Intel DC S3500 vs S3700 as a ZIL SSD Device Benchmarks

I ran some benchmarks comparing the Intel DC S3500 vs the Intel DC S3700 when being used as a SLOG/ZIL (ZFS Intent Log).  Both SSDs are what Intel considers datacenter class and they are both very reasonably priced compared to some of the other enterprise class offerings.


SLOG Requirements

Since flushing the cache to spindles is slow, ZFS uses the ZIL to safely commit random writes.  The ZIL is never read from except in the case power is lost before “flushed” writes in memory have been written to the pool.  So to make a decent SLOG/ZIL the SSD must be able to sustain a power loss in the middle of a write without losing or corrupting data.  The ZIL translates random writes to sequential writes so it must be able to sustain fairly high throughput.  I don’t think random write IOPS is as important but I’m sure it helps some. Generally a larger SSD is better because they tend to offer more throughput.  I don’t have an unlimited budget so I got the 80GB S3500 and the 100GB S3700 but if you’re planning for some serious performance you may want to use a larger model, maybe around 200GB or even 400GB.

Specifications of SSDs Tested

Intel DC S3500 80GB

  • Seq Read/Write: 340MBps/100MBs
  • 70,000 / 7,000 4K Read/Write IOPS
  • Endurance Rating: 45TB written
  • Price: $113 at Amazon

Intel DC S3700 100GB

  • Seq Read/Write: 500MBs/200MBs
  • 75,000 / 19,000 4K Read/Write IOPS
  • Endurance Rating: 1.83PB written (that is a lot of endurance).
  • Price: $203 at Amazon

Build Specifications

Virtual NAS Configuration

  • FreeNAS VM with 6GB memory and 2 cores.  Using VMXNET3 network driver.
  • RAID-Z is from VMDKs on 3×7200 Seagates.
  • SLOG/ZIL device is a 16GB vmdk on the tested SSD.
  • NFS dataset on pool is shared back to VMware.  For more information on this concept see Napp-in-one.
  • LZ4 Compression enabled on the pool.
  • Encryption On
  • Deduplication Off
  • Atime=off
  • Sync=Standard (VMware requests a cache flush after every write so this is a very safe configuration).

Don’t try this at home: I should note that FreeNAS is not supported running as a VM guest, and as a general rule running ZFS off of VMDKs is discouraged.  OmniOS would be better supported as a VM guest as long as the HBA is passed to the guest using VMDirectIO.  The Avoton processor doesn’t support VT-d which is why I didn’t try to benchmark in that configuration.

Benchmarked Guest VM Configuration

  • Benchmark vmdk is installed on the NFS datastore.  A VM is installed on that vmdk running Ubuntu 14.04 LTS, 2 cores, 1GB memory.  Para-virtual storage.
  • OLTP tests run against MariaDB Server 5.5.37 (fork from MySQL).

I wiped out the zpool after every configuration change, and ran each test three times for each configuration and took the average (in almost all cases subsequent tests ran faster because the ZFS ARC was caching reads into memory so I was very careful to run the tests in the same order on each configuration.  If I made a mistake I rebooted to clear the ARC).  I am mostly concerned with testing random write performance so these benchmarks are more concerned with write IOPS than with throughput.

Benchmark Commands

Random Read/Write:
# sysbench –test=fileio –file-total-size=6G –file-test-mode=rndrw –max-time=300 run

Random Write:
# sysbench –test=fileio –file-total-size=6G –file-test-mode=rndwr –max-time=300 run

OLTP 2 threads:
# sysbench –test=oltp –oltp-table-size=1000000 –mysql-db=sbtest –mysql-user=root –mysql-password=test –num-threads=2 –max-time=60 run

OLTP 4 threads:
# sysbench –test=oltp –oltp-table-size=1000000 –mysql-db=sbtest –mysql-user=root –mysql-password=test –num-threads=4 –max-time=60 run

Test Results

zil_random_read_write zil_random_write zil_oltp_4_theads zil_oltp_2_threads


none OLTP 2 Threads 151
Intel DC 3500 80GB OLTP 2 Threads 188
Intel DC 3700 100GB OLTP 2 Threads 189
none OLTP 4 Threads 207
Intel DC 3500 80GB OLTP 4 Threads 271
Intel DC 3700 100GB OLTP 4 Threads 266
none RNDRW 613
Intel DC 3500 80GB RNDRW 1120
Intel DC 3700 100GB RNDRW 1166
none RNDWR 273
Intel DC 3500 80GB RNDWR 588
Intel DC 3700 100GB RNDWR 569


Surprisingly the Intel DC S3700 didn’t offer much of an advantage over the DC S3500.  Real life workload results may vary but the Intel DC S3500 is probably the best performance per dollar for a SLOG device unless you’re concerned about write endurance in which case you’ll want to use the DC S3700.

