I run an nginx webserver, PHP-FPM, XCache, and MySQL for some WordPress and CakePHP websites.

I run NSD DNS servers to manage DNS for my domains. I currently use 3 different VPSs from different providers for this.

I run a Postfix mail server to handle my personal email, and retrieve it via pop3s using Dovecot. A secondary MX runs on an alternate VPS in case the main VPS is down. I also use dkim-milter to automatically add dkim signatures to outgoing email.

I run OpenVPN for personal use.

I run Duplicity and rsync for backups. Duplicity automatically runs backups to alternate VPSs. My router running TomatoUSB and optware automatically rsyncs my VPSs to a flash drive in the middle of the night.

I wrote a script to do ping tracking and monitor various resources on the VPS. This runs on two VPSs, so they can monitor each other.

I wrote a script to capture a streaming radio show to mp3 files that runs automatically on a VPS. Then my router downloads them in the middle of the night and deletes them from the VPS. So I can listen at my convenience.

What do you use your VPS for?

I used CoreNetworks service for about 4 months (February – May, 2010).

I started with their “Starter LE26″ plan for $24.95/month and upgraded after a month to the “Starter LE16″ plan for $29.95/month.

To be honest, my needs are pretty modest, but I always thought it would be neat to have my own dedicated server, and with such a low price, I thought it would make more sense than a VPS.

The “Starter LE26″ plan for $24.95 a month had a $30 setup charge, so the cost for the first month was $54.95.

It came with a 10 mbps port, 1,000 GB of transfer, and 2 IP addresses.

It had a single 2.66 GHz Celeron CPU, 512 MB RAM, and an 80 GB SATA hard drive.  They installed CentOS for me as part of the setup.

I placed the order on a Saturday morning (East Coast time).  CoreNetworks is based in Lansing, Michigan and has their own data center.

Even though they don’t guarantee weekend installs, they had my server ready later that Saturday afternoon, after 7 hours 7 minutes.

I ran Unix Bench 5.1.2, which reported a score of 421.0.

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.2)
 
   System: GNU/Linux
   OS: GNU/Linux -- 2.6.18-164.11.1.el5 -- #1 SMP Wed Jan 20 07:39:04 EST 2010
   Machine: i686 (i386)
   Language: en_US.utf8 (charmap="UTF-8", collate="UTF-8")
   CPU 0: Intel(R) Celeron(R) CPU 2.66GHz (5334.2 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext
   06:09:59 up 2 days, 10:25,  2 users,  load average: 0.07, 0.02, 0.00; runlevel
 
------------------------------------------------------------------------
Benchmark Run: Fri Feb 05 2010 06:09:59 - 06:38:06
1 CPU in system; running 1 parallel copy of tests
 
Dhrystone 2 using register variables        4209962.7 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                      944.2 MWIPS (10.6 s, 7 samples)
Execl Throughput                               1800.4 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        185884.1 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           53783.0 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        561310.1 KBps  (30.0 s, 2 samples)
Pipe Throughput                              576675.4 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 119167.1 lps   (10.0 s, 7 samples)
Process Creation                               5758.8 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   2462.6 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                    339.4 lpm   (60.1 s, 2 samples)
System Call Overhead                         586443.4 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0    4209962.7    360.8
Double-Precision Whetstone                       55.0        944.2    171.7
Execl Throughput                                 43.0       1800.4    418.7
File Copy 1024 bufsize 2000 maxblocks          3960.0     185884.1    469.4
File Copy 256 bufsize 500 maxblocks            1655.0      53783.0    325.0
File Copy 4096 bufsize 8000 maxblocks          5800.0     561310.1    967.8
Pipe Throughput                               12440.0     576675.4    463.6
Pipe-based Context Switching                   4000.0     119167.1    297.9
Process Creation                                126.0       5758.8    457.0
Shell Scripts (1 concurrent)                     42.4       2462.6    580.8
Shell Scripts (8 concurrent)                      6.0        339.4    565.6
System Call Overhead                          15000.0     586443.4    391.0
                                                                   ========
System Benchmarks Index Score                                         421.0

Here’s a hard drive test I ran on the single 80 GB SATA hard drive:

# hdparm -t /dev/sda
 
/dev/sda:
Timing buffered disk reads:  218 MB in  3.02 seconds =  72.13 MB/sec

Reverse DNS was setup within 4 minutes of my request.

