode VPS finished way ahead of time compared to my buyvm VPS. But buyvm has benefits such as internal IP/network usage which I use. Sometimes wish ramnode has such option too :D
 
Last edited by a moderator:

eva2000

Active Member
Are your Ramnode VPSes with Pure SSD or SSD Caching?

I think BuyVM uses SSD Caching for OpenVZ VPSes.
all my Ramnode VPS are either OpenVZ SSD Cached (2x VPS) or KVM SSD Cached (1x VPS) :)
 
Last edited by a moderator:

acd

New Member
RamNode uses ssd write-back caching whereas BuyVM uses ssd read caching (write-through). This gives RamNode a significant (like 3-4x, depending on the SSD used for caching) write speed advantage. I've been told the reason BuyVM doesn't use write-back is they were noticing problems if the kernel hung, it might leave the backing storage in an inconsistent state. This might not be an issue with a RAID controller supporting ssd write-back cache on-card.

edit: added clarification on what that does. Short version, write-back means write to cache then return, controller later flushes cache to disk as IO becomes available, usually in LRU order w/ maxage; write-through means write to both cache AND disk synchronously before returning.

You can read more about cache write policy on wikipedia if you're interested. Wikipedia: Cache#Writing Policies
 
Last edited by a moderator:

eva2000

Active Member
RamNode uses ssd write-back caching whereas BuyVM uses ssd read caching (write-through). I've been told the reason BuyVM doesn't use write-back is they were noticing problems if the kernel hung, it might leave the backing storage in an inconsistent state. This might not be an issue with a RAID controller supporting ssd write-back cache on-card.
cheers learn something new today :)
 

wlanboy

Content Contributer
Time to update my review.

What services are running?

  • MySQL master
  • MongoDB cluster node
  • Backup space
Support:

Not a single support ticket needed.

Overall experience:

I really enjoy my vps in Atlanta. No hassles, no downtimes, no support needed and fast and solid routings to europe.

It is my US MySQL master getting hit by 4 wordpress blogs and 3 php based galleries. And my US based MongoDB cluster node getting hit by my Twitter archive and my ping database.

Yes it is OpenVZ - but not oversold:


free -m
total used free shared buffers cached
Mem: 256 224 31 0 0 69
-/+ buffers/cache: 154 101
Swap: 256 88 167


Additionally it is collecting the backups of all of my vps and rsyncing it to my local NAS.

I prefer to have my backups located at two different continents.
 

wlanboy

Content Contributer
Want to add the current status report of my vps:

statusrna.JPG


2 hours and 3 minutes of downtime since April the 17th.
 

Pete M.

New Member
Verified Provider
Those benchmark numbers look really good, pretty much in line with I have grown to expect from SSD cached nodes :)
 
Top
amuck-landowner