Other Observations

There are a few SSDs with power loss protection that would also work.  The Seagate 600 Pro SSD, and also for light workloads a consumer SSDs like the Crucial M500 and the Crucial MX100 would be decent candidates and still provide an advantage over running without a SLOG.

I ran a few tests comparing the VMXNET3 vs E1000 network adapter and there is a performance penalty for the E1000.  This test was against the DC 3500.


Network Test TPS Avg
E1000g OLTP 2 Threads 187
VMXNET3 OLTP 2 Threads 188
E1000g OLTP 4 Threads 262
VMXNET3 OLTP 4 Threads 271
E1000g RNDRW 1101
E1000g RNDWR 564

I ran a few tests with Encryption on and off and found a small performance penalty for encryption.  This test was against the DC S3700.


Encryption Test TPS Avg
Off OLTP 2 Threads 195
On OLTP 2 Threads 189
Off OLTP 4 Threads 270
On OLTP 4 Threads 266
Off RNDRW 1209
On RNDRW 1166
Off RNDWR 609
On RNDWR 569

Total Cost of Ownership for the Honda-CRV and Subaru-Outback

It’s that time a year again to report on the annual costs.  I was a little behind this year so “2014″ is from June 2, 2013 to June 30, 2014.  In October of 2013 we acquired a 2003 Subaru Outback so I’ve started to track that as well.


So far it’s looking like an average cost around $0.50/mile for both cars.  Keep in mind some costs like Insurance and Depreciation and DMV fees are a factor of time as much as they are miles–but the benefit we get out of the cars is in miles so I think that’s the best way to look at the cost.


2003 subaru outback cost by mile



Purple Screen of Death

Well, my ASRock C2750D4I Build has not been doing so hot…

I woke-up one morning to find VMware crashed with a PSOD (Purple Screen of Death)… and then the next day it crashed again.  Over the next few days I had several crashes, often after only an hour of uptime.  Here’s a sampling of the errors I’d see:


PCPU1 Locked up.  Failed to ack TLB invalidate…


Uh oh, now it’s PCPU3!


Machine Check Exception: Fatal (unrecoverable) MCE on PCPU0 in world 36026 …looks like it got the Minecraft server that time.


And my favorite!  The recursive panic!  Recursive panic on same CPU (cpu 1, world 40471).

Machine Check Exception when the CPU is already in Panic: CPU1.


PCPU 0 locked up.  Failed to ack TLB invalidate.


#PF Exception 14 in world 35467:wmm3 …bummer.


Machine Check Exception: Fatal (unrecoverable) MCE on PCPU1 in world 35089.

I thought it might have been the ram so I swapped it out but got the same problem.  I emailed ASRock support and an automated email suggested I contact my dealer, so I’ve started an RMA process with SuperBiiz.

In the meantime I needed to get back up and running. I have an extra E3-1240-V3 so I thought I’d switch over to an 1150 socket so I bought a SuperMicro X10SL7-F-0 which is a great little board with a built in LSI-2308 which can be flashed into IT mode for ZFS… only in my haste I didn’t realize it wasn’t a Mini-ITX board.  I guess SuperMicro doesn’t even make a Mini-ITX board with an E3 socket.  So I ordered an ASRock E3C224D2I which seems like a great board, but for some reason I couldn’t get it to even post.  Had I been able to get it to work I would have had to use the IBM ServerRaid M1015 HBA to get 8 drives, but doing so would have cost me a hotswap bay or I would have had to use a riser and somehow stuff it in the case.  I think I exhausted the Mini-ITX options (SuperMicro has some, but they’ve switched to SO-DIMM on their Mini-ITX form factor and I don’t have any).   So I went back to the Supermicro X10SL7 board… but it doesn’t fit in my case…. so I’ve got sort of a makeshift setup while I look for a real server chassis… (don’t try this at home):

At least I’ve got two Ethernet connections for failover!


I think I’m going to give up on my Mini-ITX Small form factor SilverStone D380 case and stick with a Micro-ATX or ATX because there’s more options for motherboards in that size.