Support was always friendly and quick to respond.  Critical items were resolved quickly.  Non-critical items were responded to quickly, but sometimes took several days to actually resolve.  This was fine, since they were non-critical, afterall.

Initially, I found my bandwidth transfer speed was only around 3 mbps.  Support was quick to respond, but it took 5 days to actually resolve to remove a stale filter on my IP address.  They gave me an extra 5 days of service for free as a result, without having to ask.

After having the “Starter LE26″ server for a week, I changed my mind and decided to replace it with a “Starter LE16″ server instead.  I kept the first server for the remainder of the billing period so I had both online for a time so I could move things over.

The “Starter LE16″ plan has a newer 1.6 GHz dual core Celeron E1200 CPU.  It also comes with 512 MB RAM, an 80 GB SATA hard drive, a 10 mbps port, 1,000 GB of bandwidth transfer, and 2 IP addresses.  The main difference is the better CPU for an extra $5/month. It scored 959.7 with Unix Bench.

========================================================================
   BYTE UNIX Benchmarks (Version 5.1.2)
 
   System: GNU/Linux
   OS: GNU/Linux -- 2.6.18-164.11.1.el5 -- #1 SMP Wed Jan 20 07:32:21 EST 2010
   Machine: x86_64 (x86_64)
   Language: en_US.utf8 (charmap="UTF-8", collate="UTF-8")
   CPU 0: Intel(R) Celeron(R) CPU E1200 @ 1.60GHz (3199.9 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   CPU 1: Intel(R) Celeron(R) CPU E1200 @ 1.60GHz (3199.9 bogomips)
          Hyper-Threading, x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET
   08:02:31 up  7:46,  2 users,  load average: 0.17, 0.23, 0.19; runlevel
 
------------------------------------------------------------------------
Benchmark Run: Sat Feb 06 2010 08:02:31 - 08:30:25
2 CPUs in system; running 1 parallel copy of tests
 
Dhrystone 2 using register variables        8670696.1 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     1784.0 MWIPS (9.8 s, 7 samples)
Execl Throughput                               2111.4 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        327152.6 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          110430.8 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        644841.4 KBps  (30.0 s, 2 samples)
Pipe Throughput                              884142.7 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 193744.4 lps   (10.0 s, 7 samples)
Process Creation                               6126.0 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   3462.8 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                    719.0 lpm   (60.0 s, 2 samples)
System Call Overhead                        1117721.3 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0    8670696.1    743.0
Double-Precision Whetstone                       55.0       1784.0    324.4
Execl Throughput                                 43.0       2111.4    491.0
File Copy 1024 bufsize 2000 maxblocks          3960.0     327152.6    826.1
File Copy 256 bufsize 500 maxblocks            1655.0     110430.8    667.3
File Copy 4096 bufsize 8000 maxblocks          5800.0     644841.4   1111.8
Pipe Throughput                               12440.0     884142.7    710.7
Pipe-based Context Switching                   4000.0     193744.4    484.4
Process Creation                                126.0       6126.0    486.2
Shell Scripts (1 concurrent)                     42.4       3462.8    816.7
Shell Scripts (8 concurrent)                      6.0        719.0   1198.4
System Call Overhead                          15000.0    1117721.3    745.1
                                                                   ========
System Benchmarks Index Score                                         674.4
 
------------------------------------------------------------------------
Benchmark Run: Sat Feb 06 2010 08:30:25 - 08:58:19
2 CPUs in system; running 2 parallel copies of tests
 
Dhrystone 2 using register variables       17440340.8 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     3565.9 MWIPS (9.8 s, 7 samples)
Execl Throughput                               3504.5 lps   (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks        255648.9 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks           73393.8 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks        626257.6 KBps  (30.0 s, 2 samples)
Pipe Throughput                             1769598.9 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 447917.3 lps   (10.0 s, 7 samples)
Process Creation                               8704.3 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   5468.6 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                    755.2 lpm   (60.1 s, 2 samples)
System Call Overhead                        2099438.8 lps   (10.0 s, 7 samples)
 
System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   17440340.8   1494.5
Double-Precision Whetstone                       55.0       3565.9    648.3
Execl Throughput                                 43.0       3504.5    815.0
File Copy 1024 bufsize 2000 maxblocks          3960.0     255648.9    645.6
File Copy 256 bufsize 500 maxblocks            1655.0      73393.8    443.5
File Copy 4096 bufsize 8000 maxblocks          5800.0     626257.6   1079.8
Pipe Throughput                               12440.0    1769598.9   1422.5
Pipe-based Context Switching                   4000.0     447917.3   1119.8
Process Creation                                126.0       8704.3    690.8
Shell Scripts (1 concurrent)                     42.4       5468.6   1289.8
Shell Scripts (8 concurrent)                      6.0        755.2   1258.7
System Call Overhead                          15000.0    2099438.8   1399.6
                                                                   ========
System Benchmarks Index Score                                         959.7

The older 2.66 GHz Celeron CPU was perfectly fine for my needs.  I thought at the time I’d be keeping the server for a long time, though, and for an extra $5/month I thought it would be nice to have something a little faster.

The “Starter LE16″ plan had a $69 setup charge, so the first month was $98.95.

I also did a one-time upgrade to increase the RAM to 1GB and add a second 80GB hard drive.  I setup software RAID-1 myself.

CoreNetworks also includes free KVM access to the console on request, and I used this a few times.  It was hooked up promptly after emailing the request and I was able to do what I needed to do, and then let them know when I was finished with it.

There were a few network interuptions (probably DDOS attacks), but they were handled quickly.

So why did I leave?  After I began using OpenVPN on the server, I began to notice the network quality to my location (in the Philippines) was not the greatest.  While it seemed fine for web hosting, it was often not able to maintain consistent high transfer speeds, at least for the single connection that I needed for VPN use.

At one point towards the end, it became so bad that it was basically so slow as to be useless.  I contacted support, and as usual they responded very quickly, and after providing traceroutes, they said they’d have the networking team look into it.

In the meantime, I decided to sign up for a Linode at their Fremont, CA Hurricane Electric data center.  I was able to get the Linode setup very quickly, and had OpenVPN running the same day.  The Linode network quality to my location was so excellent, that I was really won over.  Since I didn’t really need all of the dedicated resources I had at CoreNetworks, the Linode Xen VPS did just fine (it was a Linode 360 at the time, which later became a Linode 512).  The network quality was most important to me.

After 3 days, I got an email from CoreNetworks support that said they’d identified an issue and made some adjustments to help with the network issues.  And when I checked again, it had indeed improved greatly.  But the network quality of the Linode was still better for me, and after 3 days with the Linode, I was spoiled and couldn’t go back.

So I moved everything over to the Linode and cancelled my CoreNetworks server.  After paying all the setup fees, it hurt a bit to switch, but it made sense.  If I had just been using CoreNetworks for web hosting, I might not have switched.  Since I was also using it for VPN, network quality to my location was important.

The server itself, and the friendly support staff were great.  If you need the resources that a dedicated server provides and you’re on a tight budget, I’d definately check them out.  If the network quality is sufficient for your needs, it’s a great bargain.  They provide test files on their website, so you can do some testing before you order.

Another thing to note, is that bandwidth overages could be pricey.  According to their TOS bandwidth overages are $0.50 per GB.  That would be $500 for an extra 1 TB. So if you do use CoreNetworks, you’ll probably want to keep a close eye on your bandwidth usage and make sure you don’t go over your quota.