BaoFeng UV-5RA, UV-B5, UV-82, and Yaesu VX-6R Comparison

I’ve been very happy with the BaoFeng UV-5RA, but BaoFeng has released some newer models on the market so I thought it was time to take a look at the current offerings: of the new models it looked like the best new models were the UV-B5 (or nearly identical UV-B6), and UV-82, so here they are:


UV-5RA, UV-B5, UV-82, VX-6R

Amazon Price
TX Power (claimed)4W, 1W5W, 2W5W, 1W5W, 2.5W, 1W, 05W
Receive Range (MHz)65-108, 136-174, 400-48065-108, 136-174, 400-47065-108, 136-174, 400-4500.5-999
Transmit Range136-174, 400-480 136-174, 400-470 136-174, 400-520144-148, 222-225, 430-450
Easy to program by handNoNoNoYes
Scan SpeedSlowMediumSlowFast
ReceiverWide, easily overloadedMuch better than other BaoFengsWide, easily overloaded.Great.
Antenna LengthShortMediumMediumMedium
Annoying AlarmYesYesNoNo
Individual band PTT buttonsNoNoYesNo
Steps (KHz)2.5, 5, 6.25, 10, 12.5, 20, 25, 50 (has 2.5 KHz steps)5, 6.25, 10, 12.5, 20, 252.5, 5, 6.25, 10, 12.5, 20, 25, 50 (has 2.5KHz steps)5, 10K, 12.5 15, 20, 25, 50, 100
Memory Banks11124
Memory Channels12799127900
Alpha Tags7-characters
Dual DisplayYesYesYesNo
Dot Matrix DisplayYesNoYesNo
Channel Encoder KnobNoYesNoYes
Dual PPT buttonsNoNoYesNo
VFO / Memory buttonYesYesNo (must hold menu button while turning on to switch)Yes
ARTS TransponderNoNoNoYes
Battery1800 mAh2000 mAh3000 mAh1500 mAh
SizeSmallSmallSmall but a little tallerSmall but a little wider
Shape / Hand FitSquareRounded (easy to hold)Rounded (great to hold).Square but rounded corners, wider so easier to hold.
Backlight ColorsPurple, Blue (RX), Red (TX)WhitePurple, Blue (RX), Red (TX)Orange
FCC Part 90 CertifiedYesNoNo (it does have a certification but mine did not come with an FCC label)No
Legal Uses in the United StatesHam Radio or FCC LicenseHam Radio onlyHam Radio onlyHam Radio only
Companding (see: Wikipedia article)NoYesNoNo
Real S-MeterNoYesNoYes

Scan Speed Compared.  You can see the UV-5RA and UV-82 are identically slow, the UV-B5 is a little faster, and the Yaesu VX-6R smokes them all.

A few thoughts about each radio:

The UV-5RA is still my favorite among the BaoFeng’s.  I think it’s the most versatile, it has 2.5KHz steps, a small size, lots of accessories, FCC part 90 certified (so this can be used on Business Bands with an FCC license).  My main complaint with this (and all BaoFeng’s) is the slow scan and difficulty programming by hand.   (As far as I can tell the UV-5RA, UV-5RB, UV-5RE, UV-5R, BF-F8+, UV-5R+, etc. are the same radio other than minor cosmetic and firmware differences).

The UV-B5 is the best radio I’ve seen from BaoFeng.  If I could only have one radio and only had $30 this is probably what I’d get.  It is shaped perfect and easy to hold.  This has the best receiver of the BaoFengs.  I tested the receivers by transmitting slightly off frequency and it would open up the other two BaoFengs but the UV-B5 would still stay quiet.  Also, there have been several times the UV-B5 will pickup a weak signal the other two BaoFeng’s didn’t.  Scanning is also faster than the UV-5RA and UV-82.  The channel encoder knob makes changing channels or settings so much faster!   Also this is the only radio that allows you to operate in VFO mode on one display and memory mode on the other. It’s only shortcomings are that it’s not FCC certified so it can’t be used on business bands, and the channel steps don’t go to 2.5KHz (which really is not a big deal, I don’t think I’ve need to be that precise for awhile).  If it was FCC certified it would take the place of my favorite from the UV-5RA.  (Note: UV-B6 is just like the UV-B5, except the 6 has a flashlight instead of the channel encoder knob).