I got a DreamHost account back in June 2007. I had hosted some small static websites in the past (a long, long time ago) with pair Networks, but this time I wanted to run WordPress.

I liked the fact that it came with huge amounts of disk space and bandwidth, though I realized CPU resources wouldn’t be unlimited. In reality, I found that I used less than 5GB of disk space, and typically less than 5GB of monthly transfer.

The plan I signed up for included 5 free domain name registrations, so it was almost like getting the hosting for $70 year ($120 – $50 for the domains). I know it’s better to keep the domain name registrations separate, but at the time, it was worth it to me because it was a good deal.

As you probably know, DreamHost uses their own custom control panel, and it worked well for me. They provided ssh access, too, and I used that often.

I was just running some small websites, using WordPress and a CakePHP app called Cheesecake Photoblog.

Overall I was happy and satisfied with the service during most of the 2 1/2 years. I know now that I probably could have done better, but I could have done worse, too. I don’t regret using DreamHost.

I did have one incident where the CakePHP app I was using was hitting their mysql server hard. A query was returning a cartesian product while trying to find related tags. They disabled the site because of that, though they left my other websites running. I ended up disabling the realted tag feature to work around the issue, and then enabled the site again and let support know I’d resolved the issue. Not a big drama, and I don’t blame them for disabling the site.

The server I was on had a pair of dual-core CPUs and 4GB RAM. They had separate servers for MySQL and email, so this was just for the webserver.

(cat /proc/cpuinfo)
 
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 15
model           : 35
model name      : Dual Core AMD Opteron(tm) Processor 175
stepping        : 2
cpu MHz         : 2194.153
cache size      : 1024 KB
 
processor       : 1
vendor_id       : AuthenticAMD
cpu family      : 15
model           : 35
model name      : Dual Core AMD Opteron(tm) Processor 175
stepping        : 2
cpu MHz         : 2194.153
cache size      : 1024 KB

(free -m)
 
             total       used       free     shared    buffers     cached
Mem:          3989       3381        608          0        353        941
-/+ buffers/cache:       2086       1902
Swap:         6165       1070       5095

I noticed the load would often get pretty high. Here are a few load averages I started taking one day after the load got really high, my sites went down, and the server got rebooted.

20:26:52 up  1:05,  2 users,  load average: 3.90, 4.03, 4.47
03:13:05 up 5 days,  7:52,  1 user,  load average: 9.91, 10.15, 10.59
04:04:57 up 5 days,  8:44,  1 user,  load average: 28.57, 26.93, 24.95
05:02:35 up 12 days,  9:43,  4 users,  load average: 19.49, 22.74, 23.84
15:00:36 up 13 days, 19:41,  5 users,  load average: 9.45, 8.01, 9.84
06:11:49 up 14 days, 10:52,  2 users,  load average: 28.77, 28.69, 28.90
00:47:29 up 16 days,  5:28,  3 users,  load average: 16.67, 20.11, 16.40
01:03:38 up 29 days,  5:45,  1 user,  load average: 24.42, 16.72, 14.82

Despite the high load, my sites seemed to function fine most of the time, at least well enough to satisfy me. After moving off DreamHost, I discovered my websites became faster, though. I didn’t know any better at the time.

Toward the end of my time with DreamHost they added a backup service, which included 50GB of backup space with rsync access. I tried to use it, but ran into problems after I accidentally overwrote some files I shouldn’t have. After that, I couldn’t get the keys for a passwordless rsync to work anymore, and tried to work with support to get my backup account reset. Unfortunately, they were never able to get it to work right after that, and then they began having other problems with their backup server that were going to take a long time to resolve.

DreamHost Status: Ongoing Hanjin (backup) server problems

It was around that time that I switched to a new service, though I kept my DreamHost account until May 2010, just before my third year with them was up. I didn’t use it much the last 4 months.

I switched to a small dedicated server, and later to a VPS. But that’s another story.

Want to see how much faster my sites were after I moved from DreamHost to a small dedicated server? Here’s a screenshot from Google Analytics for my photoblog (voxph.com). Can you tell when I made the switch?

Dynadot

I registered one domain name with Dynadot back in 2006. My only issue with them was that their whois privacy still left your name exposed, but that’s the way it was designed and I knew that when I registered. I left when I got a DreamHost account with free domain name registration.

Transfer out from Dynadot to DreamHost was smooth. Dynadot responded to the transfer request within 5 hours and the transfer was completed right away (within 5 hours).

DreamHost

For the next 3+ years, I used DreamHost. I transferred one domain from Dynadot, and registered a few more. I know what they say about keeping your domain registration and hosting separate, but I got 5 free domain names with my DreamHost account, so it was a good deal. DreamHost included free whois privacy.

The only issue I had with DreamHost domain name registration, is that their DNS control panel only works if you’re hosted with DreamHost. If you want to use it for a site hosted elsewhere, you need to change the DNS to use non-Dreamhost DNS servers (which they allow). This doesn’t seem too unreasonable since their main business is webhosting, but something to be aware of if you just want to use them as a registrar.

When I moved away from DreamHost, I transferred my domain names to Name.com. The transfer out from DreamHost to Name.com was smooth. DreamHost responded to the transfer requests within 24 hours. I don’t have a record of receiving emails from Name.com after the transfer completed, but from my memory, they were set up right away after DreamHost approved the transfer requests.

Name.com

For the next 6 months, I used Name.com. I transferred 4 domains from DreamHost, and registered one new one at Name.com.

I chose Name.com because of the good reviews and the free whois privacy they were offering at the time.

One thing to mention, at least if you’re using Name.com DNS (not sure if it applies otherwise), is that Name.com puts their own advertising page on any subdomains that aren’t defined in your DNS records. I didn’t really care for this, but setting up a wildcard DNS record stopped it.

After the Name.com whois privacy fiasco, I decided to switch registrars. I thought about it for awhile and read reviews, and decided to switch to Internet.bs.

(In case you haven’t heard, Name.com stopped giving out free whois privacy. There’s a coupon you can use to get free whois privacy now, but no telling how long it will last. I decided not to play the Name.com coupon game.)

Transfer out from Name.com to Internet.bs was not smooth. Name.com never responded to the transfer request, so I waited exactly 6 days, 4 hours, and 10 minutes for the transfer to complete.

Even one day after the transfer completed, the domains still showed up in the Name.com control panel.

I emailed Name.com support to let them know, and they responded and removed them from my Name.com account within 12 hours. They said if I had wanted the transfer to be completed sooner, I would have had to contact them so they could manually authorize the transfer out.

I had assumed since Dynadot and DreamHost could automatically respond to the transfer requests, that all registrars did. But I guess Name.com is not as automated as Dynadot and DreamHost.

Internet.bs

I just started using Internet.bs, transferring all 5 of my domains to them.

I chose Internet.bs because of the good reviews and the free whois privacy. They’re also cheaper than Name.com, though that’s not the main reason I chose them.

Internet.bs states that transfers are completed 5 days after the domains have been submitted to the registry for transfer (after the current registrant approval), if no response is received from the existing registrar. I kept track of the timing from the emails they sent me (yes, including timezone differences), and after 6 days, I emailed Internet.bs support to ask them to check into the transfer.

Even though it had already been 6 days, they were insistent that the transfers would be completed after 5 days, and that it was an automated process. Shortly after contacting them, the transfers completed, after exactly 6 days, 4 hours, and 10 minutes (to the second).

Another thing worth mentioning, is that there was no field in the whois contact details for the State/Province, so I had to include the Province in the City field.

These are minor things, though. So far I’m satisfied with Internet.bs.

Other Registrars

For completeness, I’ll mention that in the past I’ve also used Network Solutions (back when they were the only registrar), Gandi.net, and GoDaddy. I never transferred out from those registrars and no longer have any of those old domain names. Of those three registrars, I had a favorable impression of Gandi.net, and I’d probably use them again, except their prices are a little high.