Tbe UV-82 is the easiest to hold of the bunch because it is larger.  Other than that I can’t say it differs much from the UV-5RA.  It is a little bigger, has a much bigger battery, a slightly better light.  Changing from VFO mode to Memory mode is done by pressing menu while turning the radio on.  This radio is supposed to be FCC part 90 certified but mine didn’t come with an FCC label. (There are some variants of this radio like the UV-82X which are 2 meter / 1.25 CM instead of 2M / 70CM).

The VX-6R is still the radio I take with me hiking.  The difference between the Yaesu and the BaoFengs is not small.  The Yaesu is much higher quality, and it’s designed to be an Amateur radio which means it’s better at scanning and programming by hand.  The Yaesu VX-6R is not even in the same class as the BaoFengs but I included it in the comparison because it’s my favorite radio and I think most hams are going to enjoy something like this over a BaoFeng.

IMG_8804 IMG_8807 IMG_8806

I was able to use CHIRP to program all of these radios and they all worked the same way and all worked fine with the same USB Programming Cable. It was quick to copy and paste the memories from one radio image to the other inside CHIRP and upload to the new radios.

The BaoFeng’s have great versatility because they can transmit on any frequency such as  GMRS/FRS/Marine/MURS/business frequencies, etc.  However, outside of an emergency it is illegal to do so.  I don’t agree with the FCC here, I think the FCC should allow non part certified radios on those frequencies but the rules are the rules.  I’ve posted a Legal Frequencies Q&A on my UV-5RA review page.


Best Quad Core Processors for SQL Server 2014 Enterprise

SQL Server 2014 Enterprise Licensing Efficiency on Four Cores

Microsoft Core Licensing

Microsoft pricing per core is $13,748 (your price may vary depending on your licensing program) per set of two cores, with a four core minimum per processor (a dual core counts as a four core so it doesn’t make sense to go with less than 2 cores) so you’re looking at $27,500 in licensing costs.  It doesn’t matter how fast those cores are, you’re paying the same price if you run this on a Xeon E3-1220L v3 as you would on the most expensive processor money can buy… at a licensing cost of $6900/core you want to make every core count.

Quad Core Options

Most processors can quickly be eliminated.  What we’re looking for is support for as much memory as possible, a large L2 and L3 cache (see: Microsoft SQL 2012 Management and Administration by Ross Mistry pp. 685), and since licensing is core bound we want to get as high a frequency as possible to minimize the number of cores.  Xeon E7s don’t come with a quad core option.  The Xeon E3 has a memory limit of 32GB and also has a rather small 8MB L2 cache and no L3 cache so it won’t be the best option for most scenarios.  This leaves the Xeon E5 series…

Xeon E5 Processor Options

  • Xeon E5-4603 v2 (10M Cache, 2.20 GHz): $551.00 — too slow
  • Xeon E5-2407 v2 (10M Cache, 2.40 GHz): $250.00 — too slow
  • Xeon E5-2403 v2 (10M Cache, 1.80 GHz): $192.00 — too slow
  • Xeon E5-1620 v2 (10M Cache, 3.70 GHz): $294,00 — low cache and memory
  • Xeon E5-2609 v2 (10M Cache, 2.50 GHz): $294.00 — too slow
  • Xeon E5-2637 v2 (15M Cache, 3.50 GHz): $996,00 – winner
  • Xeon E5-2603 v2 (10M Cache, 1.80 GHz): $202.00 — too slow

It’s very easy to eliminate all but two… if you’re wondering why we don’t even consider the slower processors consider the E5-2637 v2 and compare it to the next fastest processor (E5-2609) and notice it’s 30% slower.  Sure you might save $800 on the CPU but you’re wasting 10 times that amount in the value of your licensing!   Microsoft isn’t going to give you a discount on a slower core so you may as well get the most performance you can get out of it.

After eliminating the slowest processors the two choices left are:

  • E5-1620 v2 – no SMP support, max memory 256GB, runs at a slightly higher clock.
  • E5-2637 v2 – Dual Processor option, larger 15MB cache, max memory 384GB,

Most Performance per Dollar: Xeon E5-2637 v2

Sandy-Bridge_EP_1Even though the E5-1620 v2 is clocked faster, that’s more than offset by the larger cache. This is especially the case with hyper-threading.  The extra 128GB of memory (make sure you’re motherboard has 12-dimm slots per CPU to get to 384GB) capacity is going to stretch licensing a lot further.