Trying to stay cool on a hot day and the power is out? I got this 12V battery pack and 12V fan more than 10 years ago, and I’m still happy to have it.

I actually found a link for the old Century BPIP-99 Electri Pac model here. I remember it having 37 AH capacity, but according to the link it’s actually two 17 AH batteries (most likely 12V, so it has 34 AH total capacity). It has a built-in 300 watt inverter, but I rarely use it. These types of battery packs are fairly common, and I’m sure you can find similar newer models around.

You could also just get an SLA battery and wire a connector onto it, but it’s kind of handy to have the nice case with the connectors built-in.

I found the 12V fan at Camping World. I don’t see the same model there any more, but there are others that look just as good or better.

The 12V fan that I have is rated at 1.4 amps. I can generally use it all day or all night on the high speed setting before the battery wears out.

It’s no replacement for an air conditioner, but on a hot day it’s much better than nothing.

UPDATE (Oct. 2010): The batteries in the Electri Pac finally gave out and will no longer take a charge. They lasted longer than I thought they would. I’ll probably look for 20-40 AH SLA battery to replace it.

Do you need a UPS that can run your computer gear during all day power outages? I put together a homemade double conversion online UPS using off the shelf components, and now when the power goes out, my computer doesn’t skip a beat, even if the power is off for an hour or two… or all day.

The main components of the UPS are:

1. A pair of 6 volt golf cart batteries wired in series
2. An automatic battery charger with a DC UPS mode
3. A 300 watt sine wave inverter

With a double conversion online UPS, your equipment runs off of batteries and an inverter full time. When there is line power, an automatic charger supplies enough current to power the inverter and keep the batteries float charged. When line power is lost everything continues running normally, with the batteries handling the load. When line power is restored, the automatic charger supplies enough current to power the inverter and recharge the batteries.

This also has the side effect of providing really clean, isolated power to your computer equipment, assuming you’re using a sine wave inverter. Power is always converted to DC and back to AC before being used (double conversion).

Batteries

To get an all day run time, you need some big batteries. Most golf cart batteries are at least 200 AH, and offer a good value at the expense of some convenience. You need to check the water levels in the batteries and add water to them periodically. Also, you should not use them indoors due to outgassing. Lead acid batteries outgas hydrogen as they charge, so they should be used outdoors and need to be well ventilated to prevent explosion.

If you’re willing to spend more money (perhaps twice as much), you could look into using sealed AGM batteries, which don’t normally outgas unless overcharged. These are often used indoors, though I think I’d still prefer keeping them outside if possible.

Here you can see the pair of golf cart batteries I’m using, inside a battery box. I’m using Water Miser Vent Caps on the batteries to help reduce battery watering and to make watering more convenient with flip-tops. They’re optional, though.

I’ve heard Trojan makes great golf cart batteries. I’m using Motolite golf cart batteries, which is what I could find locally, and they work fine. They’ve been in use 3+ years now and still going strong.

Desulfator

I’m also using a Power Pulse desulfator connected to the batteries. This is optional, but I think its really helped. Over time, sulphate will form on the lead plates in the batteries causing reduced capacity. The desulfator helps prevent and even reverse the sulfation. It’s not perfect and the batteries won’t last forever, but it helps.

Before I hooked the desulfator up to the golf cart batteries, I hooked it up to an old 37 AH SLA battery pack that had been in use for about 10 years and was starting to show substantial wear. I kept the Power Pulse desulfator on the battery pack along with a 500 mAH charger for a few weeks, putting it through a few charge/discharge cycles and it really helped improve the condition of the old battery pack. I wish I’d gotten a second Power Pulse desulfator to use on my car battery.

Battery Connections

The golf cart batteries are 6V so to get 12V for the UPS, a pair of them are wired in series. I’m using a 65A automotive fuse between the negative post of the first battery and the positive post of the second battery. It’s really important for safety to make sure everything is fused properly.

I’m using heavy gauge wire for the battery connections. On the batteries themselves, I’m using some thick yellow automotive wire (labeled with a marker +/-). For the inverter and charger connections, I’m also using heavy gauge (8mm²) wire, but it’s properly color coded (black & red). If I was doing it over again, I’d use the color coded wire for the battery connections, too.

The batteries are wired into an ordinary 50A 240V AC circuit breaker mounted on the outside of the battery box. Ideally, this would be a DC circuit breaker. AC and DC circuit breakers are designed a bit differently, and there are situations where using an AC circuit breaker with a DC load could be problematic. Since the batteries themselves are fused separately, I’m mainly using this as a disconnect switch for convenience. It seems to be working fine for me so far.

From the circuit breaker, connections are made to both the automatic battery charger mounted on the front of the battery box, and the sine wave inverter mounted on the side of the battery box underneath the circuit breaker.

Automatic Battery Charger

I’m using a Samlex SEC-1230A 12 Volt 30 Amp Automatic Battery Charger. I have the DIP switch settings set to “Battery with Load” so the charger effectively becomes a DC UPS. There’s a good manual that explains the settings, and it’s available from the Samlex website as a PDF.

The charger comes wired for 120V 60Hz input, but there’s an internal jumper that can be changed to make it 230V 50Hz. If you do that, you need to change the fuse as well. I made the change, and use it with a 220V 60Hz power source and so far I haven’t had any problems.

Sine Wave Inverter

For the inverter, I’m using a Samlex PST-30S-12A 300 Watt Pure Sine Wave Inverter (12VDC-120VAC).  I’ve used a modified sine wave inverter with my computer equipment in the past, and it mostly worked fine, but there was some buzzing on my phones. There’s no buzzing with the sine wave inverter.

Some of my gear uses 110V, some uses 220V, and some will use either.  For the stuff that needs 220V, I use a small 100 watt step-up transformer.  I also run my computer (an HP Laptop) off of 12 VDC with an HP KS474AA 90W Auto/Air/AC Adapter.  Other laptops may have similar 12 VDC adapters available.

You’ll need to calculate your load requirements.  If you’re using a desktop system instead of a laptop, you may need a larger inverter. My inverter is supplying power for a 22″ LCD monitor and an assortment of peripherals:  a pair of USB hard drives, speakers, DSL modem, router, switch, USB hub, ATA, and 3 phones.

I’ve got a power meter hooked up to the inverter output, and I’m using less than 100 watts. So for my use, a 300 watt inverter is plenty. The inverter could handle my laptop, too. I already had the 12 VDC laptop adapter, though, so I use it for maximum efficiency.

Ventilation

I keep the custom built battery box on an outside covered patio that gets plenty of ventilation and is conveniently located next to my office. There are plastic screens on the top and the sides of the box to let the batteries vent.

Connections

I’ve got a set of connectors mounted in the wall in utility boxes to keep things clean. The outdoor connectors (from left to right) are: antenna, ground, line power, UPS power inlet, and 12 VDC input posts.

There’s another set of connectors mounted on the other side of the wall in my office. The indoor connectors (from left to right) are: two 12 VDC outlets, two 110 VAC UPS outlets, line power inlet, ground, and antenna. The antenna connector isn’t used for the UPS.

Cost

The rough cost for the major components of the system was:

1. $250 Two 6 volt golf cart batteries
2. $225 Samlex SEC-1230A Battery Charger
3. $135 Samlex PST-30S-12A Sine Wave Inverter

I got the golf cart batteries locally, and ordered the charger and inverter from altEstore.

An alternative to all this, is to use a small off the shelf UPS for short power outages, and to simply take a break from the computer during longer power outages. I hear its nice to spend time outdoors. What do you do when the power goes out?

Please note: I’m not an electrician, but I did have an electrician do the wiring for this UPS system for me. There’s no guarantee any of this meets electrical code. Please consult an electrician and check your local electrical code if you want to undertake a project like this.