Also, having the dual processor option gives you flexibility to drop in another CPU on the motherboard in the future without having to acquire another server.  This is nice insurance against unexpected growth.

So, of the available quad core processors options the E5-2637 v2 is the most efficient processor from a performance per dollar and capacity per dollar for SQL Server Enterprise.

If you need more than four cores take a look at Glenn Berry’s articles: Selecting a Processor for SQL Server 2014 – Part 1 and Selecting a Processor for SQL Server 2014 – Part 2.


GPU Mining with NVIDIA GeForce GTX 750 Ti Overclocked

Energy Efficient and Quiet

Trying to cut down on the noise and power usage of my old AMD Radeon 6870 (which uses about 130 Watts idle and 250 under load) I upgraded my GPU to the new NVIDIA GeForce GTX 750 Ti (~$150 depending on model), I decided on MSI’s version which has a massive heatsync and dual fans …and it is dead silent.  Even putting a full load on the GPU I can’t hear the fans at all.  I also noticed the EVGA model (which may be slightly louder but still very quiet) is shorter so it would work in smaller cases.

Although I haven’t verified it at idle power consumption is supposed to be around 9 Watts driving two monitors, and 60 Watts under load.  Since I leave my computer on 24/7 that should cut down my energy costs by $80/year so the card could pay for itself in two years… but wait, there’s more!

Crypto-Currency Mining

600px-Bitcoin_euroOne thing in particular about this card is it is very low wattage which makes if efficient for mining crypto-currency like Bitcoins… which I’ve been wanting to try but haven’t because until now my GPU has had too loud of a fan when it’s under load.  The GTX 750 Ti is probably the best GPU for mining considering the Hashrate per Watt ratio.

To see what sort of performance I’d get I test mined a few currencies to get the hash-rates for various algorithms.  Some algorithms in use today are SHA-256 (Bitcoin), Scrypt (Litecoin), Scrypt-N (Vertcoin), X11 (Darkcoin), Kekkak (Maxcoin), Quark (QRK).  SHA-256 is what Bitcoin uses and it’s pretty pointless to mine coins with an SHA-256 algorithm because of the ASICs out there that provide an incredible hashrate compared to what a GPU can provide.  I think the best algorithms for a GPU to mine are Scrypt and Scrypt-N.  Scrypt ASICs are coming on the market so coins using Scrypt-N algorithms will probably be the best ones to mine in the future.

Hashrates of the GTX 750 Ti

  • Scrypt: 250 KH/s
  • Scrypt-N: 125 KH/s
  • Keccak: 2.5 MH/s

I found a few websites that show the most profitable coin to mine based on current exchange rates, in particular CoinWarz.  My cost per kWh is $0.0747 so I put those values and the hash rates into CoinWarz’s Crypto Currency Mining Profitability calculator and without trying to always mine the top coin it looks like it’s fairly easy to mine $0.50/day (or $15/month) after taking energy into account.  This may not seem like much, but consider the GPU costs around $150.00 that’s a 120% annual return (now this will likely go down as mining difficulty increases).  Unfortunately I think CoinWarz takes a little too long to update so I think what it shows you is what the best coin to mine would have been, but if you see a few coins that are consistently profitable you can target those.

My first mining experience…

(okay, so to be honest, this is not my first mining experience, I actually tried mining a few BitCcins several years ago but quickly discovered the energy and noise of the GPU fan wasn’t worth it).

My second mining experience… how to get started.

Pick a Coin

Execoin-64x64You can pick any coin you want.  I am most interested in coins with the Scrypt-N algorithm and they appear to have just been created this year so they’re relatively easy to mine.  So I started with EXECoin.  From the the EXECoin download page I got their Light Wallet (ExeLite) and their CUDAMiner (CUDAMiner is the program you want when mining with an Nvidia GPU).

Download the Wallet and Get a Receiving Address

Step One.  Open the Wallet and get a Receiving Address.  The receiving address is safe to give out, there’s a corresponding private key that you don’t want to give out.


One neat thing about this wallet is the ability to create a paper wallet and print it out for offline cold storage without creating it on your computer.  You can continue receiving coins at that address but you won’t be able to access those coins until you import the private key back into EXELite by typing it in or using your webcam to scan the QR code (just don’t ever lose that paper).


Here’s what a paper wallet looks like. Note: I am never going to use this address because part of the private key is shown.  You never want to show anyone your private key, it’s like leaving your debit card with your PIN number written on it right outside an ATM machine with a sign that says, “free money!”

Pick a Mining Pool


It’s possible to mine solo, but with a lot of currencies it can take days, weeks, and even longer to actually find coins.  What a mining pool does is allow several miners to pool their resources together and mine as a single unit.  When the pool finds coins it splits it up evenly  among the miners (usually after taking a small fee between 0%-2% which I’m sure doesn’t even cover the cost of running the pool).  Pools typically pay out daily but some pay out more often.

Most coins have a list of mining pools you can use, here’s EXECoin’s List of Pools.  The two things you want to consider when picking a pool are latency and node fees (I typically try to find the lowest fee pool with under 100ms).  I thought the P2Pools looked interesting so I did a quick ping and found the lowest latency for me was

Download Cudaminer (for NVIDIA)

After downloading and unzipping Cudaminer here’s the command I ran:

(the password doesn’t matter on most P2Pools)


(you’ll obviously want to change the receiving address to your own unless you want to donate some mining time to me).

The N factor will increase over time.  For example, on July 2015 you’ll need to change the N factor from 2048 to 4096.  This frequent change in the N factor is what makes EXECoin ASIC resistant which will help hold the value of your coins.

Receive Your Coins

Your address receives your coin, and you don’t need to be online.  At anytime you can view your EXECoin transactions at  Of course, without the private key which is stored in your wallet (on your computer or on paper if you made a paper wallet) you can’t spend your coins.  It’s a great idea to backup your wallet, and important to keep any media that you store the wallet on encrypted.

Other Coins

There are quite a few other coins, for now I’m sticking with coins that use Scrypt-N because I like having some ASIC resistance.  I also started to mine some Vertcoin, the process was pretty much the same.  Download the wallet to get a receiving address, find a pool, and run Cudaminer against that pool with my Vertcoin receiving address. The Cudaminer from EXECoin’s website works well to mine Vertcoin or just about any other coin, you’ll just need to change the algorithm if you’re not mining Scrypt-N.

Most Profitable Coins

Mining these ALT-coins is actually more profitable than mining Bitcoin.  In fact, you can mine ALT-coins and trade them in for Bitcoins for cheaper than you’ll mine a Bitcoin.  I’m not sure why this anomaly exists except perhaps that there is the expectation that some of these newer coins will increase in value more rapidly than Bitcoins so Bitcoin holders are willing to pay a premium for them.

There are actually mining pools that automatically switch to the most profitable coin based on their BTC (Bitcoin) exchange rate.  I looked around at a few and I think CoinSolver looks like a pretty good one, it has a Scrypt pool and is one of the few that also has a Scrypt-N pool.  I think it’s a little more up to date than CoinWarz for the most profitable coin and it mines whatever coin has the best exchange rate against BTC and automatically pays you out in BTC (so you’ll need to download the Bitcoin wallet to use it).  Once you get enough BTC–maybe after a year or so of mining you could exchange it to USD if you don’t want to hold Bitcoins.  Another strategy is to mine the most profitable coin and Exchange it into BTC, then buy other alt-coins.  I wouldn’t call this investing, this is mere speculation like buying Gold or Silver or trading other currencies.  But it can be fun.  I’d take speculating mining power on crypto-currency over gambling at a casino any day.  |:-)

* * *

Overall, I’m pretty happy with the Nvidia GTX 750 Ti.  I don’t think I’d go out and buy a GPU just for mining but if I’m going to have one anyway I may as well put it to work.  And If you want my old GPU it will be on Ebay shortly.

And of course, if you want to donate some coins to me:

EXECoin: ELFXwvb3ikLqpXC9eUGsB4Zpdyr5ZpKaA3
Vertcoin: VgdMcbLG1jausNBeEym7Kr2kuuSQbyD5Mf
Bitcoin: 1H7ycbjt6EM4Uq9wbR5qrtJaiHTupjU2DG


Moved to DigitalOcean

DO_Logo_Vertical_Black-1f85f3b9Yesterday I moved this site to DigitalOcean (referral link).  I had been running on Verzion FiOS at Jeff’s house, but I think that little N54L was too overloaded.  So I signed up for the cheapest DigitalOcean plan at $5/month and setup an Ubuntu 14.04 LTS server with Nginx.  I put it in Google’s NYC data center (Speed Test) which I think is positioned well to provide decent peering internationally.  So I hope page loads are looking faster for everyone, especially those outside the U.S.